Personality and Social Sciences Factorial validity of the Utrecht Work Engagement Scale (UWES) across occupational groups in Norway CHRISTINA G. L. NERSTAD,1 ASTRID M. RICHARDSEN1 and MONICA MARTINUSSEN2 1 BI Norwegian School of Management, Norway 2 Center for Child and Adolescent Mental Health, University of Troms��, Norway Nerstad, C.G.L., Richardsen, A.M. & Martinussen, M. (2009). Factorial validity of the Utrecht Work Engagement Scale (UWES) across occupational groups in Norway. Scandinavian Journal of Psychology, 51, 326���333. The present study investigated the factorial validity of the Utrecht Work Engagement Scale (UWES) among 1266 participants from ten different occupational groups. Confirmatory factor analyses of the total sample, as well as multi-group analyses and analyses of each of the ten occupational groups separately, indicated that a three-dimensional model of both the UWES-17 and the short version, UWES-9, provided a better fit to the data than a one- and two-dimensional model. The results of multi-group analyses and analyses of each of the groups separately, indicated that with a few exceptions, the three- factor model of work engagement provided the best fit. Results indicated factorial invariance and the internal consistencies were acceptable. The fit of the UWES-9 was slightly better than the UWES-17. It is concluded that the Norwegian short version may be recommended over the UWES-17. Key words: Factorial validity, Utrecht Work Engagement Scale, work engagement, construct validity, factorial invariance, multi-group analysis. Christina G. L. Nerstad, Department of Leadership and Organization Management, Norwegian School of Management, N-0442 Oslo, Norway. Tel: + 47 46 41 07 58 fax: + 47 46 41 07 01 e-mail: christina.nerstad@bi.no INTRODUCTION A new trend within occupational health psychology is the focus on positive functioning and well-being, rather than the preoccupa- tion with the negative and pathological in the study of human behavior (Fineman, 2006). The concept of work engagement, which has received increased research attention in the past five years, represents such a shift in focus in the area of work function- ing. According to Maslach and Leiter (1997), engagement indi- cates lack of burnout, which has been recognized as an important individual and organizational problem (Leiter & Maslach, 2001 Maslach, 1993). Maslach and Leiter (1997) proposed that engage- ment and burnout are the bipolar endpoints of a continuum of worker well-being, affected by circumstances at work. Engage- ment can therefore be measured by calculating the opposite pat- tern of scores on the Maslach Burnout Inventory (MBI, Maslach, Jackson & Leiter, 1996). The use of the MBI for assessing both burnout and engagement has been criticized by Schaufeli and colleagues (Schaufeli, Salano- va, Gonza ��les-Roma �� & Bakker, 2002b), who have suggested an alternate approach to the definition and measurement of engage- ment. According to these researchers, engagement is still conceptu- alized as the positive antithesis of burnout, but it is measured in its own right. ������Engagement is defined as a persistent, pervasive and positive affective-motivational state of fulfillment in professionals������ (Llorens, Schaufeli, Bakker & Salanova, 2007, p. 3), consisting of three different dimensions: Vigor, Dedication and Absorption. Vigor is characterized by high levels of energy, an experience of mental resilience while working, a willingness to invest effort in one���s endeavor, as well as persistence in the face of difficulties (Hakanen, Bakker & Schaufeli, 2006 Schaufeli et al., 2002b). Dedication indicates a strong involvement in one���s vocation, characterized by a feeling of significance, enthusiasm, inspiration, and pride (Hakanen et al., 2006 Llorens et al., 2007). Absorption characterizes a state of full concentration and happy devotion towards an activity, in which time seems to pass quickly and indi- viduals often experience a difficulty in detaching themselves from the endeavor (Llorens et al., 2007 Schaufeli et al., 2002b). Bakker, Schaufeli, Leiter, and Taris (2008) highlight the fact that today���s organizations expect their employees to show initiative, to be pro- active, to take responsibility for their own development of achieve- ment behaviors at work, as well as to perform at high levels. In order to satisfy these organizational expectations, it is of great importance for employees to be absorbed, dedicated, and energetic in their work (Bakker & Schaufeli, 2008 Bakker et al., 2008). In other words, employees need to experience high levels of work engagement. Thus, one important argument for the added value of the concept of work engagement is its crucial role in developing the organization���s human capital and employee well-being and health (Schaufeli & Salanova, 2007). Empirical studies, typically based on the job demands-resources model (JD-R model), on work engagement, indicate a positive association between engagement and job resources, such as autonomy, social support, performance feedback, and task variety (e.g., Richardsen, Burke & Martinussen, 2006 Bakker, Hakanen, Demerouti, and Xanthopoulou, 2007). Personal resources, such as optimism or self-efficacy are also indi- cated to be important antecedents of work engagement (Bakker et al., 2008). Work engagement has also been associated with important organizational outcomes (Schaufeli & Salanova, 2007), such as in-role and extra-role performance, as well as job satisfac- tion and organizational commitment (e.g., Bakker, Demerouti & Verbeke, 2004 Halbesleben & Wheeler, 2008 Hallberg & Schau- feli, 2006). In order to assess engagement, a self-report question- naire has been developed: the Utrecht Work Engagement Scale (UWES Schaufeli et al., 2002b). The scale consists of 17 items and includes the three dimensions of work engagement (Vigor: six �� 2009 The Authors. Journal compilation �� 2009 The Scandinavian Psychological Associations. Published by Blackwell Publishing Ltd., 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA. ISSN 0036-5564. Scandinavian Journal of Psychology, 2010, 51, 326���333 DOI: 10.1111/j.1467-9450.2009.00770.x

items Dedication: five items Absorption: six items). The test scores appear to be internally consistent (Cronbach alphas range between 0.80 and 0.90), and the instrument has shown good facto- rial validity across occupations and cross-nationally (Schaufeli, Bakker & Salanova, 2006 Schaufeli et al., 2002b). The three- dimensional structure has been found to be invariant across samples from Portugal, Spain, and the Netherlands (Schaufeli, Mart�� ��nez, Marques-Pinto, Salanova & Bakker, 2002a). However, Sonnentag (2003), did not find a clear three-factor dimensionality of work engagement in her study of German employees from six public service organizations, and thus she decided to apply a one-dimensional approach measuring work engagement with the total score of UWES. The three engagement scales often correlate highly (Schaufeli et al., 2002b, 2006), suggesting that a one-dimensional measure could be used. A short version of the UWES has recently been developed (Schaufeli et al., 2006). The UWES-9 consists of nine items (three items measuring vigor, three items measuring dedication, and three items measuring absorption). Findings from confirma- tory factor analyses across different countries have indicated acceptable psychometric properties (Schaufeli et al., 2006). Hallberg and Schaufeli (2006) conducted a study using the Swedish short version of UWES, and found support for both a three-dimensional and one-dimensional structure of work engage- ment. Latent inter-correlations (between 0.88 and 0.99) made the authors suggest that the one-dimensional structure could be favored (Hallberg & Schaufeli, 2006). Because there are inconsis- tencies regarding the dimensionality of work engagement within different countries, it is necessary to test the dimensionality of the UWES in other languages. The results so far show that further testing of the UWES is necessary in order to establish the utility of the measure. Macey and Schneider (2008) have pointed out the ambiguous role the concept of work engagement has among both practitio- ners and academic researchers, and argue that the components of work engagement have not been rigorously conceptualized. In order to clarify the added value of the concept, more research is needed on how the concept can be discriminated from other but similar concepts (Newman & Harrison, 2008). Before the measure can be reliably used in Norwegian settings, it seems necessary to test the factorial validity of the Norwegian version of both the UWES-17 and the UWES-9. The factorial structure and psychometric properties of both the 17-item version and the 9-item version across different occu- pations have to the best of our knowledge never been investi- gated in Norway. The purpose of this study was therefore to report on the factorial validity of the UWES-17 and the UWES- 9 using confirmatory factor analyses (CFA) in a Norwegian sample of 10 occupational groups (physiotherapists, child welfare workers, social workers, social rehabilitation workers, nurses, nurse assistants, teachers, journalists, police officers and air traffic controllers). Structural equation modeling is a statisti- cal technique that includes aspects of both multiple regression and factor analysis in order to estimate a series of interrelated and dependent relationships at the same time. CFA represents the possibility to test whether the theoretical reality is consistent with actual reality. The following propositions were addressed for this present study: Hypothesis 1: Confirmatory factor analysis of the Norwegian translation of the UWES-17 and UWES-9 will support a three factor structure. Hypothesis 2: The three-factor structure of the UWES-17 and the UWES-9 will be invariant across occupations. Hypothesis 3: The reliabilities of both the UWES-17 and the UWES-9 will be satisfactory for all three dimensions of work engagement. METHOD Participants Participants (N = 1266) consisted of 10 different professions including 158 physiotherapists, 130 child welfare workers, 158 social workers, 116 social rehabilitation workers, 109 nurses, 46 nurse assistants, 271 teachers, 150 police officers, 78 air traffic controllers, and 50 journalists. Ages ranged from 19 to 72 (M = 40.8, SD =10.4), and the proportion of women was 67%. Measures Work Engagement. Work engagement was assessed by the Utrecht Work Engagement Scale (UWES Schaufeli et al., 2002b). The questionnaire was translated from English to Norwegian based on the double Transla- tion-Back-Translation technique. The UWES consists of 17 items rated on a seven-point scale ranging from ������never������ (0) to ������always/every day������ (6). The items are divided into Vigor (VI, six items e.g., ������At my job, I feel strong and vigorous.������), Dedication (DE, five items e.g., ������I am enthusiastic about my job.������), and Absorption (AB, six items e.g., ������When I am working, I forget everything else around me.������). The short version of the UWES consists of 9 of the 17 items of the UWES, with three items for each of the three factors (VI1, VI2, VI3 DE2, DE3, DE4 and AB3, AB4, AB5). Procedure The data for this study were collected at the six-month follow-up of an investigation of antecedents and consequences of professional burnout across different occupational groups. The UWES was included to test the factorial validity of the Norwegian translation of the scale. Data were collected using questionnaires which were sent to the participants��� home addresses. The participants had given written consent to participate in the study. Envelopes containing a questionnaire, cover letter, a stamped envelope with a return address were sent to 2284 participants in total. After 2���3 weeks, a reminder letter went out to all participants. A total of 1349 questionnaires were returned, representing a response rate of 55.4%. Questionnaires with missing data in excess of 25% were excluded, leaving a final data set of 1266 questionnaires. Remaining missing values were replaced using a mean substitution. Statistical analyses Structural equation modeling (SEM) using the Lisrel 8.80 (Jo ��reskog & So ��rbom, 1993) was used to test the fit of the factor analytic models. Using similar analysis procedures as other studies (e.g., Hallberg & Schaufeli, 2006 Schaufeli et al., 2002a, 2006), we first tested the fit of increasingly less restricted models on the total sample, and then tested the fit of the best model across occupational groups. The correlation matrix and asymptotic covariance matrix of the items was examined using the robust maximum likelihood (RML) method. Fol- lowing Byrnes��� (1998) suggestions for multi-group analysis, this analysis was based on the correlation matrix. The SEM analysis assesses the fac- tor structure with the overall v2 value as an index of model fit (Byrne, 1998). However, in large samples, the chi-square statistic is very powerful Scand J Psychol 51 (2010) Factorial validity of the UWES 327 �� 2009 The Authors. Journal compilation �� 2009 The Scandinavian Psychological Associations.