Job crafting in Serbia: Serbian mixed-method validation of the Job Crafting Scale 1

The aim of the current study was to validate the Serbian adaptation of the Job Crafting Scale (JCS), applying both qualitative and quantitative approaches, within three studies totalling 832 employees from different industries. Job crafting is work behaviour aimed at modifying job tasks in order to enhance and maintain work motivation. In Study 1, we have tested the understanding of all items and their possible cultural embeddedness. In Study 2, we have tested the JCS factor structure, factors’ reliability and validity. In Study 3, we have tested the invariance of the proposed models (Study 2 and Study 3 models) and criterion validity by analysing the correlation between the JCS and work engagement. Qualitative analysis has revealed that the majority of items transferred the intended meaning. Special care should be taken in interpreting the decreasing hindering job demands dimension scores, since these items might point toward behaviours that were not perceived as positive. With its overall reliability, second-order four-factor model invariance and criterion validity, originally composed 21-item JCS could be considered to be a valid instrument for assessing job crafting in Serbia. The present study has also shown that the 12-item JCS-Serbian short version has satisfactory psychometric properties and that it could be considered as a valid local job crafting scale.

Within the three separate studies we have analysed the understanding of the JCS items (qualitative research); we have verified the JCS factorial structure,reliability and validity of extracted factors; and, finally, we have tested the measurement model invariance and criterion validity. Tims et al. (2012) define job crafting as a set of actions autonomously undertaken by an employee to optimize the interplay between job demands and resources in order to achieve the established work goals in line with personal abilities, needs and motivation (Bakker & Demerouti, 2017;Tims & Bakker, 2010). Being focused on redesigning the features of the existing job, job crafting can be differentiated from other proactive behaviours which can also be found in the work context, such as taking initiative. These other proactive behaviours should be more noticeable in an array of life contexts other than work (Rudolph et al., 2017). In the JDR model (Bakker & Demerouti, 2017), there are two broad groups of job characteristicsjob demands and job resources that interplay and initiate motivation and strain. These are two different processes that consequently lead to job performance. Responding to job demands and enlarging job and personal resources could act as a buffer that may, in turn, reduce exhaustion. Within the JDR model, job crafting is a part of the 'empowerment' cycle, the bottom-up, person-driven tactic for job re-design that could enrich personal and job resources. In contrast to a job (re)design initiated by the employer, some researchers claimed that employees might, by themselves, proactively change their job tasks in order to make their job more meaningful (Bakker & Demerouti, 2017). Thus, job crafting may enlarge both personal and job resources that could further lead to work engagement, motivation and productivity. Employees' searching for challenges and developing resources could in turn change work environment in a stimulative way, and increase their job satisfaction in general.

Theoretical and Research Background
As Rudolph et al. (2017) pointed out, the concept of job crafting is not new. According to them, the concept of job crafting was announced by Kulk et al. (1987), more than thirty years ago. It was described as a participative strategy, the process during which employees actively change their job characteristics to improve person-job fit. Nevertheless, the research interest about job crafting increased significantly in the past decade (Rudolph et al., 2017), due to its clear positioning and operationalizing within JDR model (Bakker & Demerouti, 2017) and consequent Job Crafting Scale development (Tims et al., 2012).
The Job Crafting Scale (JCS) was developed through three separate studies with Dutch employees (Tims et al., 2012). These studies showed that there were four dimensions of job crafting (increasing social job resources, increasing structural job resources, increasing challenging job demands, and decreasing hindering job demands), covered by a total of 21 items that depicted various activities at work related to the optimization of job resources and diminishing job demands. All intercorrelations among the dimensions were significant and moderately positive, except in the case of the decreasing hindering job demands that showed no correlation with the remaining dimensions. It was shown that the JCS dimensions correlated with proactivity, personal initiative, cynicism, as well as with the peer-rated work engagement, employability, and job performance. This implied that there was discriminatory and convergent validity of the job crafting construct as assessed by this scale. The correlations between the dimensions of increasing social job resources, increasing structural job resources, and increasing challenging job demands, on the one side, and the peer-rated work engagement, employability and productivity on the other, ranged between .31 and .46. The decreasing hindering job demands dimension did not correlate with the peers' ratings. Still, it was shown that the decreasing hindering job demands dimension correlated positively with cynicism (r = [0.30; 0.35]).
The validation of the JCS in Spain (Bakker et al., 2018) confirmed the four-factor structure (increasing structural job resources, decreasing hindering job demands, increasing social job resources, and increasing challenging job demands). The four-factor Spanish version of the JCS also showed divergent validity with significant positive correlations of the increasing structural job resources, increasing social job resources, and increasing challenging job demands on the one side with work engagement and proactivity on the other. The JCS validation study in Japan (Eguchi et al., 2016) revealed five dimensions, with the decreasing hindering job demands splitting into two dimensions (decreasing emotional and cognitive job demands). The JCS was also validated in Italy (Cenciotti et al., 2016). The Italian version of the JCS consisted of 13 items covering three dimensions: increasing structural job resources, increasing social job resources and increasing challenging job demands. Confirmatory factor analysis demonstrated the three-factor structure. Job crafting, represented through these three dimensions, correlated positively with work self-efficacy, work engagement and job performance. The Polish version of the JCS also replicated the four-factor structure of the original JCS, with satisfactory reliability and validity (Rogala & Cieślak, 2019). Likewise, the Turkish validation of the JCS (Akın et al., 2014) revealed a satisfactory fit of the four-factor solution, as well as good reliability coefficients for each dimension.
In a meta-analytic study of job crafting, based dominantly on the JCS but including some other, similar job crafting scales, combining over 122 samples from different national and cultural settings, Rudolph et al. (2017) called into question the overall conceptualization of job crafting. Meta-analytic confirmatory factor analysis has shown that the model with a single job crafting factor, on which all four job crafting dimensions loaded, gave a satisfactory fit. Nevertheless, having in mind the low standardized factor loading of the decreasing hindering job demands dimension on the latent job crafting factor, the researchers warned of the relationships of this dimension with the overall factor. Analysing the relative contributions of all job crafting dimensions to the prediction of the positive and negative work and organizational behaviours, Rudolph et al. (2017) revealed that, contrary to other three dimensions, the decreasing hindering job demands dimension was a better predictor of the negative work outcomes such as job strain or turnover intentions, whereas the other three dimensions were better predictors of positive work outcomes such as work engagement and job satisfaction.
Job crafting is applicable to assessing the adaptive behaviours of both white-and bluecollar workers. Nevertheless, some adjustments of the JCS scale are needed when it comes to the population of the blue-collar workers. Nielsen and Abildgaard (2012) adapted and adjusted JCS (Tims et al., 2012) for assessing the blue-collar workers' job crafting. They obtained a different factor structure with five dimensions: increasing challenging demands, decreasing social demands (employee endeavours to avoid the emotionally challenging work situations), increasing social job resources, increasing quantitative demands (the employees' attempts to find more tasks for themselves), and decreasing hindering demands. As Nielsen and Abildgaard (2012) pointed out, social interaction for blue-collar workers could be both a resource and a demand. In addition, the increasing structural job resources dimension was not found in the blue-collar workers' sample since it was likely that they had the limited prospects of redesigning their jobs in order to reach more autonomy and develop themselves professionally.
As Wrzesniewski and Dutton (2001) pointed out, job crafting implies the employees' freedom to "craft" what they do in their job and how they do it, be it minor alterations or more radical changes. So, it could be presumed that job crafting demands certain organizational enabling conditions, stemming not only from the industry, but from economy and culture as well. There is research evidence that work engagement varies across Europe depending on economic activity and productivity (Schaufeli, 2018). Following the JDR model and some findings based on it (e.g., Lu et al., 2014), the engaged employees craft their jobs in order to create better person-job fitting. Therefore, it could be proposed that apart from the organizational conditions, in order to craft, employees should have some favourable economic conditions as well. Serbian economy is characterized by a high unemployment rate (13.6% of the labour force, aged 15-74, compared to 7.6% for the EU-28 countries in 2017, Eurostat, 2019) and high involuntary temporary employment rate (21.2%, compared with 7.4% for the 28 EU countries average in 2018, Eurostat, 2019). So, it is interesting to explore job crafting in the specific, turbulent context of Serbia. Moreover, Serbia is a country with several decades of workers' participation approach (during socialist era) and thus overall conditions supportive of job crafting (Petrović et al., 2017;Kulk et al., 1987). In relation to this, the unique contribution of this validation study was not only in gathering data that support the etic status of job crafting (in addition to validation studies from a still limited number and variety of countries), but enriching our knowledge about its emic embeddedness.
Finally, following Hofstede's (2001) cultural dimensions framework, it may also be argued that job crafting as a "work strategy" could be perceived differently in different cultures (Gordon et al., 2015), and, hence, used in different ways. Comparing job crafting of health care professionals in the USA and the Netherlands, Gordon et al. (2015) revealed that they crafted jobs in response to job and organisational demands that were an expression of Hofstede's cultural dimension of masculinity (higher in the USA) vs. femininity (higher in the Netherlands). In the context of Hofstede's cultural dimensions (2001), Serbia is notable for being a high power distance society (a society with a larger degree of accepting unequal distribution of power) and collectivistic society In that sense, it could be of significance to test robustness of the job crafting in the specific cultural, economic and national context of Serbia, particularly by using the same instruments, and the same data gathering procedure (Llorens et al., 2006). On the other hand, by offering internationally validated scale for assessing job crafting, we also opted for inspiring and encouraging researchers and practitioners from Serbia to undertake more comparable job crafting research. This kind of research could offer an insight that would be comparable to those from other European countries, at the same time laying grounds for human resources management interventions that would exactly be tailored to the Serbian organizational context.

The Overall Aim
The aim of the research was to validate the Serbian adaptation of the Job Crafting Scale (Tims et al., 2012), applying a mixed-method approach within three studies. In particular, the goals were: 1) to analyse in detail the understanding of the JCS items; 2) to explore and verify the factorial structure of the JCS; 3) to test the reliability and validity of the proposed factors; 4) to test the measurement model invariance, and 5) to investigate the construct validity through the relationship with work engagement. The overall goal was to enrich the understanding of job crafting by exploring this concept in the specific Serbian cultural and economic context.

The JCS Scale Adaptation to Serbian and the Overall Procedure
The first step in our research was to translate JCS (Tims et al., 2012) from English and linguistically adapt it for the use in Serbian language. The JCS was translated through the committee technique in three iterations (Brislin et al., 1973). After each step in translation, we carried out back translation into English. Each item is followed by the 5-point frequency scale from 1 = never to 5 = very often.
In this paper, we report the findings based on three separate studies. We used the same questionnaire for studies two and three that were carried out on independent, convenience samples. The survey data were gathered in the paper and pencil form, with the presence of a trained field researcher. Participation was voluntary and not compensated. All three studies were conducted in accordance with the Code of Ethics of the Serbian Psychological Society (2000), with the informed consent from all participants.

Study 1
The aim of Study 1 was to analyse the employees' understanding of the JCS items (Tims et al., 2012) through exploring the following topics: work activities associated with each item, checking the conceptual understanding of each item, identifying the possible negative implications, noticing the possible limitations of item applicability in different work and job contexts and detecting the possible mutual similarity of items, as well as their possible cultural embeddedness. With the qualitative approach we have opted for searching the means by which employees interpret their job crafting experience and for giving employees 'voice' during the process of scale testing (Gioia et al., 2013). This approach would allow us not only to validate presumable meaning, but rather to discover the new ones (Gioia et al., 2013).

Procedure.
A semi-structured qualitative interview was used to collect the data. The interviews were conducted by work and organizational psychology master students from the Department of Psychology, Faculty of Philosophy, University of Belgrade. As part of their coursework, students learned about the Job Demands-Resources Model (Bakker & Demerouti, 2007) in general, and the job crafting concept in particular (Tims et al., 2012). Additionally, they were trained for conducting this interview. Students recruited study participants during their obligatory professional fellowship within 12 organisations. They adopted targeted sampling aiming at those employees who worked for more than six months within the organisation, regardless of their age, profession, education and position. They were approaching respondents directly and also used the snowball technique in order to reach potential interviewees. The study was generally well accepted among employees (with 100% response rate) and all interviews were included in the analysis.
The interviews lasted between 30 and 60 minutes. In accordance with the stated aims, the data were gathered based on the questions covering the work activities associated with items, the meaning of each item, the resemblance with other items and whether each item indicated a positive or a negative work behaviour and possible mutual similarities of items.
Firstly, respondents were asked to fill in a paper and pencil form of JCS in order to familiarize themselves with items. The quantitative data gathered in this study were not included in further quantitative analyses. Then, a researcher proceeded with an interview aiming to discuss the meaning of JCS items. After reading each item interviewer asked the following questions: "Which work activity do you associate with this item?"; "What is the meaning of this item?"; "Do you perceive the activity described in the item as a positive or a negative one?; Does this item resemble any of the other items from this questionnaire?".
The data were analysed through qualitative content analysis (Flick, 2014) as one of the extensively used qualitative techniques for analyzing and reducing diverse textual materials (Bauer, 2000). In analyzing the materials from the interviews, we searched for the similarities and recurring themes in order to aggregate second order themes. Deducing from the available data, we made a general conclusion about the possible limitations of the items' applicability and their possible cultural embeddedness.
Sample. The sample comprised 63 employees from different sectors and industries (41 were women; 44 with university education; 50% of the sample were in the age group 21-40, and age of remaining 50% of participants was in the range 41-65 years. Participation in the study was voluntary, non-compensated and confidential, all participants were informed about the possibility to withdraw from interviews at any point.

Results and Discussion
The answers were first sorted within the previously defined topics: work activities associated with items, item meaning, resemblance with other items, whether each item indicated a positive or negative work behaviour, the possible limitations of items' applicability in different work and job contexts, possible mutual similarity of items (Table 1a-d, arranged by original subscales), as well as specific cultural embeddedness ( Table 2). All the topics mentioned in respondents' answers were covered, regardless of their frequencies. Table 2 lists the topics related to the items that may be culturally specific.
[ The items from the first dimension, increasing structural job resources, were mainly understood as related to personal development. With the exception of item 5 (Table 1a), all other items were understood as intended. Besides the intended meaning, item 5 ("I decide on my own how I do things") could also be understood as an expression of autonomy at work. If understood as an indicator of autonomy in decision making, this behaviour could point to arrogance, disdainful behaviour, and disregarding colleagues at work. From the collectivistic cultural context perspective, this could be perceived as a negative indicator of work behaviour (Table 2).
Concerning the dimension decreasing hindering job demands, we could say that we identified the problem with negative and/or ambiguous implications of the meaning of items.
The items depict the employees' reaction to irregular, stressful working conditions and work overload. The employees from Serbia perceived the decreasing hindering job demands in line with a local saying: "if employees are paid little, they can always work less." decreasing mental, social, emotional demands could be perceived as "doing just enough" and could be a negative indicator of work engagement.
Concerning the dimension increasing social job resources (Table 3c), the items were tied with the meaning of exchanging ideas and experiences with colleagues. Furthermore, we could observe that, although they conveyed the intended meaning, all items, except item 14, were perceived as almost equal in meaning and highly mutually alike. In Serbia, as a "power distance culture" (Hofstede, 2001), turning to a supervisor is highly acceptable, but at the same time, the behaviours covered by these items could be perceived as a sign of weakness, and, moreover, as a sign of flattering ( Table 2).
The items included in the increasing challenging job demands dimension (Table 1d) were understood as indicators of motivation and proactivity. Nevertheless, they were perceived as related to the organization, work and organizational commitment rather than as an expression of a personal career benefit. The expressed expectations that workers should be paid for every bit of work they do can be perceived as one of the "side-effects" of the transition from socialism to market economy. Item 21 was the most complex and abstract for understanding, regardless of the educational level and type of job (Table 1d). As expressed by some workers, "examining the underlying relationships between aspects of my job" can be perceived as a "philosophical issue" (Table 2).
In conclusion, the majority of items conveyed the intended conceptual meaning.
Qualitative analysis pointed to several issues with the items from the decreasing hindering job demands and increasing social job resources dimension. The items of these dimensions were perceived as pointing to some behaviours that could be perceived as negative and having applicability limitations in relation to jobs and work contexts; further, employees perceived them as mutually similar. They were also significantly embedded in the cultural context. In other words, the specificities of our cultural context caused the interpretations that were not in line with the job crafting concept. Items 5, 14, and 21 were challenging regarding their meaning and ambivalent implications as it was previously explained.

Study 2
The aim of Study 2 was to explore and verify the factorial structure of the JCS and to test the reliability and validity of the proposed factors. We have tested both the first-order fourfactor JCS model and the second-order JCS factor model that both presumed the existence of factors envisaged by Tims et al., (2012): increasing structural job resources, decreasing hindering job demands, increasing social job resources, and increasing challenging job demands. In further analyses, we have tested the reliability, convergent and discriminant validity of JCS dimensions.

Method
Procedure and Sample. The convenience sample consisted of 382 employees from 37 different organizations across Serbia (Belgrade, the capital city, and other towns). Research participation was anonymous, voluntary, and non-compensated. Employees filled out paperand-pencil questionnaires at work, during the break, under the supervision of trained fieldresearchers. Field-researchers were graduate psychology students on their professional internship. They approached employees, asked for participation and arranged surveying at the time most convenient for the respondent. The response rate was very high, with rejection rate 2-5% per field-researcher. Data were gathered in the period February-July 2017.
Participants' average age was 40 years; 33.94% were men; 57.9% had university education, 10.9% completed vocational school, 30.2% high school, 0.5% elementary school and 0.5% did not specify the education. Almost three quarters, i.e. 84.8% were employees, and 15.2% were in some managerial position (i.e., managers, supervisors, or team leaders). The majority of respondents, 57.6%, came from the state-owned companies. Most of them worked in companies that had between 51 and 250 employees (46.2%). Participants were employed in different sectorsmainly the education sector (36.1%), followed by the health sector (22.8%), and the IT sector (7.9%). About one third of respondents worked for some other industries, each represented with less than 5% of the participants. Data were analysed using IBM SPSS 21 and IBM SPSS Amos.

Results and Discussion
Factorial Structure. Firstly, data were analysed in order to estimate their suitability for exploratory factor analysis. The value of the Kaiser-Meyer-Olkin (KMO) measure was .817 for the complete model. The KMO values for the items were greater than .70, which is above the satisfactory limit of 0.50 (Field, 2013). Bartlett's sphericity test showed statistical significance (χ 2 [210] = 3342.53; p < .001). Applying Guttmann-Kaiser criterion (Eigenvalues > 1), five components were retained that explained 62.78% of the total variance. The first component explained 24.90%, the second component14.42%, the third component 11.45%, the fourth component 6.95%, and the fifth component 5.10% of variance. However, Cattell's scree test and Horn's parallel analysis (eigen values higher than the threshold value taken out from the equally large matrix of random numbers -21 variables * 382) suggested a four principal component solution. As the four-factor solution gave the expected number and structure of the components, we retained it for further analysis. For further component analysis we used maximum likelihood, with promax rotation (Table 3).
[ Table 3 around here] As can be seen from Table 3, all the dimensions had good reliability with Cronbach's alphas all above .75. Cronbach's alpha for the whole scale was .83. The number and structure of the components almost perfectly correspond to the initial version of the JCS. The only exception was item 5,"I decide on my own how I do things", that loaded on the fourth dimension (increasing challenging job demands) instead of the first one (increasing structural job resources) ( Table 3). "Alpha if item deleted" analysis for all items from subscale increasing structural job resources (including item 5), indicates that the reliability for this dimension increases to .84 if this particular item is deleted. Still, if the same item is included in calculating the Cronbach's alpha for the fourth dimension (where this particular item eventually belongs in this analysis), the reliability somewhat decreases (to .77). The item 5 was also "singled out" in qualitative analysis and perceived differently (Table 1a), more in line with the items that were mainly understood as proactivity. The correlations between the factors (Table 5)

Testing the Factorial structure -Confirmatory Factor Analysis
The JCS factorial structure was tested with the Confirmatory Factor Analysis (CFA) in the AMOS software. We tested the JCS within the first (Table 4, Models 1 to 3) and secondorder (Table 4, Models 4 to 6) confirmatory factor analysis because they were both supported in the previous research (e.g., Rudolph et al., 2017;Tims, et al., 2012). The first-order fourfactor model had a generally poor fit (Table 4), especially judging by the values of the RMSEA and PClose (Hu & Bentler, 1999). Item 5 had a particularly low factor loading (.25), followed by items 6 (.46) and 7 (.47). In addition, items 20, 21, and item 4 could be taken into the consideration for being omitted from the scale, concerning the fact that they had factor loadings close to, but still below .60.
Searching for the way to improve the discrepancies between the proposed and the estimated model, we looked for the modification indices. Thus, we added covariance between the error terms between the items 6 and 7; 8 and 9;10 and 11, and 17 and 18. All these covariances between the errors were included only within the same factor. The respondents from the first study perceived these items as mutually similar. It should be noted that the majority of these items belong to the decreasing hindering job demands dimension, except for the items 17 and 18 that belong to the increasing challenging job demands dimension. As can be seen from Table 4, these error covariances gave a more acceptable, but not a satisfactory fit.
In addition, it should be noted that including error covariances could be problematic since this calls into question the construct validity (Hair et al., 2010). These covariances actually point towards the problematic relationship between some items, as well as their content per se. So, we tested the third modelthe first-order, four factor model (the 12-item scale)in which we omitted the most problematic items. We decided to omit the items based on both qualitative and quantitative analysis (based on the low factor loadings and modification indices). Hence, we excluded items 4,5,6,7,8,14,16,17,21, and came up with the 12-item JCS-Serbian short version scale. As can be seen from Table 4, this solution produced a good fit (Hu & Bentler, 1999). Apart from that, Cronbach's alpha for this short version was .74, which indicates solid reliability.
Finally, we tested the second-order job crafting model that represents the relationship of job crafting as an underlying factor with the first-order factorsincreasing social support, increasing structural job resources, increasing challenging job demands, and decreasing hindering job demands. As in the previous analysis, we tested the model with and without error covariances, as well as the12-item Serbian short version. The results are given in Table 4. As can be seen, the Serbian short version (Model 6) gave the best fit.

Reliability and Validity of Factors
In further analysis, we tested the reliability and convergent and discriminant validity of the JCS factors (Table 5) using the Master validity tool for AMOS (Gaskin & Lim, 2016). For reliability, we used the measures of composite reliability (CR), for convergent validity, the average variance extracted (AVE), and for discriminant validity we used the maximum shared variance (MSV) and its relation to AVE (MSV < AVE), as well as the relation of the square root of AVE to inter-construct correlations. As can be seen from Table 5, in general, the indicators of composite reliability, convergent and discriminant validity were more satisfactory (Hu & Bentler, 1999) for the 12-item JCS-Serbian short version than for the original version of the JCS-Serbian. The average variance extracted was under .5 for the dimensions decreasing hindering job demands and increasing challenging job demands for the short version of the JCS. For the long version, it was below .5 for all dimensions, except for the increasing social job resources dimension. This certainly points to some issues already observed in the qualitative analysis, but since the AVE is considered to be a quite stringent indicator of convergent validity (Malhotra & Dash, 2011), we can rely on CR that is generally reasonable in both cases.

Study 3
The Study 3 had two aims. The first one was to test the measurement model invariance across samples (from Study 2 and Study 3). The second one was to investigate the construct validity through the relationship with work engagement. We expected that measurement invariance of the hierarchical structure of JCS, would be confirmed (Rudolph et al., 2017;Zhang & Parker, 2019). Apart from that, we expected that the correlation between job crafting and work engagement would be significant and positive since work engagement could be considered both an outcome and an antecedent of job crafting (Rudolph et al., 2017). Field-researchers were graduate psychology students on their professional internship. They approached employees, asked for participation and arranged surveying at the time most convenient for the respondent. Thus, the response rate was very high, with rejection rate 2-5% per field-researcher. Data were gathered in the period February-July 2018.

Method
Participants' average age was 39 years; 62.70% were women. The majority of respondents (55.30%) graduated from university; 35% had a high school diploma, and 8.40% completed a vocational technical college. The vast majority (85.80%) were employees, while 14.20% were in some managerial position (i.e., managers, supervisors or team leaders). The majority of employees, 52.90%, worked in state-owned companies. The largest number of participants were employed at the companies that had between 51 and 250 employees (43.8%).
Employees worked in different economy sectors -almost one third, 31.30%, were employed in the education sector, and the remainder worked in the health sector (20.20%), IT (11.1%), the technology hardware sector (6.5%), and trade (6%). Other industries were represented with less than 5% of employees each.

Measures
Apart from the JCS scale (Tims et al., 2012), we assessed work engagement using the Serbian version of the Utrecht Work Engagement Scale, the UWES-Serbian (Petrović et al., 2017). It consists of 17 items followed by a seven-point scale (from 0 = Never, to 6 = Always).
It was shown that the Serbian version of the UWES-17 had a high reliability and that the factorial structure was in line with the envisaged model (vigour, dedication and absorption).

Results and Discussion
Testing the invariance of the second-order JCS factor model. We tested the invariance of the second-order JCS factor model (both the JCS-Serbian and the JCS-Serbian, short) following the procedure proposed by Chen et al. (2005), as well as Rudnev et al. (2018).
We analysed the second-order factor invariance with error covariance of the whole version of the JCS, as well as that of the short, 12-item JCS-Serbian version.
In order to identify the model (in both cases), we applied the marker item method. As marker items in both models we chose the following ones: "I try to develop myself professionally" (for increasing structural job resources); "I ask whether my supervisor is satisfied with my work" (for increasing social job resources); and "When there is not much to do at work, I see it as a chance to start new projects" (for increasing challenging job demands).
All these items are actually the first indicators of each construct, which is usually the default option when building and testing models in AMOS. The marker variable for the dimension decreasing hindering job demands was "I manage my work so that I try to minimize contact with people whose problems affect me emotionally". When testing the invariance of the short, Serbian version (in which this item was not listed), we used the item "I organize my work so as to minimize contact with people". Both of these items, in both models, are the first indicators of each construct. For the second-order factor, Job crafting, the dimension increasing challenging job demands was chosen as the marker variable, since it could be understood as the most representative dimension of job crafting (e.g., Rudolph et al., 2017). Error covariances imposed in the previous study that lead to a better model fit were also included in  Table 6.
[ Table 6 around here] As can be seen from Table 6, the fit measures for all five models for the long JCS form were acceptable (with RMSEA < .06 and CFI close to .95). Still, concerning these values, especially CFI, it could be concluded that the JCS-Serbian, short form, produced a good fit.
Nevertheless, the χ2 difference test for both JCS forms (the original, 21-item and the Serbian short form) did not show a significant difference or deterioration when comparing each step with the previous one. This indicates that both forms could be considered as invariant, but it should be noted that for testing the invariance of the original JCS form we included the error covariance.

Relationships of the JCS and UWES
Since we lacked the means to explore in depth the relationships of the JCS with other job crafting measures, we wanted to examine the relationships of the JCS with the conceptually related JD-R theory construct of work engagement, assessed by the UWES (Bakker & Demerouti, 2017). The correlations of the JCS-Serbian sub-scales and the total scores with the UWES-Serbian sub-scales and total scores are presented in Table 7 for the 21-item version and in Table 8 for the Serbian short 12-item version.
[ Regarding the level of correlations of the JCS long and short versions with the UWES (tables 7 and 8), we could note that the JCS long version gave significant correlations with the UWES total and all three UWES dimensions, except the decreasing hindering job demands dimension that is generally not correlated with the UWES. In line with the findings of the qualitative analysis, the decreasing hindering job demands dimension is understood as a remedial activity rather than as a correlate of work engagement. Somewhat higher correlations of increasing structural job resources and increasing challenging job demands with all the UWES dimensions and UWES total confirm that the employees from Serbia that are more prone to craft their jobs are at the same time more engaged in their jobs, whereas investment and crafting of the social aspects of the job are more general issues rather than strictly related to the job itself.

General Discussion
Wrzesniewski and Dutton (2001) pointed out that job crafting implies the employees' freedom to "craft" their jobs. Clearly, actual crafting behaviours are closely tied with the context-specific nature of the jobs in general and job crafting in particular, on the one side, and the conceptualization of crafting as a culturally specific "work strategy" (e.g., Gordon et al., 2015), on the other. Thus, to fully understand the concept of job crafting, it should be explored within a culturally sensitive approach in applying the JDR model. Having in mind a turbulent and sensitive social, economic and cultural context of Serbia as a specific context for job crafting, the aim of this research was to validate the Serbian adaptation of the Job Crafting Scale (JCS) applying the mixed-method, qualitative and quantitative approach within three studies.
The qualitative analysis of the JCS-Serbian adaptation revealed that the majority of items transferred the intended meanings. Nevertheless, it pointed to the problems with the understanding of some of the items from the decreasing hindering job demands and increasing social job resources dimensions.
The decreasing hindering job demands dimension is mainly understood as a protective strategy, oriented towards conserving the resources in "bad times" (highly demanding organizational circumstances), rather than as a proactive behaviour related to work engagement. The interpretation of hindering job demands as a protective strategy may set this dimension apart from the rest of the job crafting dimensions. However, this interpretation is in line with Tims et al. (2012) validation study results from Netherlands about significant moderate correlations of the scores on the decreasing hindering job demands subscale and the cynicism scores. This interpretation could also be supported by the results of the meta-analytic study of the concept of job crafting by Rudolph et al. (2017), that has shown that the decreasing hindering job demands dimension was a better predictor of negative work outcomes such as job strain or turnover intentions. That is why some researchers in their national JCS validation studies omitted the inclusion of decreasing hindering job demands dimension from the start (e.g., Cenciotti et al., 2016). Nevertheless, before putting it aside from the concept of job crafting, the decreasing hindering job demands dimension has to be further investigated within different occupational groups, industries and cultures. As it was evident from Nielsen and Abildgaard (2012) research, mentioned early in this article, there are some occupational groups (e.g., blue-collar workers) in which some crafting behaviours (e.g., increasing structural job resources) were not found. Decreasing hindering job demands dimension could prove to be a proactive form of job crafting behaviour within a smaller group or cluster of employees or jobs.
The respondents from Serbia understood the items of the increasing social job resources sub-scale as intended, but they also understood them as quite mutually similar. This means that they expressed lower sensitivity in understanding them. It should be noted that four out of five items from this subscale cover the relationship with supervisor, and only one item deals with increasing social resources through colleagues. In Serbia, as high "power distance culture" (Hofstede, 2001), self-initiated approaching to a supervisor could be perceived as a sign of flattering. Based on lower level of agreement with the items from the increasing social job resources scale, we could conclude that respondents could perceive these behaviours as flattering which is not an accepted form of behaviour in high power distance culture. From standpoint of a collectivistic culture, the concept of increasing social job resources should be widespread and widely accepted behaviour, but these behaviours should be connected with colleagues, friends and team-members, rather than supervisors.  (Bakker et al., 2020).

Limitations and Implications of this Research
As a final point, there are some limitations of this research that have to be pointed out.
Though we had multiple samples, totalling more than eight-hundred employees from different sectors and vocations, the sample was not composed as a representative one. Including the blue-collar workers who work in different economic, social and cultural circumstances could be a valuable addition to the existing research. In relation to that, what may be a limitation of our research is that we did not explore and analyse the specificity of crafting in Serbia. On the one hand, this was justified by the intention to reach the necessary understanding and a reasonable 'coverage' and grasp the fullness of this phenomenon while keeping the comparability of the proposed solution. On the other hand, it is reasonable to further explore the culturally, economically and socially embedded crafting stories. Starting from twelve JCS items selected in this study that proved to be understandable, reliable and invariant, we could further explore some 'missing' stories that would cover the specificities of the Serbian context.     Thinking upfront about all the steps needed to be taken for a repetitive tasks, well organizing work in advance (production jobs) Meaning: sounds like "doing just enough". Negative indicator of work behaviour: To be well performed, every job demands focused concentration that has to last for some period of time, this could lead to mistakes at work; "I think that work has to be done with more focus and higher concentration in order to be well done and with the purpose" (Engineer); Unethical work behaviour; Using shortcuts; Avoiding work; Rather avoiding responsibilities than being resourceful. Possible limitation: Depending on the job and expertise  Limitation: There are jobs in which employees are overloaded with work all the time and/or there is no spare time at work (production jobs); If you do not have too much to work you should work at a slower pace and preserve energy and recover; Certain level of autonomy is needed at work to be in a position to think about new projects 20. I regularly take on extra tasks even though I do not receive extra salary for them Being engaged in extra activities related to well-being of one's clients/users of services (service/education sector); Completing colleagues' work when they cannot do it themselves (service job) Meaning: Overtime work; Diligence; Work overload; Work commitment; Organizational commitment Negative indicator of work behaviour: Exploitation Limitation: Employees perceive this activity in relation to the organization's overall attitude towards them, whether they would want to support the organization's endeavours without being paid for extra work; Employees should be paid for all the work they do, employees should not often perform additional tasks without being paid for them; Employees do not have time for the work that is not generating income 21. I try to make my work more challenging by examining the underlying relationships between aspects of my job Motivating oneself by reminding about the broader meaning and/or aims of the job; Reflecting on the quality of work (service job); When upset about something at work, reminding oneself about the mission of the job (service/education job); Choosing work activities in line with personal values and/or mission of the job (office jobs) Meaning: Job importance; Mission of the job Problem with understanding the item: The item was too abstract for respondents, the most difficult to understand Could be perceived as an indicator of insecurity and weakness Increasing challenging job demands 17. When an interesting project comes along, I offer myself proactively as project co-worker It is better to preserve energy than to spread too much 19. When there is not much to do at work, I see it as a chance to start new projects It is better to preserve energy than to spread too much 20. I regularly take on extra tasks even though I do not receive extra salary for them Possible social, economic and cultural embeddedness, employees often experience economic hardship and do not have resources to do unpaid work, masochism, workaholism, Burning out and exhausting yourself; Collectivism, helping colleagues, and organization, volunteering at work 21. I try to make my work more challenging by examining the underlying relationships between aspects of my job Perceived as a "philosophical issue"  Note. *Covariations of error terms that are part of the same factor. Note. * p < .050; ** p < .010; *** p < .001; Square root of AVE is in bold, diagonally.