Workplaces are spaces characterized by rules defined as appropriate in a specific society. Sometimes such rules might be broken and people might carry out behaviors considered as deviant at that moment and in that context (Pawar, 2013), affecting negatively employees’ well-being and satisfaction. In the last few years, several researchers have focused their attention on deviant behaviors at work and many of them have been identified (e.g., Hershcovis, 2011): social undermining (Duffy et al., 2002), bullying (Hoel et al., 1999; Einarsen, 2000), ostracism (Williams, 2007), or abusive supervision (Tepper, 2000) are only some of them. Quite recently, workplace incivility (WI), defined as “low-intensity deviant behavior with ambiguous intent to harm the target, in violation of workplace norms for mutual respect” (Andersson and Pearson, 1999, p. 457), has captured the attention of many researchers. WI is characterized by the ambiguity about the real intention of the perpetrator. In fact, uncivil acts might be the result of an intentional mistreatment; but it might also be the consequence of a lack of nicety (Pearson et al., 2000) without any goal of harming the other part. Moreover, following the original definition, WI might pass as unseen, given the low-intensity of acts that enter in its domain. In the original works of (Andersson and Pearson, 1999; Pearson et al., 2001) and in the following research on this issue (e.g., Cortina et al., 2001, 2013) behaviors of different nature where identified as WI. Shouting against a coworker, isolating him or her, devaluating a colleague’s job, making jokes at others’ expense: all these behaviors enter into the realm of what is defined as WI. The subtlety of such acts makes difficult to define them as mistreatments and can be attributed to a rude environment (Leiter, 2013). Various researchers (e.g., Hershcovis, 2011) have pointed out the growth and, often, the overlapping between constructs related to negative acts at work. One of them is social undermining (Duffy et al., 2002) which refers to those behaviors whose intent is undermining victims’ work relational ties and reputation. Although related concepts, they can be differentiated by the intent to harm which is clearer in social undermining and ambiguous in case of incivility.

Although uncivil acts are not blatantly aggressive and victims often are not aware of the perpetrator’s intent to harm, they might be the cause of several prejudicial consequences for people’s well-being and organizations (Di Marco et al., 2015; Garrosa et al., 2015; Kabat-Farr et al., 2016; Schilpzand et al., 2016). Such consequences depend on the intensity of the act and its frequency (Hershcovis, 2011; Leiter, 2013). Incivility is an organizational stressor (Griffin, 2010) and being victim of uncivil behaviors produces negative outcomes. Previous studies demonstrated that WI affects physical and psychological well-being, diminishes job satisfaction and increases the desire of revenge (Andersson and Pearson, 1999; Cortina et al., 2001; Pearson et al., 2001; Diaz et al., 2011; Moreno-Jiménez et al., 2012; Kabat-Farr et al., 2016); it also reduces job performance, increases the intention to leave and organizational costs (Griffin, 2010; Porath and Pearson, 2010; Taylor et al., 2012; Schilpzand et al., 2016; Mao et al., 2017). Moreover, recent research has demonstrated that being victim of WI can lead to counterproductive work behaviors (voluntary actions which are prejudicial for the organization or their members) (Sakurai and Jex, 2012; Welbourne and Sariol, 2017), increasing the cost of incivility for the organization. People who perceive WI spend more time thinking on uncivil acts experienced, are less satisfied and the probability to quit the organization is higher (Cortina et al., 2001).

The cost of WI is paid by victims and organizations, but also by bystanders and all those people who enter into the uncivil spiral (Montgomery et al., 2004; Porath and Erez, 2009). In fact, not paying attention to WI might involve the escalation of conflict. Workers who are victims of uncivil acts might ignore such treatment or might desire to retaliate with the original perpetrator or with another person. Indeed WI might be contagious (Andersson and Pearson, 1999) and a spiral of incivility might rise, involving many people within the organization. All the organizational actors might be implicated. Past studies found that supervisors, coworkers and clients might perpetrate incivility (Schilpzand et al., 2016). Specially, people at the top of the hierarchy might use more often incivility as a mean to exercise power or to push productivity (Porath and Pearson, 2010; Kabat-Farr et al., 2017).

Rudeness and uncivil behaviors are largely widespread at the workplace and many workers confirmed they have been victims of them at least once in their working life (Cortina et al., 2001; Pearson and Porath, 2013). In order to measure the impact of this phenomenon, numerous qualitative (Andersson and Pearson, 1999; Pearson et al., 2001; Di Marco et al., 2015) and quantitative (Welbourne et al., 2015) studies have been carried out. Several scales have been developed with the goal of creating a comprehensive instrument in order to measure quantitatively and objectively uncivil acts at the workplace (e.g., Cortina et al., 2001, 2013; Martin and Hine, 2005). However, “WI is not an objective phenomenon; it reflects people’s interpretation of how actions make them feel” (Porath and Pearson, 2010, p. 64) and several factors might play a role in the process of interpretation of behaviors considered uncivil (e.g., hierarchical status) (Schilpzand et al., 2016).

Although different measures have been developed (e.g., Cortina et al., 2001, 2013; Martin and Hine, 2005; Wilson and Holmvall, 2013), Cortina and colleagues’ Workplace Incivility Scale (WIS), in the original (Cortina et al., 2001) and revised version (Cortina et al., 2013) is the most known and applied instrument to date. The WIS is a unidimensional scale that brings together 7 (Cortina et al., 2001) or 12 items (Cortina et al., 2013) which refer to several uncivil behaviors experienced in a variable recall window: “the past 5 years” in the former version, and “the last year” in the latter. In fact, researchers were conscious that results can be affected significantly by the length of the recall window (Blau and Andersson, 2005; Matthews and Ritter, 2016), due to distractions that a longer recall window might create (Matthews and Ritter, 2016).

There have been several attempts to reduce or adapt the WIS to different research goals (e.g., Blau and Andersson, 2005; Matthews and Ritter, 2016). Recently, Matthews and Ritter (2016) elaborated a shorter (four-item scale) WIS based on the second version of Cortina et al.’s (2013) WIS. Applying a shorter instrument is important in order to improve the clarity of the construct that researcher would measure; it enhances the probability that respondents complete the questionnaire, even more when several scales are applied at the same time (Fisher et al., 2015; Balducci et al., 2017). Reduced scales are necessary when researchers carry out longitudinal studies and, also, when responses are given by computer, due to the largely use of electronic surveys (Balducci et al., 2017).

The four item-reduced version of the WIS (Matthews and Ritter, 2016) was the result of a complex analysis which first had the goal to explore the judgmental qualities of the scale, understanding which items of the 12-items WIS version (Cortina et al., 2013) respected the initial conceptualization that Andersson and Pearson (1999) gave to the construct. In other terms, researchers explored if both the distinguishing features of the original construct -the ambiguous intent to harm and the low intensity of acts- were reflected by the WIS. Then convergent and divergent validity were analyzed and the four items of the reduced version were identified. The reduced WIS showed good psychometric properties with a Cronbach’s α = 0.79 and factor loadings ranged from 0.57 to 0.77. The reduced scale correlated 0.93 with the 12-items scale and both explained similar levels of variance in the dependent variables analyzed (e.g., civility norms, interpersonal deviance, organizational deviance, etc.).

The goal of this research is to validate a Spanish version of the WI reduced scale elaborated by Matthews and Ritter (2016). The recent increment of workers’ demands and the escalation of work insecurity due to the financial crises have involved several countries. For instance, in Spain, only 59% of workers perceive job security (Eurofound, 2017). According the 6th Eurofound Working Conditions Survey, 16% of European workers perceived adverse social behavior at work and 16% have not been treated fairly in their workplace. In Spain the prevalence of these issues is 10 and 21%, respectively (Eurofound, 2017). Thus, this trend might have reinforced negative acts such as WI which might be instigated by a demanding workplace (Schilpzand et al., 2016). The few studies on WI in Spain showed that uncivil behaviors generate workers less satisfied, who carried out more counterproductive behaviors (Diaz et al., 2011); the emotional exhaustion increases (Garrosa et al., 2015) as well as the intention to leave the organization (Moreno-Jiménez et al., 2012). Although research carried out in Spain to date does not show high levels of incivility, they highlight its prejudicial effects at individual and organizational level. In order to recognize and evaluate the impact of WI, it is necessary to analyze the psychometric properties of the WIS, making available a valid and reduced measure of such construct for those studies carried out in Spain. In the present study we analyze, firstly, the items reliability and factor loading of the Spanish reduced WIS; secondly, we analyze the convergent validity of the reduced WIS with another related concept, such as social undermining; and the concurrent validity with several outcomes propensity to leave, job satisfaction, and psychological well-being. Finally, as Matthews and Ritter (2016) highlighted in their work, it is necessary to obtain an invariant scale based on the characteristics of the employment type. While their measure results to be invariant when applied to full-time versus part-time workers and to men and women, they did not consider other features of the employment, such as the type of contract (permanent vs. temporal contract) and the position occupied in the hierarchical scale (supervisors vs. non-supervisors). In fact, people who are in a temporary contract might be more likely victims of WI, because their transit within the organization is perceived as limited and perpetrators might care less about consequences derived from negative behaviors against this group of workers (Salminen and Saloniemi, 2010). The same process might arise when we are in front of people with different level of formal power. As research has underlined previously (e.g., Schilpzand et al., 2016), subordinates experience WI more than superiors do. Consequently, the goal to validate an invariant measure will be to find out those behaviors that are uncivil beyond the employment status.

Participants were 407 employees from a broad range of professional backgrounds, including trade (28%), education (25.3%), insurance (10.8%), health and welfare (8.8%), catering (6.4%), industry (4.7%), culture and leisure (4.2%), construction (3.2%), transport (1.7%), business service (1.7%), agricultural sector (1.2%), government (1.2%), communication (1%), and others (2.8%). The percentage of female employees was 59.2%. Regarding their highest education, 58.5% had a university degree, 30% had completed high school or vocational education, 10.1% had completed elementary studies, and 1.5% had not completed any formal education. Respondents’ mean age was 38.73 (SD = 10.45) years old. Most respondent held a fulltime employed (70.3%) and no managerial position (57.5%). About 72% held permanent contract, 18.9% held temporal contract and 2.2% were freelance. Participants worked an average of 26.54 (SD = 14.37) hours per week. Their average seniority at the current company was about 12 years (SD = 9.87).

According to the International Test Commission guidelines (Hambleton et al., 2005; International Test Commission [ITC], 2005), four expert researchers first translated the WIS into Spanish trying to preserve the original meaning of items. Second, the translated items were discussed and the four experts got a consensus about the content of the items. In general, item translation did not generate debate and consensus was reached easily. Third, a back-translation was conducted by an English speaking translator and the equivalence between the two versions was assured. The final Spanish version of the scale is shown in “Appendix A”.

Employees from various organizations in Spain were recruited using an incidental sampling scheme. They voluntary fulfilled a cross-sectional survey in approximately 15 min. Participants were informed about the anonymity and confidentiality of the survey.

Confirmatory factor analysis (CFA) was conducted using AMOS 19.0 and all other statistical procedures were computed with SPSS 24.0. Since there was evidence of data’s deviation of normal distribution and the sample is moderated we computed analysis with Asymptotic Distribution-Free (ADF) method (Jones and Waller, 2015). Model evaluation included an examination of the model fit indices and squared multiple correlations. As widely recommended (Byrne, 2010; Mueller and Hancock, 2010; Garson, 2012; West et al., 2012) the goodness-of-fit of the factor structure was evaluated from three different perspectives (incremental, parsimonious, and absolute) using the normed chi-square value, the comparative fit index (CFI), the non-normed fit index (NNFI), the root mean square error of approximation (RMSEA), and the standardized root mean square residual (SRMR). Moreover, the measurement invariance across gender and work characteristics groups was analyzed through a multi-group CFA. After configural invariance across groups was established, testing for metric invariance and scalar invariance was performed using Δχ² and ΔCFI tests for two nested models (Brown, 2006; Byrne, 2010; Dimitrov, 2010). A change of equal to or < 0.01 for CFI indicates that invariance holds (Cheung and Rensvold, 2002).

The internal consistency, concurrent and convergent validity of the scale were examined by Cronbach’s alpha coefficients and Spearman correlation coefficients, respectively. Moreover, correlational analyses and analysis of variance were used to perform group comparison.

Regarding the distributional properties of the 4 items, means ranged from 1.96 to 1.43 (see Table 1). When studying negative acts at work, such as WI, low means are common (e.g., Kabat-Farr et al., 2016). Standard deviations ranged from 0.99 to 0.78, skewness ranged from 0.90 to 2.45 and kurtosis ranged from 0.35 to 6.35. Some items are below indices for acceptable limits of 2 (Field, 2000, 2009; Trochim and Donnelly, 2006; Gravetter and Wallnau, 2014). Hence, as shown in Table 1, there is evidence of deviation of normal distribution.

Confirmatory factor analysis model with the unconstrained factor loadings are shown in Figure 1. The factor loadings ranged from 0.48 to 0.77 and parameters show a good fit (see Figure 1). Two CFAs were conducted for each pair of work characteristics –men vs. women; permanent vs. temporal contract; full-time vs. part-time and supervisor vs. non-supervisors- separately. Taken together, values suggest a good fit. A non-significant chi-square indicates a perfect fit, and a good fit is indicated when the ratio between the chi-square statistic and the degree of freedom is less than 5 (West et al., 2012). All CFI and NNFI values are > 0.95 indicating a good fit (Byrne, 2010). The RMSEA for the scale with all participants indicates a well-fitting model (Browne and Cudeck, 1993). With only one exception (for part-time employees), RMSEA values are lower 0.08, indicating an acceptable fitting (Hu and Bentler, 1999). All SRMR values are under 0.08 indicating a good fit (Hu and Bentler, 1999). Standardized residuals revealed that all values fell well below the 0.15, which is consistent with the notion of an acceptable fit (Byrne, 2010) (see Figure 1).

Based on selective discrimination’s assumptions, different ANOVA was conducted with all demographic variables: sex, age, studies, civil status, type of contract, employment status, responsibility position, and sector. There were no differences in WI perceptions.

Using the criterion by Nunnally (1978), the alpha coefficient of the reduced measure showed a good internal consistency (Cronbach’s α = 0.75). The mean and standard deviation of each item and the inter-item correlations are presented in Table 1. Sequential deletion of items generated alphas ranging from 0.66 to 0.72.

To test the convergent validity, Spearman’s correlation coefficients were computed between the reduced WIS and social undermining, the propensity to leave a job, job satisfaction and psychological well-being measures. In addition, guidelines Cohen’s (1988) were used to determine the strengths of the correlation coefficients between the reduced WIS and other measures, with r around 0.1 indicating small effect sizes, around 0.3 indicating medium effect sizes, and r > 0.5 indicating large effect sizes. Assessing for convergent validity, as expected, the reduced WIS had a large positive effect with social undermining. Regarding the concurrent validity, results showed that the scale was significantly, although moderately, associated with propensity to leave a job (r = 0.21, p < 0.01), job satisfaction (r = -0.44, p < 0.01) and psychological well-being (r = -0.20, p < 0.01) (see Table 2).

To test measurement invariance across groups, configural, metric and scalar models were compared (see Table 3). The results showed that these four types of invariant models had a good fit in three out of four pair of groups analyzed (permanent vs. temporal, full-time vs. part-time and supervisor vs. non-supervisors groups). Multigroup analysis showed no invariance between men vs. women groups (see Table 3).

In the original model, CFA for men showed a good fit (see Table 3 for the fit indices). However, the CFA for the original model did not fit for women (see Table 3). Therefore, no invariance models between men and women are analyzed. According with Byrne (2010), we can conclude about metric invariance for permanent vs. temporal contract groups. Although the value of the ΔCFI was below the cut-off value (ΔCFI = 0.030), Δχ² is not significant (1.77, p > 0.005), showing invariance. For full-time vs. part-time and supervisor vs. non-supervisors groups, when configural and metric invariant models were compared, no significant changes occurred in fit indices. Both ΔCFI and Δχ² were smaller than the cut-off values (ΔCFI = 0.001 and Δχ² = 1.85, p > 0.05, for full-time vs. part-time groups; and ΔCFI = 0.000 and Δχ² = 4.70, p > 0.05, for supervisor vs. non-supervisors groups). These results suggest that factor loadings were invariant across the each pair of groups.

When metric and scalar invariant models were compared, neither significant changes occurred in fit indices (ΔCFI = 0.000 and Δχ² = 5.51, p > 0.05, for full-time vs. part-time groups; and ΔCFI = 0.000 and Δχ² = 2.29, p > 0.05, for supervisor vs. non-supervisors groups). These results suggest that factor loadings and intercepts were invariant across the each pair of groups. However, for permanent vs. temporal contract, again, although Δχ² = 11.241, p > 0.05, ΔCFI = -0.095, so invariance could not be concluded.