Abstract
Introduction
The new forms of work organization and the progress of information and communication technologies have been considered a dynamizer for the transformation of human life in the organizational environment, gradually moving away from the use of rigid and hierarchical structures for an organizational system based on initiative and cooperation (Andrade, 2017). In this sense, People Management in organizations needs to increasingly analyze a growing list of different variables in the work environment, aiming at an action that provides competitive advantage to organizations, allied to a good working environment for employees.
To this end, discretionary actions and practices of employees in the work environment, not listed in contracts, positions and functions in the company, also known as Organizational Citizenship Behaviors, play an important role in achieving these results (Gouvea et al., 2020).
Organizational Citizenship Behaviors (OCB) can be considered an important measure of Organizational Behavior and essential in issues related to employee performance (N. Podsakoff et al., 2014). The concept of Organizational Citizenship Behavior (OCB) is based on the concepts of willingness to cooperate (Barnard, 1938) and innovative and spontaneous behaviors (Katz & Kahn, 1978), and was initially defined by Organ (1988, p. 4) as a “discretionary individual behavior, not directly or explicitly recognized by the formal reward system and which as a whole promotes the effective functioning of the organization” and later redefined as “activities that support the social and psychological environment in which the performance of tasks has a place” (Organ, 1997, p. 95).
Organ (1988) spread out the idea related to the need for organizations to resort to the spontaneous, innovative and cooperative behaviors of their members, so that organizations can obtain answers to constantly changing demands, thus obtaining greater efficiency and competitive advantage.
In the same way, Siqueira (2003) states that OCB are informal actions of workers that benefit the organization, through acts of social exchange, voluntarily promoted by workers and offered to organizations spontaneously, exempt from legal or contractual requirements. A discretionary organizational citizenship means a personal choice that cannot be imposed as an obligation, nor induced by the guarantee of a formal reward. As well as the absence of this behavior, it should not be understood as punishable.
From this perspective, it is believed that organizations should resort to the spontaneous, innovative and cooperative behavior of their members, so that the responses to the demands of a context in constant change are effective, thus generating a competitive advantage (Organ, 1997).
According to V. F. Costa et al. (2017), in the corporate environment found in the current economic scenario, whether private or public, organizations increasingly need professionals with OCB, that is, individuals with initiative and proactivity who are not restricted only to prescribed tasks in the formal system. The OCB occupy a strategic space in achieving the most complex goals of organizations as they have the ability to inspire co-workers to voluntarily share the tacit knowledge, which was obtained by each individual in the execution of tasks in the organization (Neves & Cerdeira, 2018). The OCB can promote an enrichment in human relations at work, as well as in the relations between people and companies, resulting in gains for both (Gouvea et al., 2020).
According to Eisenberger et al. (2001), when analyzing citizenship behavior from an exchange perspective, attention is focused on the exchange between the organization and the individual. This theory suggests that OCB is more likely to occur when the employee is satisfied with the organization and feels motivated to respond reciprocally.
The Theory of Social Exchanges indicates that, when they perceive support from the organization and colleagues, individuals propose something in return, reciprocating with behaviors valued by the organization and that also benefit their co-workers (Swift & Virick, 2013). This interaction between the worker and the organization is characterized by exchange relations influenced by expectations of reciprocity (Eisenberger et al., 1986).
The norm of reciprocity proposed by Gouldner (1960) is based on the idea that individuals will exhibit positive behaviors in exchange for their benefits, which can result in perceptions of psychological contract by employees and influence spontaneous behaviors bringing benefits to the organization.
Despite presenting numerous benefits for the organizational environment, evaluating OCBr is a complex task, and there is still no consensus on its measurement. Nationally, the contributions of Siqueira (1995, 2003) and Porto and Tamayo (2003) stand out. Siqueira (1995) is considered the pioneer to study the subject in the country, when developing the first Brazilian scale aimed at measuring OCB. The scale developed by Siqueira (1995) contains 18 items and five factors, using the nomenclature suggested by Katz and Kahn (1978): creation of a favorable climate for the organization in the external environment, creative suggestions for the system, protection of the system, self-training and cooperation with co-workers.
In the study proposed by Porto and Tamayo (2003), based on the SOCB of Siqueira (1995), the authors developed the Civism Scale in Organizations (CSO), consisting of 41 items and five factors. The use of the term Civism did not interfere with the creation of the factors, since the dimensions of the CSO are consonant with the OCB construct, using the same categories proposed by Katz and Kahn (1978), Siqueira (1995), and George and Brief (1992).
Considering the transformations that the nature of work has undergone, Dekas (2010) and Dekas et al. (2013) found that some dimensions, such as obedience, might not be appropriate for the social context of knowledge workers. Thus, the authors proposed the OCB-KW scale, which includes the dimensions of employee sustainability and social participation, representing an advance in studies on OCB in relation to changes in the world of work and in the profile of knowledge workers. This measure was adapted for the Brazilian context by Andrade (2017), being applied with workers in the private educational sector.
Inspired by studies already carried out using the Organizational Citizenship Behavior—Knowledge Workers (OCB-KW) Scale and the literature on OCB, this study aimed to analyze the empirical implications derived from the use of the Organizational Citizenship Behavior Scale in the context of Knowledge Workers through additional tests not previously performed with the samples obtained in the different empirical studies in which the scale was used.
Based on these tests, it is intended to confirm or refute hypothesis 1 of this study:
The justification for this study resides in the fact that studies carried out with the translated version of the scale found different results regarding the dimensions and items excluded from the scale. With different and more robust tests, we intend to understand the behavior of the translated scale in more depth. Through the results of the tests that will be carried out in this study, we seek to expand the discussions on Organizational Citizenship Behaviors, presenting a comparative analysis of the applications of the Organizational Citizenship Behavior—KW scale. In addition, this study sought to contribute to the process of construction, adaptation and validation of the OCB-KW scale, presenting suggestions for improving the scale, as studies (Andrade, 2017; G. L. Á. Costa, 2019; Malheiros, 2021; Parcianello, 2022; Rebolho, 2018) developed using this scale in different contexts indicated some divergences in relation to the proposal of Dekas (2010) and Dekas et al. (2013).
The understanding and evaluation of OCB depends on the existence of a valid and consistent measurement model. Thus, the search of one or more efficient ways to assess OCB can contribute to the development of research in the area. In addition, research on the behavior of individuals can support and assist in decision-making by managers in companies, in the most diverse organizational spheres, thus providing relevant information for strategic business purposes, which can be used in organizations, in order to explain the differences in individual and group performances (Gouvea et al., 2020).
Method
The aim of this study was to analyze the empirical instabilities and adopt new validation procedures for the Organizational Citizenship Behavior—Knowledge Workers (OCB-KW) scale. For this purpose, we decided to divide the analyses into two stages.
In the first stage, the performance of the OCB-KW scale was analyzed in the different empirical studies in which the scale was used. To identify the sample at this stage, a survey was carried out of the articles that cited the work of Dekas et al. (2013)“Organizational Citizenship Behavior, Version 2.0: A Review and Qualitative Investigation of OCBs for Knowledge Workers at Google and Beyond” in the Scopus and Web of Science databases.
Only articles published from 2013 to 2023 were used, the searches were carried out in January 2023. Fifty-three documents were found in the Web of Science and 69 documents in Scopus. The authors performed the reading of all the works to identify which used the OCB-KW scale as a research tool and which only cited the work of Dekas et al. (2013), with two studies found. In addition to the research in Scopus and the Web of Science, an academic google search was carried out to check for theses or dissertations using the scale. The search term used was “Scale of Organizational Citizenship Behaviors—Knowledge Workers”Dekas et al. (2013) and three dissertations, two thesis and an article were found, totaling a sample of six works.
In the second stage, aiming to adopt new estimation procedures to identify the best structure for the OCB-KW scale, we used a sample set consisting of three databases already used in previous studies with knowledge workers, characterized as a secondary database. In these studies, the answers were obtained through an online and printed questionnaire distributed to knowledge workers. Thus, a database was built, with a total of 1,696 cases from the databases provided by authors of studies that used the Portuguese version of the OCB-KW scale.
Also in the first stage, the six works were read in detail for a general characterization of the production, presenting year of publication, authors, analyzed sample, country in which the study was developed, translation, items that the authors used, analysis techniques used by the authors, the number of factors and items excluded from the studies that used the OCB-KW. Thus, based on these data, it was possible to characterize the applications of the scale and assess its possible instabilities.
The second, more exploratory, stage of analysis, using the secondary database, was divided into three stages. The first moment consisted of using parallel analysis (Horn, 1965). In a second moment, exploratory factor analysis was used for the initial estimation and validation of the constructs. The third moment, eminently confirmatory and comparative, was characterized by the confirmatory estimation of the constructs and the comparison of different models.
Parallel analysis (Horn, 1965) is a statistical Monte-Carlo simulation that consists of the random construction of a hypothetical set of variable correlation matrices, using as a basis the same dimensionality as the real data set. The number of factors in the real data to be retained refers to those that have an eigenvalue >1 and that have a value greater than the respective eigenvalue obtained from the random data. To increase the accuracy of the method, the 95% confidence interval obtained for random eigenvalues should be considered (Timmerman & Lorenzo-Seva, 2011).
In the use of exploratory factor analysis, estimation by principal components was adopted as a method of extracting the dimensions, and Varimax was selected as the rotation method. In addition, initial tests of the Kaiser-Meyer-Olkin (KMO) factor analysis and Bartlett’s Sphericity test were performed. Confirmatory factor analysis was applied to validate the scale constructs and to compare different models. For model estimation, the direct procedure was used and as input matrix the variance-covariance matrix estimated by maximum likelihood via was applied. For the analysis of discriminant validity, correlations were compared to the average variance extracted. Correlations between the constructs must be less than 0.85 (Kline, 2015) and the average variance estimates extracted for two factors must be greater than the square of the correlation between the two constructs (Fornell & Larcker, 1981).
To assess convergent validity, the magnitude and significance of standardized coefficients, the average variance extracted (AVE) and the following adjustment indices were considered: Residual Root Mean Square (RMR), Mean Square Root of the Approximation Error (RMSEA), statistics chi-square (χ2), Goodness-of-Fit Index (GFI), and also by the following comparative fit indices: Tucker-Lewis Index (TLI), Comparative Fit Index (CFI), and Normed Fit Index (NFI).
For the Average Variation Extracted, values equal to or greater than 0.5 are desirable (Fornell & Larcker, 1981). For the chi-square/degrees of freedom relationship, values smaller than five are indicated and values greater than 0.950 are suggested for NFI, TLI, CFI, and GFI; and the RMSEA and RMR should be less than 0.060 and 0.080 respectively (Byrne, 2016; Hooper et al., 2008; Kline, 2015). And, as a third criterion for analyzing the discriminant validity, the chi-square difference test was used, for which differences between the restricted model and the free model greater than 3.84 indicate discriminant validity (Bagozzi et al., 1991).
The reliability of the constructs was evaluated by Cronbach’s alpha (Cronbach, 1951) and Composite Reliability (Fornell & Larcker, 1981), with values above 0.7 are desirable. The one-dimensionality was evaluated from the standardized residuals, constructs that presented values below 2.58 are considered one-dimensional (Hair et al., 2014).
Thus, after model validation, a comparison was made between non-aligned models or competing models (they differ in structure and not just in individual paths). For this, we used information criteria, such as Akaike Information Criterion (AIC) (Akaike, 1973) and Bayesian Information Criterion (BIC) (Raftery, 1993; Schwarz, 1978). In addition, the Expected Cross-Validation Index (ECVI) was also used to compare the models. The model with the lowest values of AIC, BIC, and ECVI, was considered the most fit model (Byrne, 2016; Hu & Bentler, 1999; Kline, 2015).
Analysis of Applications of the Organizational Citizenship Behavior Scale—Knowledge Worker of Dekas et al. (2013)
In this step, the results of the analyses of the six works that used Organizational Citizenship Behavior—Knowledge Worker (2013) are presented. In a search in the Web of Science and Scopus databases using the surname Dekas and year 2013, the original article of the scale and the number of citations in this study were found. All citations were analyzed and two studies were found that used the Dekas scale for measurement (Shen et al., 2017; Srour et al., 2020). Academic Google searches were also performed using the term OCB-KW (2013), where the other four works were found (Andrade, 2017; Andrade et al., 2018; G. L. Á. Costa, 2019; Rebolho, 2018), which also used the scale of Dekas et al. (2013) as a way of measuring.
Considering the changes that the world of work has been undergoing and also the context of knowledge workers, Dekas et al. (2013) developed a new model of OCB, aimed at knowledge workers. Based on surveys carried out in the existing literature on models of OCBr (Bateman & Organ, 1983; Katz & Kahn, 1978; Organ, 1988, 1997; P. M. Podsakoff et al., 2000; Van Dyne et al., 1995; Williams & Anderson, 1991); Dekas (2010) and Dekas et al. (2013) identified that dimensions such as Obedience may not be adequate to measure OCB with knowledge workers.
In this sense, the scale of Organizational Citizenship Behaviors—Knowledge Workers (OCB-KW) by Dekas (2010) and Dekas et al. (2013) was prepared with the following dimensions: Civic Virtue (5 variables), Employee Sustainability (4 variables), Voice (4 variables), Social Participation (4 variables), and Help (6 variables) evaluated by a Likert Scale of 5 points ranging from 1 (strongly disagree) to 5 (strongly agree). The 23 items are provided in the Appendix 1 of this document, and the variable numbers used in the Appendix 1 persist throughout the article.
The Employee Sustainability dimension was defined by Dekas et al. (2013) as participation in activities for the maintenance and prevention of individual and collective health and well-being. The Social Participation dimension refers to participation in social activities during work that are not directly related to core work tasks. The Civic Virtue dimension can be understood as the taking of actions indicative of a macro level of interest for the organization as a whole and the recognition of the responsibilities of being a member of the organization. The Voice dimension means participating in activities, making suggestions or speaking with the intention of improving the organization’s products or some aspect of individual, group or organizational functioning and the fifth and final dimension, Helping, has its concept related to voluntarily helping colleagues from work on work-related problems and prevent problems from occurring.
During replications of the OCB-KW scale, its reliability was satisfactory, presenting a Cronbach’s alpha between .78 and .88 (Dekas, 2010; Dekas et al., 2013). Of these categories, three are in line with the dimensions of citizenship existing in the literature: Civic Virtue, Voice, and Help. Two other categories, Employee Sustainability and Social Participation, emerged in the social context related to knowledge workers, reflecting the impact of changes in the world of work in recent decades.
Initially, in order to analyze the applications of the OCB-KW, the initial characteristics of the works, year of publication, authors, analyzed sample, country in which the study was developed, translation, and items the authors used are listed in Table 1.
Year, Authors, Sample, Country, Translation, and Items of Studies That Used the OCB-KW.
The studies presented are organized in chronological order, however it is noted that the year 2018 was the only one that had more than one study performed. Regarding the analyzed sample, there is a diversity of knowledge workers examined, such as knowledge workers in general (Shen et al., 2017), employees of educational institutions (Andrade et al., 2018; Andrade, 2017), industry workers (Rebolho, 2018), and employees of TI companies (G. L. Á. Costa, 2019; Srour et al., 2020), employees of a city hall (Malheiros, 2021) technical-administrative in education (Parcianello, 2022).
With regard to the country where the study was developed, the original scale was created by Dekas et al. (2013) in the United States. There is a predominance of application of the scale in Brazil with six studies, demonstrating the interest of Brazilian scholars in this subject, followed by China and Egypt with one study each. The results of the search for studies using the scale in question did not return expressive numbers of use, which may be because the scale is a recent adaptation and update of OCB measures.
With respect to the process of scale translation, it is clear that the studies carried out in Brazil (G. L. Á. Costa, 2019; Malheiros, 2021; Parcianello, 2022; Rebolho, 2018) used the version translated by Andrade et al. (2018), which used validation by experts and the protocol suggested by Beaton et al. (2000). This measure was adapted to the Brazilian context by Andrade et al. (2018), through the process of translation and back-translation, semantic analysis, factor analysis, and analysis of internal consistency of items making the necessary cultural adaptations.
The studies using the scale internationally used the original version by Dekas et al. (2013) and did not mention the translation and adaptation to other languages. With regard to the items used, all studies used the 23 items proposed in the original version by Dekas et al. (2013).
Table 2 presents the analysis techniques used by the authors, estimation, rotation, number of factors and variables excluded from studies using the OCB-KW.
Year, Authors, Analysis Techniques, Estimation, Rotation, Number of Factors, and Variables Excluded from Studies Using the OCB-KW.
With Table 2, it is possible to observe the predominance of the exploratory factor analysis technique used by the authors who applied the OCB-KW. The studies carried out by Andrade et al. (2018), Rebolho (2018), G. L. Á. Costa (2019), Malheiros (2021), and Parcianello (2022) used exploratory factor analysis, which obtained different numbers of factors (6, 5, 3, 4, and 5 respectively) in each application, as well as a significant difference in the number of excluded variables (2, 5, 9, 10, and 0 respectively). The authors Shen et al. (2017) opted for the Analysis of Variance, where the number of factors obtained remained as the original by Dekas et al. (2013), not indicating whether any variable was excluded or if any factor had modification in the distribution of variables. The study by Srour et al. (2020) also found the same factors as the original study by Dekas et al. (2013), not reporting if any variable was excluded or if any factor had a change in the distribution of variables.
It is also noteworthy that in Andrade (2017), the two factors Employee Sustainability and Social Participation were not significant to represent the OCB and, based on a confirmatory factor analysis, they were excluded from the final scale.
Importantly, in some of the analyzed studies, other tests were developed with other constructs that were not listed in the Table, as they are not part of the object of analysis of this study.
Then, we sought to evaluate the composition of factors, aiming to analyze the stability of the scale structure. Table 3 presents the authors, the factors formed in the studies carried out by these authors and the items that were grouped into each factor, in addition to the Cronbach’s alpha of each factor. It is noteworthy that the numbers of variables used in Appendix 1 persist throughout all analyses.
Authors, Factors, Items of the Works Using the Exploratory Factor Analysis, and Cronbach’s Alpha.
Table 3 shows a comparison of the different results obtained from the exploratory factor analysis of the OCB-KW scale. When applying the OCB-KW scale model translated by Andrade et al. (2018), it was possible to observe a certain instability in the results from different contexts, which affects both the construction of the content of the factors and in the grouping of items. In the six cases (Andrade, 2017; Andrade et al., 2018; G. L. Á. Costa, 2019; Malheiros, 2021; Parcianello, 2022; Rebolho, 2018), there was a divergence in the formation of factors, or for needing to rename factors due to the grouping (Andrade, 2017; Andrade et al., 2018; Parcianello, 2022; Rebolho, 2018), or because the number of factors (Andrade, 2017; G. L. Á. Costa, 2019; Malheiros, 2021) is smaller than the original version by Dekas et al. (2013).
Another element that deserves to be highlighted refers to the reduction in the number of items. Only in the study of Parcianello (2022) the 23 items remained, And it is noteworthy that items 3 and 4 were not retained in any of the other five studies that conducted the factor analysis (Andrade, 2017; Andrade et al., 2018; G. L. Á. Costa, 2019; Malheiros, 2021; Rebolho, 2018). The Employee Sustainability factor only appears in the study by Andrade et al. (2018), and item 2 of this factor was excluded from almost all other studies analyzed (Andrade, 2017; G. L. Á. Costa, 2019; Rebolho, 2018), except in Parcianello (2022). The smallest number of factors for this scale was reported by G. L. Á. Costa (2019), totaling three factors.
Cronbach’s alpha values ranged from .780 to .880, applying the scale in the original study by Dekas et al (2013), from .713 to .769 in Andrade (2017), from .626 to .832 in Andrade et al. (2018), from .602 to .747 by Rebolho (2018), from .770 to .829 in G. L. Á. Costa (2019), from .715 to .788 in Malheiros (2021) and from .703 to .825 in Parcianello (2022) demonstrating that, in general, the constructs present internal consistency.
Exploratory Analysis of the Organizational Citizenship Behaviors—Knowledge Worker Scale by Dekas et al. (2013)
Due to different evidence regarding the number of dimensions, initially, it was decided to assess the scale’s dimensionality. Although empirical studies have predominantly used the criterion of eigenvalues greater than one to define the number of dimensions, we chose to use the parallel analysis due to evidence that the eigenvalues criterion overestimates the number of factors and the good performance of the parallel analysis (Timmerman & Lorenzo-Seva, 2011). Table 4 lists the results.
Parallel Analysis of Horn (1965) for the Organizational Citizenship Behavior—KW Scale.
Horn’s parallel analysis indicates that the scale should have five dimensions, as there are five eigenvalues greater than one and with real values greater than the simulated ones. These results are in line with the original dimensionality proposal by Dekas et al. (2013) and was also obtained by Rebolho (2018).
Next, a factor analysis of the items of the OCB-KW Scale was run, initially using the 23 questions of the instrument. Firstly, the adequacy of the factorability of the sample was attested, since a coefficient of 0.894 was obtained for the KMO, and the Bartlett’s Sphericity test (value = 13,792, Sig < 0.001) rejected the null hypothesis.
The five factors generated together explain 58.42% data variability. In Table 5, these factors were analyzed in detail, considering the component items, the original factor and the factor loading for each of the items, as well as the naming of the created factors.
Exploratory Factor Analysis of OCB-KW.
The results indicate the structure with five factors as the most adequate, retaining the 23 questions. It is noticed that only Factor 3, Civic Virtue, remained the same as its original version in the Dekas scale with all items (items 9, 10, 11, 12, 13). Factor 1—Voice (items 14, 15, 16, 17, 18, 19) found in this study approached the factor Voice found by G. L. Á. Costa (2019), which grouped items 14, 15, 16, 17, 19. The second factor—Employee sustainability (1, 2, 3, 6, 8, 20) was not found in any of the studies previously conducted. The third factor—Civic Virtue, composed of items 9, 10, 11, 12, 13, in addition to maintaining the original measure, was also reported by Andrade et al. (2018) and Andrade (2017).
Also in this study, the fourth factor—Helping grouped items 4, 21, 22, 23. The study by Andrade et al. (2018) and Andrade (2017) found a factor with a very similar grouping (21, 23, 24), considering that item 4 was excluded by the analysis of commonalities in the four studies analyzed here. Finally, the fifth factor—Social Participation (items 5 and 7) was not observed in this grouping composition in any of the previous studies analyzed.
All factor loadings are greater than 0.4 and no cross-factor loadings were found, indicating the proper allocation of variables in the factors. In summary, it is clear that when analyzing the behavior of the OCB-KW Scale, the factors continue to show instability, with no constancy either in the number of factors or in the grouping of items.
Confirmatory and Comparative Analysis of the Scale
Considering the differences in the results obtained in the exploratory analysis and the structure proposed by Dekas et al. (2013) for the scale, it was decided to initially compare two models: model 1, with the original structure proposed by Dekas et al. (2013) and model 2, with the factor structure obtained in this study. Initially, Table 6 lists the results of the confirmatory factor analysis (CFA) of the constructs of the OCB-KW Scale, according to the original model.
CFA Results for the Constructs of the Original Model by Dekas et al. (2013).
The results demonstrate that the initial models of the constructs have inadequate values for the fit measures considered. Thus, using the model improvement strategy, some measures were used, specifically: (1) for Employee sustainability, variable 4 was removed (loading = 0.433); (2) in Social Participation, variable 8 was excluded (loading = 0.286); (3) in Civic Virtue, correlations were set between e9–e10 and e9–e13; (4) in the Voice construct, correlations between e16 and e17 were inserted; and (5) for Helping, variables 23 (loading = 0.035) and 22 (0.391) were excluded and a correlation between e19 and e20 was established.
After the changes, the results of the constructs with regard to the fit indices are satisfactory. However, the same cannot be said for the composite reliability and the average variance extracted. Cronbach’s alpha for Social Participation was below .6, indicating low internal consistency. This factor was also the one with the lowest alpha values in the studies by Andrade et al. (2018) and Rebolho (2018).
Next, Table 7 was constructed to test the discriminant validity of the constructs of the model.
Discriminant Validity for the Original Model by Dekas et al. (2013).
Other values—correlation values between constructs.
The results of this table reveal that most constructs have discriminant validity, since the values of the main diagonal are greater than the values of the correlations between the constructs. This confers discriminant validity according to Fornell and Larcker (1981). The exception is the relationships highlighted in bold, for which the values of correlations between the constructs are higher than the square root of the AVE, suggesting the lack of discriminant validity. For these cases, we decided for the chi-square difference suggested by Bagozzi et al. (1991), whose results are listed in Table 8.
The results in Table 8 indicate discriminant validity between the exposed constructs given that the chi-square difference for all tested relationships is greater than 3.84. Therefore, considering that all constructs have correlations with each other of less than 0.85, as suggested by Kline (2015) and that even the three relationships not confirmed in the comparative analysis of the square root of the AVE with the correlation showed significant differences in the chi-square difference test, it can be concluded that the constructs of the original model by Dekas et al. (2013) are discriminating among themselves.
However, as demonstrated throughout the present study, the composition of factors (constructs) of this scale has shown differences in several studies and also here. Therefore, the CFA procedures and discriminant validity for the model resulting from the work were performed. The CFA results are listed in Table 9.
CFA Results for the Model Constructs Resulting from the Work.
These estimates were calculated from the loadings of these variables in the integrated model.
Again, it is observed that the initial models of the constructs have inadequate values for the fit measures. Thus, the following procedures were performed: (1) for Voice, variable 17 was removed (loading = 0.429) and correlations between e18–e19 and e18–e14 were established; (2) for Employee sustainability, variable 6 was excluded (loading = 0.360) and correlations between e3–e8 and e1–e20 were inserted; (3) in Civic Virtue incorporated correlations between e9–e10 and e9–e13; and (4) in Helping, correlations were established between e16 and e17.
With these fits made, it can be seen that the model fit indices became adequate. However, as for the constructs of the original model, the composite reliability measures are not suitable for two constructs and, especially, the results of the average variance extracted are not suitable for any construct. It should be noted that in most of the studies analyzed (Andrade, 2017; Andrade et al., 2018; G. L. Á. Costa, 2019; Rebolho, 2018), as a measure of reliability, only Cronbach’s alpha was used, most of which were superior to .7 or at least .6 allowing the authors to conclude for the consistency of the constructs. The only study that used the AVE also presented values below desirable (Andrade, 2017).
To assess the discriminant validity of the constructs of the model, the same tests applied to the original model were performed, the results of which are presented in Tables 10 and 11.
Discriminant Validity for the Model Resulting from the Work.
Other values—correlation values between constructs.
Chi-Square Difference Test for the Model Constructs.
All correlations between constructs are below 0.85, suggesting discriminant validity according to the rule of Kline (2015). Nevertheless, according to the criterion of Fornell and Larcker (1981), only the Social Participation construct is discriminated from the others, since the values of the main diagonal are greater than the values of the correlations between the constructs. For values in bold, the discriminant validity cannot be concluded. Therefore, the chi-square difference test for these relationships was also performed. The results are presented in Table 11.
The chi-square difference for all tested relationships is greater than 3.84, indicating that there is discriminant validity between the constructs. Therefore, as all constructs were considered discriminating in at least two of the three criteria used, it is considered that there is discriminant validity of the constructs.
As the constructs of the two models were considered valid according to the criteria used, the final step consisted of the construction of the structural model, which contains all the constructs and their respective correlations, aiming to identify the most parsimonious model. Therefore, in addition to the fit measures already used, three additional measures for comparing models were added, namely: AIC, BIC, and ECVI. The results are presented in Table 12.
Results of the comparison measures of the tested models.
Note. MI = initial model; MF = final model.
The initial models were not fit, so the strategy of improving the models was adopted, in which correlations between the errors of the variables that made theoretical sense were inserted. The results show that both models have adequate fit measures after the adjustments made.
Analyzing the three comparison measures between two models (AIC, BIC, and ECVI), it appears that the final version of the original model by Dekas et al. (2013) has the same values in the three indices, demonstrating that this is the most parsimonious model. Therefore, the results show that the OCB-KW Scale has the five dimensions proposed by Dekas et al. (2013) and that they are correlated with each other.
Rebolho (2018) reported correlations when specifically analyzing the factors that make up the OCB-KW construct, with all positive associations. With regard to the intensity of the associations detected by Rebolho (2018), the Civic Virtue factor stands out, which when related to the factors Voice (0.436) and Voluntary Participation (0.407), presents moderate coefficients.
G. L. Á. Costa (2019) analyzed the factors composing the OCB-KW construct, and observed that the factors are positive and have the same sense of association, with moderate intensity between Voice and Civic Virtue (0.507), between Helping and Civic Virtue (0.512), and the highest coefficient among the factors refers to Helping and Voice (0.576). Figure 1 illustrates the final original model.
Final original model.
This model presents the five dimensions proposed by Dekas et al. (2013), however, not all 23 questions proposed in the original scale are present. In the Employee Sustainability construct, the variable “I praise others when they succeed” was excluded. This variable was also excluded in five of the six studies that used the scale, not repeating in the study by Parcianello (2022). From the Social Participation construct, the variable “I’m excited about work environment interactions” was excluded, which had already been excluded in Andrade (2017) and G. L. Á. Costa (2019). In the Helping construct, the variables “I consider the impact of my actions on co-workers” and “I communicate with others before taking actions that may affect them” which had only been excluded in the study of Malheiros (2021).
The correlations between the constructs vary between 0.302 and 0.719. The existence of positive correlations between the five dimensions of the scale is in accordance with the perception that all dimensions are directly associated, that is, all their interrelationships contribute to the construction of a measure of OCB.
Alternatively, the model in Figure 1 was also compared to a model in which OCB is a second-order construct, formed by the five dimensions. Despite good fit indices (CFI = 0.953, NFI = 0.94, RMSR = 0.027, RMSEA = 0.044), the model was less parsimonious (AIC = 625.17, BIC = 1,022.00, ECVI = 0.369). Therefore, the hypothesis that it is valid to measure the OCB of knowledge workers based on the dimensions: voice, civic virtue, help, social participation, employee sustainability and their correlations is confirmed. It is noteworthy that the best structure found includes 19 measurement items.
Final Considerations
With the aim of analyzing the empirical instabilities and adopting new validation procedures for the OCB-KW scale, the characterization of eight studies found in the Scopus, Web of Science, and academic Google databases was carried out.
From the general characterization of the production, it is possible to observe that more than half the works that used the scale by Dekas et al. (2013) were developed in Brazil. It should be noted that only the study by Andrade et al. (2018) aimed at the cross-cultural validation of the scale, and that out of the studies that used translation, none achieved the same results.
When examining the exploratory factor analyses developed by Andrade (2017), Andrade et al. (2018), Rebolho (2018), G. L. Á. Costa (2019), Malheiros (2021), and Parcianello (2022), the instability of the OCB-KW scale is evident: (1) there was no constancy in the number of factors, with scales with three (G. L. Á. Costa, 2019), four (Andrade, 2017; Malheiros, 2021), five (Parcianello, 2022; Rebolho, 2018), and six factors (Andrade et al., 2018) having been obtained. Another issue is that the way of grouping the variables was not maintained, since many items started to be grouped with a set of items different from the original composition and, in addition, there was a variation in the number of items that make up the scale.
From the verification of instability in the applications of the scale, we sought to apply validation techniques through four procedures: the analysis of dimensionality, the evaluation of the constructs, the verification of the discrimination between the constructs and the comparison between different models. The dimensionality analysis indicated that the solution with five factors was the most adequate for the scale, corroborating the dimensionality proposed by Dekas et al. (2013). In the individual assessment of the constructs, it was observed that the initial models did not meet the established fit criteria, which were achieved using the model improvement strategy, except for the average variance extracted, composite reliability and Cronbach’s alpha of some constructs. From the application of three criteria to assess the discriminant validity, it was possible to observe that the constructs are different from each other. Thus, after validating the factorial composition of the scale, we sought to find the best structural model. Again, after adopting some adjustments, two fitted models can be identified. Finally, from the comparison indices between models, it can be concluded that the best structure for the OCB-KW scale is built from five dimensions correlated with each other.
In future applications of the scale, the following points are suggested to researchers. If they want to evaluate the scale with exploratory techniques, prioritize the use of parallel analysis to identify the dimensionality instead of the eigenvalues criterion, since the parallel analysis has shown better results. When deciding for the rotation method, it does not seem appropriate to use the Varimax predominantly used so far, since there is evidence that the factors are correlated. In this context, oblique rotation methods would be more suitable (Browne, 2001) such as promin. Another important issue regarding the use of exploratory analyses is the adequacy of the Pearson correlation matrix, which can be replaced by the polychoric matrix, especially in cases where the scale items present asymmetry and/or kurtosis (Muthén & Kaplan, 1985).
With an adequate sample size, it is understood that, according to the evidence presented in this study, the researcher could decide to use confirmatory factor analysis. The solution that presented the best fit in this study was structured with 19 items; four of them excluded from the original scale. Nonetheless, because the items to be excluded are diverging between the different studies, with the exception of item 4, which was eliminated in four studies, it is suggested that researchers start the validation with the proposed 23 items and that they use the strategy of improvement of the model, mainly with the removal of variables with low factor loadings and insertion of correlations between errors in the search for the best fit.
Importantly, the removal of variables with lower factor loadings, generally smaller than 0.7, tends to improve convergent and discriminant validity problems, as it increases the composite reliability and the Average Variance Extracted, however, there will be a loss of content validity. Since the OCB-KW scale presented some problems in these indicators, especially for the Social Participation construct, the removal of the variable with low loading could be a solution, which was not adopted as it would reduce by half the number of variables forming the construct, which would have only two items. As this construct was not maintained in two of the studies and in others it presented the lowest Cronbach’s alpha, a valid attempt would be the proposition and validation of new items for this dimension in future applications of the scale.
Identifying the correct structure for a scale has a fundamental role in the construction of measuring instruments (Timmerman & Lorenzo-Seva, 2011). The use of scales is an important way to understand human behavior and identify perceptions in organizations. The practical contributions of this study lie in the fact that Organizational Citizenship Behaviors are associated with individual, discretionary actions that contribute to the effective functioning of the organization. Such behaviors are valued within organizations, as they contribute to their competitiveness. In this way, organizations are encouraged to adopt measures to create a favorable climate for citizenship, encouraging the CCO, instead of just trying to hire people who have personal characteristics that may predispose to engage in such behavior.
The theoretical contributions of this study imply the existence of a valid and robust measure to identify or not the existence of OCB in certain contexts, providing a reliable assessment of the construct. A limitation of this study is the lack of comparative tests with other organizational citizenship scales, which may be carried out by future research. It is recommended to replicate the scale in other samples through quantitative research and statistical analysis, such as confirmatory factor analysis, to compare the results. It is also recommended to apply the scale in different organizations and in studies that assess OCB as antecedents, consequences or related to other constructs, such as organizational justice and ethical leadership, which address fair treatment in the workplace, helping employees gain trust, reduce uncertainty, and behave ethically (Al Halbusi et al, 2022), which may favor OCBs. Thus, it is expected that further research will contribute new empirical evidence and more psychometric testing of the scale.
Footnotes
Declaration of Conflicting Interests
Funding
ORCID iDs
