Authentic Leadership Questionnaire: invariance between samples of Brazilian and Portuguese employees

Authentic leadership model

Conceived by Luthans and Avolio (2003), the authentic leadership (AL) model considers the leader as self-effective, hopeful, resilient, and optimistic. AL is represented by high moral character, altruism, and virtuousness (Avolio and Gardner, 2005). According to Avolio and Gardner (2005), the authentic leader presents the capacity to analyze and make decisions on ethical dilemmas. Walumbwa et al. (2008) add to AL attributes the promotion of a climate of ethical work and development of followers’ psychological capacities. The theoretical model of AL is formed of four dimensions. The first, relational transparency, consists of presentation of the leader’s genuine “self” rather than a modification with regard to the external context (Ilies et al. 2005). The second, moral and ethics, deals with the leader’s attitudes, values, and moral standards (Walumbwa et al. 2008). The third, balanced processing, deals with others’ consideration of the leader’s deepest information and reflections (Walumbwa et al. 2008). And self-awareness includes knowledge of self and one’s own potential (Avolio and Gardner, 2005).

The Authentic Leadership Questionnaire (ALQ) is a self-administered instrument that assesses the perception of AL (Walumbwa et al. 2008) and has been applied in several cultures and organizational contexts (Gardner et al. 2011). The instrument has two versions, one for leaders to answer about themselves (self-report) and another for employees to rate their leaders.

The ALQ is used in research correlating AL with, for example, working atmosphere (Nelson et al. 2014), team effectiveness (Xiong and Fang, 2014), positive emotions (Ilies et al. 2013), and psychological capital (Laschinger and Fida, 2014). Creativity (Rego et al. 2012) and innovation (Černe et al. 2013) are also variables studied with this measure. Research samples have involved professionals in such different contexts as the navy and army (Borgersen et al. 2014), health (Laschinger and Fida, 2014), and education (Monzani et al. 2014). AL is also studied in institutions of different segments and sizes (Gardner et al. 2011; Hsiung, 2012).

The ALQ has been translated into several languages, and these versions are provided by the authors. Brazilian studies of AL (Lanzoni and Meirelles, 2011; Sobral and Gimba, 2012) have used a Portuguese version of the instrument but without validating it and assessing its psychometric qualities (Esper and Cunha, 2015; Fonseca et al. 2015).

Measurement invariance in authentic leadership studies

Considering that the ALQ assesses the level of authenticity attributed to the leader, understanding of “being authentic” can be affected by cultural influences (Nyberg and Sveningsson, 2014). For Diddams and Chang (2012), authenticity has distinct mobilizing elements and should therefore not be conceived linearly, as proposed in the initial model. For Milfont and Fischer (2010), cultural idiosyncrasies should be investigated in careful applications of the same measure to distinct groups and individuals. Cross-cultural research examines variations in psychological constructs in different cultural contexts (Rašković and Kržišnik, 2010). This kind of study contributes to advancing organizational research, allowing valid use of an instrument in different groups of individuals (Milfont and Fischer, 2010).

In 1994, Graen and Wakabayashi (1994) stated that leadership studies could not ignore variations in characteristics between countries (Hofstede, 1980, 2001). The AL is always expressed within a specific culture, a specific organization, and a specific society, which leads Owusu-Bempah et al. (2011) to debate its universal and specific character. That discussion converges with the assumptions of the study of Global Leadership and Organizational Behavior Effectiveness (GLOBE). For Minkov and Blagoev (2012), the current complex organizational scenario means leaders must develop a particular set of skills. In a survey, GLOBE was applied to executives in 62 cultures and societies (House et al. 2004). This identified the main similarities and differences in the contexts analyzed and found three structuring dimensions of the leader’s global competences. Those competences, also called the global mindset, unite the leader’s intellectual capital, social capital, and psychological capital.

Although criticism of GLOBE is found in the literature (e.g., Hofstede, 2006), House et al. (2004) argue that the model provides a development path of the global mindset for organizations and leaders. Čater et al. (2013) consider that this leadership model contributes to making global objectives and strategies suitable for operation in local markets. Due to the many demands of the globalized market, organizations need flexible leadership (Minkov and Blagoev, 2012). A leader with global skills will adjust better to that situation (House et al. 2004). To strengthen that understanding of leadership, GLOBE imports the concept of psychological capital (psycap), by Luthans (2002). The fact that psycap is one of the ingredients of global leadership suggests that AL itself can also be considered a global leadership.

Nevertheless, despite this more or less universal character of leadership, existing cultural specificities should not be neglected and it is important to refer to them when we approach AL. Lawler and Ashman (2012) understand the formation of authenticity as a continuous process of the individual’s interaction, influenced by environmental elements. For Ford and Harding (2011), leaders adjust their authenticity in adapting to a collective context. This view is corroborated by research that identifies a relational and cultural basis for authenticity (Costas and Taheri 2012; Nyberg and Sveningsson, 2014).

Cross-cultural studies indicate that some elements of leadership can be considered universal (Zander and Romani 2004). However, the reasons activating these factors vary according to the values prized in each culture and should therefore be observed in discussions on the subject (Ensari and Murphy, 2003). Exemplifying the potential character of culture in the understanding and expression of authenticity, Zhang et al. (2012) made a study of this model of leadership in Chinese organizations. The results show that Chinese leaders’ main source of authenticity lies in the relationship with others. This result reinforces the roles of cultural and social elements in AL. In the same connection, Owusu-Bempah et al. (2014) study the perception of leaders and followers in Ghana and New Zealand as to the formation of AL. They conclude that the leadership model referred to is idiosyncratic but common among cultures.

In this context, Dorfman et al. (2012) argue that when assessing a cultural situation in the framework of a certain theoretical model, other dimensions can emerge. So when a group of people use different expressions to conceptualize the same beliefs, values, behaviors, and attitudes, there may be different perceptions in interpreting the items of a measure (Milfont and Fischer 2010). To respond to this challenge, the measurement of equivalence/invariance (ME/I) measures whether the instrument activates similar cognitive, perceptive, and interpretative processes in different groups (Vandenberg and Lance, 2000). That is, identifying whether the measurement invariance confirms that the latent construct can be generalized and is open to analogical interpretation between specific groups (Milfont and Fischer 2010). The non-existence of ME/I can imply differences in the instrument’s functioning in the populations studied and not necessarily reflect differences in the model of AL (Sirigatti et al. 2013). In these circumstances, the interpretations and inferences resulting from applying the instrument are affected negatively (Milfont and Fischer 2010).

This study compares the ME/I of ALQ for Brazil and Portugal. It contributes to understanding the universality and specificities of authenticity in the domain of leadership. Brazilian culture was influenced by Portuguese culture, among others, and both countries share the same language and some cultural traits. In the understanding of Hofstede (1980, 2001), common to the two cultures is the dimension of power distance. This form of hierarchy in society, in both countries, reinforces the valuation of individuals’ social status. On the other hand, compared to Portugal, Brazil has a culture tending slightly more towards individualism, masculinity, indulgence, and long-term orientation. The dimension of aversion to uncertainty is clearly more expressive in Portugal (Hofstede, 2001).

Participants

Both groups contain public and private firms, and the three economic sectors of activity are represented. However, the service sector predominates, since 90.3 % of Brazilian organizations and 38.6 % of Portuguese firms in this research belong to this sector. Regarding commerce, 4.5 % of Brazilian and 17.8 % of Portuguese firms in the sample belong to this sector. Industry accounts for 4.1 % of Brazilian and 17.2 % of Portuguese firms. The Portuguese sample contains 25.1 % of organizations with activities of a different nature from those mentioned above.

Concerning the size of organizations, 95.2 % of Brazilian companies in the sample are large and in the Portuguese sample 14.4 % are large. Regarding this aspect, 3 % of Brazilian companies and 60.5 % of Portuguese ones are medium-sized, with 0.8 % of Brazilian companies and 24.3 % of Portuguese ones being small. It should be noted that both samples were selected following the criterion of accessibility and convenience.

Measures

This study used the ALQ, conceived by Walumbwa et al. (2008). The research project, which the measurement invariance is a part of, was presented to the Mind Garden Institute (Copyright © 2007 Authentic Leadership Questionnaire by Avolio et al. 2007), and permission for scientific use of the ALQ was obtained. The ALQ was translated and adapted for Portugal by Rego et al. (2012).

The study validating the ALQ for Brazil includes a first stage of translation and back translation (Brislin, 1980). Four bilingual academics were instructed to translate the original English version of the ALQ, observing conceptual similarity. These translations were reviewed by three experts, fluent in English and experienced in the area of organizational psychology. A first version of the ALQ was elaborated and submitted to assessment by two focus groups (leaders and subordinates). Fine-tuning improvements were suggested by the groups concerning idiomatic expressions. An adapted version of the instrument was sent for back translation to two other academics also fluent in English. The same three experts analyzed the translations of the scale comparing them with the original version, defining the Brazilian version of the ALQ. To check the suitability of the instructions and items, a pilot application was carried out, in a sample of 42 followers, 23 women and 19 men. Thirty-eight employees were aged between 21 and 30. Regarding length of service, 35 had been in the same organization between 5 and 10 years. Those characteristics are similar to the sample that participated in the study to validate ALQ for Brazil.

The ALQ is composed of 16 items measured on a five-point Likert scale (from 0—never, to 4—frequently, if not always), divided in four dimensions: relational transparency (items 1 to 5, e.g., item 4: Tell you the hard truth), moral and ethics (items 6 to 9, e.g., item 6: Demonstrate beliefs that are consistent with actions), balanced processing (items 10 to 12, e.g., item 11: Analyze relevant data before coming to a decision), and self-awareness (items 13 to 16, e.g., item 16: Show I understand how specific actions impact others). The final scores and those for each dimension were calculated based on the average score of the corresponding items.

This study used the hetero-assessment version of ALQ, in which followers assess their immediate superiors. Participants also answered an inventory of socio-demographic data, with information also being gathered about organizations’ size and sector of activity.

Procedures

Data collection was carried out by duly trained Brazilian and Portuguese researchers. In Brazil, the group of researchers contacted companies and presented the research proposal. With their consent, the study was explained to the employees in the position of followers, and with their agreement, collection on location was arranged. Data collection took place in two ways: on location and online. In the former, collection was arranged previously with the participants and took place in the work environment. To employees in the organizations that opted for online collection, a link to the research was provided via e-mail (using SurveyMonkey). One week later, a reminder was sent to all participants who had not yet answered the questionnaire. In total, data was collected from 1019 employees.

The Portuguese sample was collected through research involving the University of Coimbra and the University of Évora. University students who were duly trained for this specific data collection contacted follower employees in different organizations. The study was explained in detail to each participant and doubts clarified, with collection taking place on location. A total of 842 questionnaires were gathered in.

All employees were informed of the voluntary nature of participation, the confidentiality and anonymity of the information given, and the possibility of withdrawing at any moment. They all signed an informed consent form and were given the researchers’ contact number for any clarification required.

Data analysis

The data were recorded and processed on a database in version 22.0 of SPSS and AMOS. Missing values, all MCAR (completely at random) and under 3 %, were substituted by the expectation maximization method (Ibrahim et al. 2005; Kline, 2011). The factorial validity of the ALQ was evaluated with confirmatory factor analysis (CFA) with AMOS (Arbuckle, 2013), maximum likelihood estimation method (Jöreskog and Sörbom 2004). The composite reliability and the average variance extracted for each factor were evaluated as described in Fornell and Larcker (1981).

In both samples, the skewness and kurtosis of all individual items did not show values excessively distant from those considered suitable for the assumption of normality (Finney and DiStefano, 2006; Kline, 2011): −0.66 ≤ sk ≤ −0.21 and −0.98 ≤ ku ≤ 0.05 for the Brazilian sample; −0.58 ≤ sk ≤ −0.20 and −0.68 ≤ ku ≤ −0.02 for the Portuguese sample. Goodness of fit was evaluated by the indices of X/df² (acceptable fit < 5; Arbuckle, 2013; Loehlin, 2004; Wheaton, 1987; X ² was irrelevant for N > 500; Bentler 1990; Schumacker and Lomax, 2010), normed of fit index (NFI; acceptable >.80; Bentler and Bonett, 1980; Schumacker and Lomax, 2010), comparative fit index (CFI; good fit >.90; Bentler, 1990), parsimony normed fit index (PNFI; very good fit ≥.80; Marôco, 2010; Mulaik et al. 1989), standardized root mean square residual (SRMR; appropriate fit <.08; Brown, 2015; Hu and Bentler, 1999), root mean square error of approximation (RMSEA; acceptable fit <.10, good fit <.08 for Marôco 2010; acceptable fit <.08 for Kline 2011, Schumacker and Lomax, 2010, and Steiger, 1990), and Akaike’s information criterion (Model AIC; applicable when maximum likelihood estimation is used, Burnham and Anderson, 1998; there is no rule of thumb; the values depend on actual datasets and the model—the model with the lowest AIC value is chosen; Arbuckle, 2013; Burnham et al. 2011). Reliability was estimated by Cronbach’s alpha (Cronbach, 1951). Measurement invariance was tested by multi-group confirmatory factor analyses (maximum likelihood estimation); configural, metric, scalar, and full uniqueness invariance values were compared with CFI, RMSEA, and SRMR, since they are not affected by the size of the sample or by the model’s complexity (Chen, 2007; Cheung and Rensvold, 2002; Meade et al. 2008).

Fit of the model proposed by Walumbwa et al. (2008)

Invariance tests for ALQ

To test the fit of each measurement invariance model, we turn to the values of the CFI and the RMSEA, since, in general, they are not influenced by the size of the sample nor by the complexity of the model and are not associated with global adjustment measures (Chen, 2007; Cheung and Rensvold, 2002). As can be seen in Table 4, CFI and RMSEA decreased by .001 from configural to metric invariance for both first- and second-order models, which is suitable for invariance, considering the guidelines for RMSEA and for CFI—changes in RMSEA of ≥.010 to .015 and in CFI of ≤−.005 to −.010 (Chen, 2007) or changes in CFI ≤ −.001 (Cheung and Rensvold, 2002) or ≤−.002 to −.008 (Meade et al. 2008), indicate invariance. Additionally, considering that a change in SRMR of ≥.025 to .030 is recommended for invariance testing of factor loadings (Chen, 2007) and our differences were only of ΔSRMR = .003 for the first-order model and ΔSRMR = .002 for the second-order model, metric invariance was shown.

The differences in RMSEA and CFI between scalar and metric invariance were ΔRMSEA = .005 and ΔCFI = −.020 for both first- and second-order models; these differences can also be accepted for invariance since, according to Chen (2007), for a cutoff point of .05, only a value equal to or less than −.005 in the CFI associated with a value equal to or greater than .010 in the RMSEA shows non-invariance (for a cutoff point of .01, change of <−.010 for CFI and >.015 for RMSEA). Also, considering SRMR, a change of ≥.005 to .010 is recommended for intercept level and residual variance tests (Chen, 2007), where our difference from scalar to metric invariance was ΔSRMR = .000, strengthening scalar invariance.

Full uniqueness measurement invariance (all parameters equal across groups) is rarely tested (Raju et al. 2002), because “this scenario of invariant measurement residuals is considered too restrictive and, in general, invariance analysis does not require residual invariance” (Marôco, 2010, p. 288). Despite this, for the first- and second-order models, the ΔCFI were −.020 and the ΔRMSEA were .003, which also indicates invariance for these higher order models, since non-invariance exists only when the decrement in model fits exceeds RMSEA ≥ .010 and CFI ≥ −.005 (Chen, 2007), which does not occur in either the first- or second-order model.

Authentic leadership model: comparison between Brazil and Portugal

Limitations

One limitation of this research lies in comparing the invariance of the ALQ in only two cultures. Considering cultural differences in studies on leadership is relevant in understanding how authenticity is developed and perceived in different nationalities. This study sought to contribute to cross-cultural research into AL, including a Brazilian sample, and also intended to address the shortage of studies in Latin-American countries. However, it is recommended that future studies extend the number of cultures analyzed.

A second limitation of this study concerns analysis of the ALQ only in the hetero-assessment version. The authors formulated a self-assessment version which was not considered in this study. The purpose of this research was to deepen understanding of how subordinates perceive authenticity in their superiors. But this aim should be extended and deepened with samples of leaders. Finally, we highlight the need to validate the results obtained here in longitudinal studies.