Test of the Bifactor Model of Job-Related Affective Well-Being Research Reports Test of the Bifactor Model of Job-Related Affective Well-Being Mariola Laguna* a, Emilia Mielniczuk a, Wiktor Razmus a [a] Institute of Psychology, The John Paul II Catholic University of Lublin, Lublin, Poland. Abstract The multidimensional measure of the job-related affective well-being developed by Warr (1990) is a frequently used tool estimating affect in the work context. Alternative factorial models of this measure were tested in previous studies. Recently a bifactor model has been proposed as alternative factorial structure recommended for multifaceted constructs. It allows capturing the global aspect of the construct along with the specificity of its subdimensions. We conducted two studies to test a bifactor model on Warr’s measure and to compare it to factor models proposed in earlier studies. This bifactor model identified one general factor in addition to four unique factors. Two studies were conducted among employees (Study 1; N = 869) and entrepreneurs (Study 2; N = 204). Results of both studies corroborate a four correlated factors model as superior to the bifactor model. The model with four unique but correlated factors representing anxiety, comfort, depression, and enthusiasm is a good representation of job-related affective well-being measured by Warr’s instrument, both in a sample of employees and entrepreneurs. Keywords: affect, well-being, bifactor model, employees, entrepreneurs Europe's Journal of Psychology, 2019, Vol. 15(2), 342–357, https://doi.org/10.5964/ejop.v15i2.1632 Received: 2018-04-06. Accepted: 2018-05-10. Published (VoR): 2019-06-07. Handling Editors: Vlad Glăveanu, Department of Psychology, Webster University Geneva, Geneva, Switzerland; Tiziana Lanciano, University of Bari "A. Moro", Bari, Italy *Corresponding author at: Institute of Psychology, The John Paul II Catholic University of Lublin, Al. Raclawickie 14, 20-950 Lublin, Poland. E-mail: laguna@kul.pl This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 International License, CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Affective dimensions of work have become an important research area in work and organizational psychology (Brief & Weiss, 2002; Mäkikangas, Hyvönen, Leskinen, Kinnunen, & Feldt, 2011; Seo, Barrett, & Bartunek, 2004), as well as in the psychology of entrepreneurship (Baron, 2008; Gorgievski & Stephan, 2016; Stephan, 2018). Studies have shown that people’s affective states play a central role in explaining links between work environments, employees behavioral responses and work performance (Parker, 2014). With the increased in- terest in job-related affective well-being comes a renewed interest in the validity of its measurement instru- ments. It is vital to accurately measure individuals’ affective experiences, in order to reach valid conclusions for explaining their role in a work environment (Mäkikangas, Feldt, & Kinnunen, 2007). One of the most widely used tools for testing affect in the work context is Warr’s (1990) job-related affective well-being measure (Laguna, Mielniczuk, Razmus, Moriano, & Gorgievski, 2017; Mäkikangas et al., 2007; Salanova, Llorens, & Schaufeli, 2011; Van Horn, Taris, Schaufeli, & Schreurs, 2004). As Warr has not proposed a name for this measure, authors customarily refer to it as “Warr’s (1990) job-related affective well-being meas- ure,” “the measure developed by Warr (1990),” or like phrases. This name may seem similar to the name of the Europe's Journal of Psychology ejop.psychopen.eu | 1841-0413 https://creativecommons.org/licenses/by/4.0/ https://creativecommons.org/licenses/by/4.0/ https://ejop.psychopen.eu/ https://ejop.psychopen.eu/ https://www.psychopen.eu/ Job-Related Affective Well-Being Scale (JAWS) developed by Van Katwyk et al. (2000), however, these are two different instruments. To clarify, JAWS is also based on Warr’s model of affective well-being at work, which was primarily operationalized by his own measure. Previous analyses (Daniels, Brough, Guppy, Peters-Bean, & Weatherstone, 1997; Gonçalves & Neves, 2011; Mäkikangas et al., 2007; Sevastos, Smith, & Cordery, 1992) have confirmed the validity and reliability of the instrument developed by Warr (1990). What has not been tested, however, is the validity of a potential bifactor model, which has recently been postu- lated as a recommended alternative method for testing multifaceted constructs (Chen, Hayes, Carver, Laurenceau, & Zhang, 2012). Research has demonstrated that bifactor models which identify one general fac- tor in addition to multiple unique factors, allow the superior representation of different personality constructs (Alessandri, Vecchione, Eisenberg, & Łaguna, 2015; Chen, Watson, Biderman, & Ghorbani, 2016), as well as of well-being (Chen, Jing, Hayes, & Lee, 2013) and affect (Leue & Beauducel, 2011). Such bifactor representa- tion allows capturing the global aspect of the construct along with the specificity of its subdimensions. This sug- gests that a bifactor model may be a better representation of job-related affect than previously tested models. Moreover, for some research purposes, partialing out the general factor may be a preferred way of analyzing data. In the case of well-being, we can expect that such a general factor may be influenced by stable, internal processes rather than momentary, environmental factors. Therefore, the aim of our research was to determine whether a bifactor model may represent properly, and bet- ter than alternative models, job-related affective well-being as captured by the measure developed by Warr (1990). Our research tested the factorial structure of this measure not only on a sample of employees, but also, for the first time, on a sample of entrepreneurs – people who established themselves and manage their own firms (Baron, 2007). Using relatively large, diverse samples, and a sound analytic framework, we were able to comprehensively test the factorial structure of the widely-used measure proposed by Warr (1990) and provide recommendations how the results can be interpreted in future studies and counseling practices. Job-Related Affective Well-Being Measure Affect as consciously accessible feelings evident in moods and emotions (Fredrickson, 2001; Walter & Bruch, 2008) is a vital sign of an individual’s subjective well-being (Chen, Jing, et al., 2013). It also constitutes the most central dimension of occupational subjective well-being (Van Horn et al., 2004) which has become of in- terest in work and organizational psychology (Mäkikangas et al., 2011; Seo et al., 2004), in career counseling (Cuyper, Philippaers, Vanhercke, & Witte, 2017; Gati et al., 2011; Hutchison & Gerstein, 2017; Konstam, Celen- Demirtas, Tomek, & Sweeney, 2015), and in entrepreneurship (Stephan, 2018). There is growing evidence which shows that employees’ job-related affect is linked to their work engagement (Bledow, Schmitt, Frese, & Kühnel, 2011; Salanova et al., 2011), direction, intensity, and persistence of actions (Seo et al., 2004). Affect also impacts performance at work (Beal, Weiss, Barros, & MacDermid, 2005), as well as satisfaction from per- formed tasks and responsibility for those tasks (Isen & Reeve, 2005). Affect is important not only in the activity of employees, but also that of entrepreneurs (Baron, 2008; Stephan, 2018) and this is related to their work en- gagement and goal attainment (Laguna, Alessandri, & Caprara, 2016; Laguna, Razmus, & Żaliński, 2017). All these findings show that job related affect should be taken into account when discussing career choices, job satisfaction and performance. Laguna, Mielniczuk, & Razmus 343 Europe's Journal of Psychology 2019, Vol. 15(2), 342–357 https://doi.org/10.5964/ejop.v15i2.1632 https://www.psychopen.eu/ This broad interest in job-related affective well-being raises the question of how to accurately estimate it to rep- resent the structure of an individual’s emotional experience (Mäkikangas et al., 2007). A conceptualization of affective well-being, accompanied by a measurement tool, was proposed by Warr (1990). The measure is based on the circumplex model of affect (Russell, 1980) and conceptualizes well-being as a job-specific rather than as a context-free phenomenon. According to Warr (1990), job-related affective well-being is defined by two principal dimensions: pleasure (negative—positive) and arousal (low—high activation level). The measure is in- tended to capture the content and intensity of job-related emotional states: Comfort (positive affect of low arousal), Enthusiasm (positive affect of high arousal), Depression (negative affect of low arousal), and Anxiety (negative affect of high arousal). The instrument consists of 12 adjectives (e.g., calm, worried) which are answered in relation to the job: “Think- ing of the past few weeks, how much of the time has your job made you feel each of the following?” It is also possible to measure context-free well-being with different phrasing of the instruction. The Factorial Structure of the Job-Related Affective Well-Being Measure The factorial structure of the measure proposed by Warr (1990) was assessed in a several studies (Cifre & Salanova, 2002; Daniels et al., 1997; Gonçalves & Neves, 2011; Laguna, Mielniczuk, et al., 2017; Mäkikangas et al., 2007; Mielniczuk & Laguna, 2018; Sevastos et al., 1992; Van Horn et al., 2004; Warr, 1990) in which six alternative models were tested (for details see Appendix). Model 1 contained four correlated factors that were postulated in Warr’s model (1990): Anxiety, Comfort, Depression, and Enthusiasm. Model 2 posited two correla- ted factors, capturing two dimensions: Anxiety–Comfort and Depression–Enthusiasm. Model 3 posited two cor- related factors: Positive Affect and Negative Affect. Model 4 represented three dimensions: Positive Affect, Negative Affect, and Pleasantness–Unpleasantness. Model 5 was a single factor model capturing affective well-being as a single common factor. The hierarchical Model 6 captured the Global Affective Well-being as a second-order factor and four first-order latent factors. Investigations of the general factor of job-related affective well-being (Gonçalves & Neves, 2011; Van Horn et al., 2004) raise the question of how well-being should be understood: Does it primarily refer to specific affective dimensions, or should it be considered as a more global phenomenon that captures the general subjective esti- mation of whether a person feels well or unwell at work? Both conceptualizations are of interest in counselling and in work and organizational psychology research. A potential solution may be the consideration of a bifactor model as an alternative to the single factor or hier- archical models (Chen, Hayes, et al., 2012). The bifactor model can be considered when (a) there is a general factor accounting for the commonality of the items; (b) there are multiple domain specific factors; and (c) the common factor together with the domain-specific factors can be interesting for researchers (Chen, Hayes, et al., 2012). This proposition was validated as a valuable and unsurpassed alternative in explaining dimensionali- ty of personality traits (Chen et al., 2016), self-esteem (Alessandri et al., 2015), context-free well-being (Chen, Jing, et al., 2013), context-free affect (Leue & Beauducel, 2011). This would suggest that the bifactor model may be a better estimate of job-related affective well-being than previously tested models. Hence, we propose a new model: Model 7 as a bifactor representation of job-related affective well-being (Fig- ure 1). It captures the global phenomenon of job-related well-being in addition to specific affective dimensions: Anxiety, Comfort, Depression, and Enthusiasm. Model 7 is proposed as an alternative way of representing the Bifactor Model of Job-Related Affective Well-Being 344 Europe's Journal of Psychology 2019, Vol. 15(2), 342–357 https://doi.org/10.5964/ejop.v15i2.1632 https://www.psychopen.eu/ core features of an underlying global latent dimension than a hierarchical model (Gonçalves & Neves, 2011) or a single factor model (Van Horn et al., 2004). Figure 1. The bifactor model for the Warr’s (1990) measure (Model 7). An attractive feature of this bifactor interpretation is that it casts the specific factors in a manner consistent with previous empirical research that confirmed four specific affective well-being dimensions from Warr’s item pool. However, the simple procedure of summing the items representing each dimension left room for possible con- tamination of the scale scores, since variance from general job-related affective well-being was not partialled- out from the scores. This problem can be avoided by the bifactor approach (Chen, Hayes, et al., 2012). Never- theless, even if successfully applied towards other measures (Alessandri et al., 2015; Chen et al., 2016; Leue & Beauducel, 2011), it is not guaranteed that this approach will be fruitful in estimating emotional experiences at work. Therefore, the aim of our research was to validate the bifactor model on Warr’s instrument (1990) in com- parison to alternative models established in previous studies. The alternative measurement models considered in our research are depicted in Appendix. We conducted two studies to investigate the relative fit of seven alternative factor models described above, in- cluding newly proposed the bifactor model. In both studies, we relied on relatively large samples which allowed us to use Structural Equation Modeling (SEM) in order to address several important psychometric questions pertaining to the assessment of affective well-being. We focused on the dimensionality of the instrument, based on the accepted methodological procedures (Chen, Hayes, et al., 2012). The relative fit of the hypothesized al- ternative models was iteratively tested in samples of employees (Study 1) and entrepreneurs (Study 2). Study 1 The aim of this study was to examine the relative fit of seven alternative models, attesting the factorial structure of the job-related affective well-being measure (Warr, 1990) on a sample of employees. Laguna, Mielniczuk, & Razmus 345 Europe's Journal of Psychology 2019, Vol. 15(2), 342–357 https://doi.org/10.5964/ejop.v15i2.1632 https://www.psychopen.eu/ Method Procedure The data was gathered using the paper-and-pencil questionnaires (62.5% of the sample) and on-line question- naires (37.5%). In additional analyses, measurement invariance for the sampling method was tested and full scalar invariance was achieved. Participants were invited to take part in the study through social and business networks using direct solicitation. Paper-and-pencil questionnaires were delivered to participants at their homes or at work and collected from them after completion. Participation in the study was voluntary and the partici- pants did not receive any reward. The confidentiality and anonymity of data was ensured. We computed the minimum sample size for the model structure (N = 187) as a function of the required significance level of α = .05, and the desired statistical power of 1-β = .80. Participants The sample included 818 employees from Poland, of whom 59.5% were women, and it far exceeded the mini- mum sample size. The participants were 19 to74 years old (M = 33.92, SD = 10.13), and worked full time (61.8%), part time (30.8%), or on other forms of work contracts (7.3%). They worked in services and trade (45%), production (15.1%), construction (1.7%), or in other sectors (38.2%). The sample generally matched the characteristics of the population of employees in Poland, except that it included a few more women (54.9% men and 45.1% of woman in the population; GUS 2017). Measure Job-related affective well-being was evaluated using the Polish version (Mielniczuk & Laguna, 2018) of the measure proposed by Warr (1990). It consisted of 12 items and answers were given on a 6-point scale (1 = never; 2 = occasionally; 3 = some of the time; 4 = much of the time; 5 = most of the time; 6 = all of the time). Data Analysis Strategy The data were analyzed by means of confirmatory factor analysis (CFA) using Amos (Arbuckle, 2005). The pa- rameters were estimated using the maximum likelihood method. Missing data, which did not exceed 1.5% in any item, were handled using regression imputation. We examined a series of seven measurement models. To assess model fit, we used the χ2 Goodness-of-Fit Statistic, the Comparative Fit Index (CFI), the Root Mean Square Error of Approximation (RMSEA), the Stand- ardized Root Mean Square Residual (SRMR), and the Akaike Information Criterion (AIC). Values below .08 for RMSEA and SRMS indicate an acceptable model fit (Kline, 2005), while values below .10 indicate mediocre model fit (Hooper, Coughlan, & Mullen, 2008). For the CFI values greater than .90 indicate a good fit (Brown, 2006; Hooper et al., 2008). The lower the AIC index, the better is the fit of the model to the data (Brown, 2006). To test differences between alternative models, we calculated the chi-square difference tests (Δχ2) and differ- ence in CFI (ΔCFI; Brown, 2006; Kline, 2005). An absolute difference in CFI that is less than .01 (ΔCFI < .01) would indicate a lack of significant difference in model fit (Cheung & Rensvold, 2002). Results and Discussion Examination of alternative models showed that only the four correlated factors model (Model 1) obtained ac- ceptable fit in all fit indices (Table 1). Bifactor Model of Job-Related Affective Well-Being 346 Europe's Journal of Psychology 2019, Vol. 15(2), 342–357 https://doi.org/10.5964/ejop.v15i2.1632 https://www.psychopen.eu/ Ta bl e 1 G oo dn es s- of -F it In di ce s of A lte rn at iv e M ea su re m en t M od el s in S am pl e of E m pl oy ee s (N = 8 69 ) M od el χ2 df p R M S E A S R M R C FI A IC M od el co m pa ri so n Δ χ2 Δ df p Δ C FI M od el 1 . F ou r co rr el at ed fa ct or s: A nx ie ty , C om fo rt , D ep re ss io n, E nt hu si as m 28 6. 04 48 < .0 01 .0 76 .0 43 .9 60 37 0. 04 - - - - - M od el 2 . T w o co rr el at ed fa ct or s: A nx ie ty -C om fo rt , D ep re ss io n- E nt hu si as m 15 60 .3 6 53 < .0 01 .1 81 .1 23 .7 48 16 10 .3 6 M 7 vs M 1 12 74 .3 2 5 < .0 01 .2 1 M od el 3 . T w o co rr el at ed fa ct or s: P os iti ve A ffe ct , N eg at iv e A ffe ct 97 9. 14 53 < .0 01 .1 42 .0 83 .8 45 10 29 .1 4 M 2 vs M 1 69 3. 10 5 < .0 01 .1 2 M od el 4 . T hr ee fa ct or s: P os iti ve A ffe ct , N eg at iv e A ffe ct , P le as an tn es s- U np le as an tn es s 12 70 .9 8 51 < .0 01 .1 66 .1 07 .7 96 13 24 .9 8 M 8 vs M 1 98 4. 94 3 < .0 01 .1 6 M od el 5 . S in gl e fa ct or : W el l-b ei ng 18 55 .5 3 54 < .0 01 .1 96 .1 16 .6 99 19 03 .5 3 M 4 vs M 1 15 69 .4 9 6 < .0 01 .2 6 M od el 6 . H ie ra rc hi ca l: fo ur fa ct or s an d hi gh er o rd er fa ct or W el l-b ei ng 41 7. 98 50 < .0 01 .0 92 .0 73 .9 39 47 3. 98 M 5 vs M 1 13 1. 94 2 < .0 01 .0 2 M od el 7 . B ifa ct or : f ou r fa ct or s an d W el l-b ei ng 37 2. 89 48 < .0 01 .0 95 .0 69 .9 45 46 8. 89 M 6 vs M 1 86 .8 5 0 < .0 01 .0 2 N ot e. R M S E A = R oo t M ea n S qu ar e E rr or o f A pp ro xi m at io n; S R M R = S ta nd ar di ze d R oo t M ea n S qu ar e R es id ua l; C F I = C om pa ra tiv e F it; A IC = A ka ik e In fo rm at io n C rit er io n. Laguna, Mielniczuk, & Razmus 347 Europe's Journal of Psychology 2019, Vol. 15(2), 342–357 https://doi.org/10.5964/ejop.v15i2.1632 https://www.psychopen.eu/ None of the alternative models reached the criteria of acceptable fit (Brown, 2006; Kline, 2005), with the excep- tion of SRMR for a bifactor model (SRMR = .069), which is inconsistent with other fit indices. Moreover, tests of differences (Δχ2 < .05; ΔCFI > .01) confirmed that the four correlated factors model fit the data significantly bet- ter than any other alternative model, including the bifactor model (lower χ2 value and ΔCFI > .01). The bifactor solution is therefore not confirmed in this study. In this best fitting four correlated factors model (Model 1) regression weights were high, varied between .87 and .69 for each of four latent factors. Descriptive statistics, reliability and correlations between scales are pre- sented in Table 2. Cronbach’s alpha range from .78 to .90 indicated satisfactory to high internal consistency of all scales. Table 2 Scales Reliability, Descriptive Statistics and Correlations Between Scales in Samples of Employees (Study 1, N = 869) and Entrepreneurs (Study 2, N = 204) Employees α M SD 1 2 3 Entrepreneurs α M SD 1 2 3 1. Anxiety .84 2.89 1.01 1. Anxiety .82 2.71 .79 2. Comfort .83 3.53 1.00 -.61 2. Comfort .78 3.25 .71 -.54 3. Depression .79 2.12 0.92 .55 -.48 3. Depression .81 1.60 .59 .67 -.53 4. Enthusiasm .89 3.81 1.09 -.39 .69 -.40 4. Enthusiasm .90 4.27 .92 -.35 .78 -.40 Note. All correlations are statistically significant at the level of at least p < .01 (two-tailed). In conclusion, the results of this study demonstrated that the model representing job-related affective well-being as comprised of four correlated dimensions is better representation of emotional experiences of employees than the bifactor model. Study 2 In Study 2, the bifactor model was again compared with alternative models of job-related affective well-being (Warr, 1990) on a sample of entrepreneurs. This allowed us to check whether the findings from Study 1 on em- ployees could be generalized to other samples. As entrepreneurs, we consider people who established and manage their firms themselves (Baron, 2007), and were, therefore, not only managers, but also new organiza- tion builders. Method Procedure The snowball sampling method was used where the initially-contacted entrepreneurs from different regions of Poland were asked to provide contacts to two other entrepreneurs. The participants were invited to take part in the study if they fulfilled four criteria, being: (1) founders, (2) owners, and (3) managers of their firms, and (4) had firms that survived on the market for at least one year. The respondents were asked to fill in the paper-and- pencil questionnaires. Participation in the study was voluntary and the participants did not receive any reward. The data were anonymized to ensure confidentiality. The minimum sample size (N = 187) was calculated simi- larly as in Study 1. Bifactor Model of Job-Related Affective Well-Being 348 Europe's Journal of Psychology 2019, Vol. 15(2), 342–357 https://doi.org/10.5964/ejop.v15i2.1632 https://www.psychopen.eu/ Participants A sample of 204 entrepreneurs took part in the study, of whom 123 were male (61.9%). It exceeded the mini- mum sample size. The participants were 22-71 years old (M = 44.07, SD = 11.53). The firms they managed had existed for between one and 47 years (M = 10.10, SD = 8.36). Sectors of operation included services (79.9%), production (11.3%), and construction (8.8%). All of these firms were small and medium-sized enterprises, em- ploying up to 82 employees for full-time contracts (M = 6.29, SD = 13.12), and up to 32 employees for part-time contracts (M = 1.55, SD = 3.43). The sample matched the characteristics of a population of entrepreneurs in Poland (Tarnawa & Zadura-Lichota, 2013). Measure Job-related affective well-being was assessed using the Polish version (Mielniczuk & Laguna, 2018) of the measure proposed by Warr (1990). Data Analysis Strategy We used Amos (Arbuckle, 2005) to perform SEM analyses; the maximum likelihood estimation was applied. Missing data, not exceeding 1.5% in any item, were handled using the regression imputation method. CFA was applied to examine alternative models. Results and Discussion To further validate the results of Study 1, we examined the seven alternative models again. CFA results (Table 3) show that the model with four correlated factors (Model 1) obtained an acceptable fit in all fit indices (χ2(48) = 129.601, p < .001, CFI = .96, AIC = 213.60, RMSEA = .092, SRMR = .059), except for RMSEA which exceeded the criterion value of .08. Neither the bifactor model nor any of the alternative models reached the acceptable level of model fit. Moreover, the comparison of differences (Δχ2 < .05; ΔCFI > .01) between Model 1 and all other alternative models show that it fits the data significantly better. Therefore, again, the bifactor model was not supported. Regression weights for each of the four correlated factors varied between .86 and .62. Descriptive statistics and correlations between scales are presented in Table 2. Cronbach’s alpha for the four scales ranged between .78 and .90, indicating satisfactory to high internal consistency of four scales. Laguna, Mielniczuk, & Razmus 349 Europe's Journal of Psychology 2019, Vol. 15(2), 342–357 https://doi.org/10.5964/ejop.v15i2.1632 https://www.psychopen.eu/ Ta bl e 3 G oo dn es s- of -F it In di ce s of A lte rn at iv e M ea su re m en t M od el s in S am pl e of E nt re pr en eu rs ( N = 2 04 ) M od el χ2 df p R M S E A S R M R C FI A IC M od el co m pa ri so n Δ χ2 Δ df p Δ C FI M od el 1 . F ou r co rr el at ed fa ct or s: A nx ie ty , C om fo rt , D ep re ss io n, E nt hu si as m 12 9. 60 1 48 < .0 01 .0 92 .0 59 .9 56 21 3. 60 1 - - - - - M od el 2 . T w o co rr el at ed fa ct or s: A nx ie ty -C om fo rt , D ep re ss io n- E nt hu si as m 49 2. 62 6 53 < .0 01 .2 02 .1 52 .6 57 54 2. 62 6 M 2 vs M 1 36 3. 02 5 5 < .0 01 .2 79 M od el 3 . T w o co rr el at ed fa ct or s: P os iti ve A ffe ct , N eg at iv e A ffe ct 26 3. 47 3 53 < .0 01 .1 40 .0 88 .8 36 33 7. 47 3 M 3 vs M 1 13 3. 87 2 5 < .0 01 .1 00 M od el 4 . T hr ee fa ct or s: P os iti ve A ffe ct , N eg at iv e A ffe ct , P le as an tn es s- U np le as an tn es s 43 0. 12 0 51 < .0 01 .1 91 .1 35 .7 04 48 4. 12 0 M 4 vs M 1 30 0. 51 9 3 < .0 01 .2 32 M od el 5 . S in gl e fa ct or : W el l-b ei ng 54 8. 67 2 54 < .0 01 .2 12 .1 41 .6 14 59 6. 67 2 M 5 vs M 1 41 9. 07 1 6 < .0 01 .3 22 M od el 6 . H ie ra rc hi ca l: fo ur fa ct or s an d hi gh er o rd er fa ct or W el l-b ei ng 16 2. 70 4 50 < .0 01 .1 05 .0 99 .9 12 21 8. 70 4 M 6 vs M 1 33 .1 03 2 < .0 01 .0 24 M od el 7 . B ifa ct or : f ou r fa ct or s an d W el l-b ei ng 13 8. 42 1 42 < .0 01 .1 06 .0 85 .9 25 23 4. 42 1 M 7 vs M 1 8. 82 0 2 .0 12 .0 11 N ot e. R M S E A = R oo t M ea n S qu ar e E rr or o f A pp ro xi m at io n; S R M R = S ta nd ar di ze d R oo t M ea n S qu ar e R es id ua l; C F I = C om pa ra tiv e F it; A IC = A ka ik e In fo rm at io n C rit er io n. Bifactor Model of Job-Related Affective Well-Being 350 Europe's Journal of Psychology 2019, Vol. 15(2), 342–357 https://doi.org/10.5964/ejop.v15i2.1632 https://www.psychopen.eu/ Discussion A bifactor solution to factorial structure of measures was recently postulated for testing multifaceted constructs (Chen, Hayes, et al., 2012). Such bifactor models allow to identify a general factor in addition to multiple unique factors, which are of interest in personality psychology (Alessandri et al., 2015; Chen et al., 2016), and in well- being research (Chen, Jing, et al., 2013; Leue & Beauducel, 2011). We expected that such bifactor model may be also a good representation of job-related affective well-being. Results of our two studies, however, did not confirm this expectation. Our research findings supported the four correlated factors model over the proposed bifactor model. According to our results, neither the bifactor, nor a single factor or hierarchical model fit the data better than the four factor model; this is consistent with the results of other studies (Gonçalves & Neves, 2011; Laguna, Mielniczuk, et al., 2017; Mäkikangas et al., 2007; Mielniczuk & Laguna, 2018). All of those findings confirm that job-related affective well-being, as measured by Warr’s instrument (1990) capture four specific, in- terrelated affective dimensions. This altogether supports the job-related affective well-being dimensionality, and subsequently, the circumplex model of affect (Russell, 1980). It postulates that each affective experience is the result of a combination of two independent neurophysiological systems of valence and arousal, and is then in- terpreted as representing rather a particular affects (Posner, Russell, & Peterson, 2005), and not a global affec- tivity. Therefore, even if in case of self-esteem (Alessandri et al., 2015), context-free well-being (Chen, Jing, et al., 2013), context-free affect (Leue & Beauducel, 2011) the global aspect of the construct along with the specif- icity of its subdimensions can be distinguished, in case of job-related affective well-being (as measured by the Warr instrument) a global affectivity dimension is not confirmed. Strengths and Limitations One of the strengths of this research lies in the variation of samples included in our two studies. We involved a large sample of employees of different professions and, for the first time, made an evaluation of psychometric properties of the measure on the sample of entrepreneurs. This validates the instrument as useful in research on the role of affect in entrepreneurship – a dynamically evolving field of study (Baron, 2008; Laguna et al., 2016; Stephan, 2018). Another strong point of this study is that we tested a broad spectrum of alternative struc- tural models. We checked all potential models tested in previous analyses and added a new, theoretical and psychometric proposition of the bifactor model (Chen, Hayes, et al., 2012). The studies were, however, not free of some shortcomings. First, our analyses were based on samples coming from one country, Poland. Even if the factorial structure of the measure was confirmed in other samples of Pol- ish employees (different than one presented in this paper, i.e., employees of small and medium size companies Laguna, Mielniczuk, et al., 2017; Mielniczuk & Laguna, 2018) and shows cross-culture invariance (Laguna, Mielniczuk, et al., 2017; Mielniczuk & Laguna, 2018), the bifactor model should be examined further in other cultural contexts. Conclusions In conclusion, our results suggest that in future analyses four job-related affective well-being dimensions of the Warr (1990) measure should be applied rather than the bifactor model. The four scales show satisfactory relia- bility in samples of employees of different profession and of entrepreneurs. Laguna, Mielniczuk, & Razmus 351 Europe's Journal of Psychology 2019, Vol. 15(2), 342–357 https://doi.org/10.5964/ejop.v15i2.1632 https://www.psychopen.eu/ As work activity constitutes one of the central domains of functioning for most adults, job-related affect is a vital part of the human experience which should be taken into account when dealing with problems of people seek- ing professional help. This may be essential in career guidance and counseling where there is also an increas- ing focus on the emotional functioning of clients (Gati et al., 2011). Knowledge on what kind of emotions a client experiences toward work may help to prepare interventions targeted at specific problems. This can be done through the use of different affect-related interventions (Koydemir & Sun-Selisik, 2016; Saedpanah, Salehi, & Moghaddam, 2016). Warr’s (1990) measure can be applied as a useful tool for evaluating affective experiences related to work and distinguishing four affective dimensions differing in valence (i.e., positive and negative af- fect) and activation level rather than a single global affectivity dimension. Funding This research was financed by the National Science Centre, Poland, grant no. DEC-2013/10/M/HS6/00475. Competing Interests The authors have declared that no competing interests exist. Acknowledgments The data set described in Study 1 was never used in any of published article. The data set described in Study 2 was presented in Laguna, M., Razmus, W., and Żaliński, A. (2017). Dynamic relationships between personal resources and work engagement in entrepreneurs. Journal of Occupational and Organizational Psychology, 90(2), 248–269, however only one job-related affective well-being scale (enthusiasm) was used in that study. R ef er enc es Alessandri, G., Vecchione, M., Eisenberg, N., & Łaguna, M. (2015). On the factor structure of the Rosenberg (1965) General Self-Esteem Scale. Psychological Assessment, 27(2), 621-635. https://doi.org/10.1037/pas0000073 Arbuckle, J. (2005). Amos 6.0 user’s guide. Chicago, IL, USA: Amos Development Corporation. Baron, R. A. (2007). Entrepreneurship: A process perspective. In J. R. Baum, M. Frese, & R. A. Baron (Eds.), The psychology of entrepreneurship (pp. 19–40). London, United Kingdom: Lawrence Erlbaum Associates. Baron, R. A. (2008). The role of affect in the entrepreneurial process. Academy of Management Review, 33(2), 328-340. https://doi.org/10.5465/amr.2008.31193166 Beal, D. J., Weiss, H. M., Barros, E., & MacDermid, S. M. (2005). An episodic process model of affective influences on performance. The Journal of Applied Psychology, 90(6), 1054-1068. https://doi.org/10.1037/0021-9010.90.6.1054 Bledow, R., Schmitt, A., Frese, M., & Kühnel, J. (2011). The affective shift model of work engagement. The Journal of Applied Psychology, 96(6), 1246-1257. https://doi.org/10.1037/a0024532 Brief, A. P., & Weiss, H. M. (2002). Organizational behavior: Affect in the workplace. Annual Review of Psychology, 53(1), 279-307. https://doi.org/10.1146/annurev.psych.53.100901.135156 Brown, T. A. (2006). Confirmatory factor analysis for applied research. New York, NY, USA: Guilford Press. Bifactor Model of Job-Related Affective Well-Being 352 Europe's Journal of Psychology 2019, Vol. 15(2), 342–357 https://doi.org/10.5964/ejop.v15i2.1632 https://doi.org/10.1037%2Fpas0000073 https://doi.org/10.5465%2Famr.2008.31193166 https://doi.org/10.1037%2F0021-9010.90.6.1054 https://doi.org/10.1037%2Fa0024532 https://doi.org/10.1146%2Fannurev.psych.53.100901.135156 https://www.psychopen.eu/ Chen, F. F., Hayes, A., Carver, C. S., Laurenceau, J.-P., & Zhang, Z. (2012). Modeling general and specific variance in multifaceted constructs: A comparison of the bifactor model to other approaches. Journal of Personality, 80(1), 219-251. https://doi.org/10.1111/j.1467-6494.2011.00739.x Chen, F. F., Jing, Y., Hayes, A., & Lee, J. M. (2013). Two concepts or two approaches? A bifactor analysis of psychological and subjective well-being. Journal of Happiness Studies, 14(3), 1033-1068. https://doi.org/10.1007/s10902-012-9367-x Chen, Z., Watson, P. J., Biderman, M., & Ghorbani, N. (2016). Investigating the properties of the general factor (M) in bifactor models applied to Big Five or HEXACO data in terms of method or meaning. Imagination, Cognition and Personality, 35(3), 216-243. https://doi.org/10.1177/0276236615590587 Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modeling, 9(2), 233-255. https://doi.org/10.1207/S15328007SEM0902_5 Cifre, E., & Salanova, M. (2002). Multidimensionality and bipolarity of a Spanish version of Warr’s (1990) well-being measures. The Journal of Psychology, 136(1), 69-74. https://doi.org/10.1080/00223980209604139 Cuyper, N. D., Philippaers, K., Vanhercke, D., & Witte, H. D. (2017). The reciprocal relationship between resources and psychological distress among unemployed job seekers. Journal of Career Development. Advance online publication. https://doi.org/10.1177/0894845317730413 Daniels, K., Brough, P., Guppy, A., Peters-Bean, K. M., & Weatherstone, L. (1997). A note on a modification to Warr’s measures of affective well-being at work. Journal of Occupational and Organizational Psychology, 70(2), 129-138. https://doi.org/10.1111/j.2044-8325.1997.tb00638.x Fredrickson, B. L. (2001). The role of positive emotions in positive psychology: The broaden-and-build theory of positive emotions. The American Psychologist, 56(3), 218-226. https://doi.org/10.1037/0003-066X.56.3.218 Gati, I., Gadassi, R., Saka, N., Hadadi, Y., Ansenberg, N., Friedmann, R., & Asulin-Peretz, L. (2011). Emotional and personality-related aspects of career decision-making difficulties: Facets of career indecisiveness. Journal of Career Assessment, 19(1), 3-20. https://doi.org/10.1177/1069072710382525 Gonçalves, S. P., & Neves, J. (2011). Factorial validation of Warr’s (1990) well-being measure: A sample study on police officers. Psychology, 2(7), 706-712. https://doi.org/10.4236/psych.2011.27108 Gorgievski, M., & Stephan, U. (2016). Advancing the psychology of entrepreneurship: A review of the psychological literature and an introduction. Applied Psychology, 65(3), 437-468. https://doi.org/10.1111/apps.12073 GUS. (2017). Działalność innowacyjna przedsiębiorstw w latach 2014–2016. Warszawa, Polska: Urząd Statystyczny w Szczecinie. Hooper, D., Coughlan, J., & Mullen, M. R. (2008). Structural Equation Modeling: Guidelines for determining model fit. Electronic Journal of Business Research Methods, 6(1), 53-60. Hutchison, A., & Gerstein, L. (2017). Emotion recognition, emotion expression, and cultural display rules: Implications for counseling. Journal of Asia Pacific Counseling, 7(1), 19-35. https://doi.org/10.18401/2017.7.1.3 Laguna, Mielniczuk, & Razmus 353 Europe's Journal of Psychology 2019, Vol. 15(2), 342–357 https://doi.org/10.5964/ejop.v15i2.1632 https://doi.org/10.1111%2Fj.1467-6494.2011.00739.x https://doi.org/10.1007%2Fs10902-012-9367-x https://doi.org/10.1177%2F0276236615590587 https://doi.org/10.1207%2FS15328007SEM0902_5 https://doi.org/10.1080%2F00223980209604139 https://doi.org/10.1177%2F0894845317730413 https://doi.org/10.1111%2Fj.2044-8325.1997.tb00638.x https://doi.org/10.1037%2F0003-066X.56.3.218 https://doi.org/10.1177%2F1069072710382525 https://doi.org/10.4236%2Fpsych.2011.27108 https://doi.org/10.1111%2Fapps.12073 https://doi.org/10.18401%2F2017.7.1.3 https://www.psychopen.eu/ Isen, A. M., & Reeve, J. (2005). The influence of positive affect on intrinsic and extrinsic motivation: Facilitating enjoyment of play, responsible work behavior, and self-control. Motivation and Emotion, 29(4), 295-323. https://doi.org/10.1007/s11031-006-9019-8 Kline, R. B. (2005). Principles and practice of structural equation modeling. New York, NY, USA: Guilford Press. Konstam, V., Celen-Demirtas, S., Tomek, S., & Sweeney, K. (2015). Career adaptability and subjective well-being in unemployed emerging adults: A promising and cautionary tale. Journal of Career Development, 42(6), 463-477. https://doi.org/10.1177/0894845315575151 Koydemir, S., & Sun-Selisik, Z. E. (2016). Well-being on campus: Testing the effectiveness of an online strengths-based intervention for first year college students. British Journal of Guidance & Counselling, 44(4), 434-446. https://doi.org/10.1080/03069885.2015.1110562 Laguna, M., Alessandri, G., & Caprara, G. V. (2016). Personal goal realization in entrepreneurs: A multilevel analysis of the role of affect and positive orientation. Applied Psychology, 65(3), 587-604. https://doi.org/10.1111/apps.12061 Laguna, M., Mielniczuk, E., Razmus, W., Moriano, J. A., & Gorgievski, M. (2017). Cross-culture and gender invariance of the Warr (1990) job-related well-being measure. Journal of Occupational and Organizational Psychology, 90(1), 117-125. https://doi.org/10.1111/joop.12166 Laguna, M., Razmus, W., & Żaliński, A. (2017). Dynamic relationships between personal resources and work engagement in entrepreneurs. Journal of Occupational and Organizational Psychology, 90(2), 248-269. https://doi.org/10.1111/joop.12170 Leue, A., & Beauducel, A. (2011). The PANAS structure revisited: On the validity of a bifactor model in community and forensic samples. Psychological Assessment, 23(1), 215-225. https://doi.org/10.1037/a0021400 Mäkikangas, A., Feldt, T., & Kinnunen, U. (2007). Warr’s scale of job-related affective well-being: A longitudinal examination of its structure and relationships with work characteristics. Work and Stress, 21(3), 197-219. https://doi.org/10.1080/02678370701662151 Mäkikangas, A., Hyvönen, K., Leskinen, E., Kinnunen, U., & Feldt, T. (2011). A person‐centred approach to investigate the development trajectories of job‐related affective well‐being: A 10‐year follow‐up study. Journal of Occupational and Organizational Psychology, 84(2), 327-346. https://doi.org/10.1111/j.2044-8325.2011.02025.x Mielniczuk, E., & Laguna, M. (2018). The factorial structure of job-related affective well-being: Polish adaptation of the Warr’s measure. International Journal of Occupational Medicine and Environmental Health, 31(4), 429-443. https://doi.org/10.13075/ijomeh.1896.01178 Parker, S. K. (2014). Beyond motivation: Job and work design for development, health, ambidexterity, and more. Annual Review of Psychology, 65(1), 661-691. https://doi.org/10.1146/annurev-psych-010213-115208 Posner, J., Russell, J. A., & Peterson, B. S. (2005). The circumplex model of affect: An integrative approach to affective neuroscience, cognitive development, and psychopathology. Development and Psychopathology, 17(3), 715-734. https://doi.org/10.1017/S0954579405050340 Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and Social Psychology, 39(6), 1161-1178. https://doi.org/10.1037/h0077714 Bifactor Model of Job-Related Affective Well-Being 354 Europe's Journal of Psychology 2019, Vol. 15(2), 342–357 https://doi.org/10.5964/ejop.v15i2.1632 https://doi.org/10.1007%2Fs11031-006-9019-8 https://doi.org/10.1177%2F0894845315575151 https://doi.org/10.1080%2F03069885.2015.1110562 https://doi.org/10.1111%2Fapps.12061 https://doi.org/10.1111%2Fjoop.12166 https://doi.org/10.1111%2Fjoop.12170 https://doi.org/10.1037%2Fa0021400 https://doi.org/10.1080%2F02678370701662151 https://doi.org/10.1111%2Fj.2044-8325.2011.02025.x https://doi.org/10.13075%2Fijomeh.1896.01178 https://doi.org/10.1146%2Fannurev-psych-010213-115208 https://doi.org/10.1017%2FS0954579405050340 https://doi.org/10.1037%2Fh0077714 https://www.psychopen.eu/ Saedpanah, D., Salehi, S., & Moghaddam, L. F. (2016). The effect of emotion regulation training on occupational stress of critical care nurses. Journal of Clinical and Diagnostic Research: JCDR, 10(12), VC01-VC04. https://doi.org/10.7860/JCDR/2016/23693.9042 Salanova, M., Llorens, S., & Schaufeli, W. B. (2011). “Yes, I can, I feel good, and I just do it!” On gain cycles and spirals of efficacy beliefs, affect, and engagement. Applied Psychology, 60(2), 255-285. https://doi.org/10.1111/j.1464-0597.2010.00435.x Seo, M.-G., Barrett, L. F., & Bartunek, J. M. (2004). The role of affective experience in work motivation. Academy of Management Review, 29(3), 423-439. https://doi.org/10.5465/amr.2004.13670972 Sevastos, P., Smith, L., & Cordery, J. L. (1992). Evidence on the reliability and construct validity of Warr’s (1990) well-being and mental health measures. Journal of Occupational and Organizational Psychology, 65(1), 33-49. https://doi.org/10.1111/j.2044-8325.1992.tb00482.x Stephan, U. (2018). Entrepreneurs’ mental health and well-being: A review and research agenda. The Academy of Management Perspectives, 32(3), 290-322. https://doi.org/10.5465/amp.2017.0001 Tarnawa, A., & Zadura-Lichota, P. (Eds.). (2013). Raport o stanie sektora małych i średnich przedsiębiorstw w Polsce w latach 2011–2012. Warsaw, Poland: Polska Agencja Rozwoju Przedsiębiorczości. Van Horn, J. E., Taris, T. W., Schaufeli, W. B., & Schreurs, P. J. G. (2004). The structure of occupational well-being: A study among Dutch teachers. Journal of Occupational and Organizational Psychology, 77(3), 365-375. https://doi.org/10.1348/0963179041752718 Van Katwyk, P. T., Fox, S., Spector, P. E., & Kelloway, E. K. (2000). Using the Job-Related Affective Well-Being Scale (JAWS) to investigate affective responses to work stressors. Journal of Occupational Health Psychology, 5(2), 219-230. https://doi.org/10.1037/1076-8998.5.2.219 Walter, F., & Bruch, H. (2008). The positive group affect spiral: A dynamic model of the emergence of positive affective similarity in work groups. Journal of Organizational Behavior, 29(2), 239-261. https://doi.org/10.1002/job.505 Warr, P. (1990). The measurement of well-being and other aspects of mental health. Journal of Occupational Psychology, 63(3), 193-210. https://doi.org/10.1111/j.2044-8325.1990.tb00521.x Laguna, Mielniczuk, & Razmus 355 Europe's Journal of Psychology 2019, Vol. 15(2), 342–357 https://doi.org/10.5964/ejop.v15i2.1632 https://doi.org/10.7860%2FJCDR%2F2016%2F23693.9042 https://doi.org/10.1111%2Fj.1464-0597.2010.00435.x https://doi.org/10.5465%2Famr.2004.13670972 https://doi.org/10.1111%2Fj.2044-8325.1992.tb00482.x https://doi.org/10.5465%2Famp.2017.0001 https://doi.org/10.1348%2F0963179041752718 https://doi.org/10.1037%2F1076-8998.5.2.219 https://doi.org/10.1002%2Fjob.505 https://doi.org/10.1111%2Fj.2044-8325.1990.tb00521.x https://www.psychopen.eu/ App endi x Figure A.1. Six alternative measurement models for the Warr’s (1990) measure. Bifactor Model of Job-Related Affective Well-Being 356 Europe's Journal of Psychology 2019, Vol. 15(2), 342–357 https://doi.org/10.5964/ejop.v15i2.1632 https://www.psychopen.eu/ A bout the A uthor s Mariola Laguna, PhD, is associate professor at the Institute of Psychology, The John Paul II Catholic University of Lublin, Poland. Head of the Department of General Psychology, Editor-in-chief of “Roczniki Psychologiczne/Annals of Psychology”. Her research interests include social and personality psychology, organizational behavior, psychology of entrepreneurship. Emilia Mielniczuk, PhD, is assistant at the Institute of Psychology, The John Paul II Catholic University of Lublin, Poland. Her research interests include work and organizational psychology, psychology of entrepreneurship, health and positive psychology. Wiktor Razmus, PhD, is assistant professor at the Institute of Psychology, The John Paul II Catholic University of Lublin, Poland. His research interests include marketing and economic psychology, consumer behaviour, work and organizational psychology. Laguna, Mielniczuk, & Razmus 357 Europe's Journal of Psychology 2019, Vol. 15(2), 342–357 https://doi.org/10.5964/ejop.v15i2.1632 PsychOpen GOLD is a publishing service by Leibniz Institute for Psychology Information (ZPID), Trier, Germany. www.leibniz-psychology.org https://www.leibniz-psychology.org/ https://www.psychopen.eu/ Bifactor Model of Job-Related Affective Well-Being (Introduction) Job-Related Affective Well-Being Measure The Factorial Structure of the Job-Related Affective Well-Being Measure Study 1 Method Results and Discussion Study 2 Method Results and Discussion Discussion Strengths and Limitations Conclusions (Additional Information) Funding Competing Interests Acknowledgments References Appendix About the Authors