The “Working conditions and control questionnaire” (WOCCQ): Towards a structural model of subjective stress

Revue européenne de psychologie appliquée 58 (2008) 253–262 Original article The “Working conditions and control qu Towards a structural model of su Le questionnaire sur le contrôle de l’activ vers un modèle structural du stress subjectif Abstract This pape naire” (WO the control stress? The way. The co control of r effect of the perspective © 2008 Els Résumé L’objecti du « Questi langue fran subjectif ? e comprend 8 même mani des tâches s également d contrôle sont semblables en termes de R-carré. Ces résultats sont discutés non seulement dans la perspective théorique du stress au travail, mais aussi en termes d’interventions relatives à la gestion du stress. © 2008 Elsevier Masson SAS. All rights reserved. Keywords: Jo Mots clés : C � This resea E-mail ad 1162-9088/$ doi:10.1016/j b control; Subjective stress; Working conditions; Psychosocial risks ontrôle de la situation de travail ; Stress subjectif ; Conditions de travail ; Risques psychosociaux rch has been supported by the Belgian Science Policy. dress: [email protected]. 1. Introduction Job control is one of the most popular concepts in occupa- tional psychology literature. The complexity of this concept is largely recognized and discussed in the scientific literature (e.g. Aronsson, 1989; Frese, 1989). More particularly from the per- spective of stress studies, the feeling of uncontrollability on job – see front matter © 2008 Elsevier Masson SAS. All rights reserved. .erap.2008.09.008 I. Hansez Department of Work Psychology, University of Liège, boulevard du Rectorat, 5 (B32), 4000 Liège, Belgium r examined the structural model of subjective stress using the job control dimensions of the “Working conditions and control question- CCQ), a psychosocial risk diagnosis widely used in French-speaking countries. Two research questions were investigated: (1) Do all facets influence subjective stress in the same way? and (2) Are certain control scales more important than others in the prediction of sample used includes 816 workers of a public employment agency. First, not all of the facets of job control influence stress in the same ntrol of resources dimension is important in indirectly influencing the stress process. Planning control is a partial mediator between esources and other dimensions of control. The model suggests considering future control as an exogenous variable. Finally, the direct four job control subscales on stress is identical in terms of R-square. These results are discussed not only with regards to the theoretical of stress at work but also the stress intervention perspective. evier Masson SAS. All rights reserved. f de la recherche est de présenter un modèle structural du stress au travail permettant de prendre en compte les dimensions de contrôle onnaire sur le contrôle de l’activité de travail » (WOCCQ), un diagnostic des risques psychosociaux largement utilisé dans les pays de c¸aise. Deux questions guident cette recherche : (1) Les dimensions de contrôle influencent-elles toutes de la même manière le stress t (2) Certaines dimensions de contrôle sont-elles plus importantes que d’autres dans la prédiction du stress ? L’échantillon utilisé 16 travailleurs d’une agence publique pour l’emploi. Tout d’abord, toutes les dimensions de contrôle n’influencent pas le stress de la ère. Le contrôle sur les ressources est primordial même s’il n’influence le stress que de manière indirecte. Le contrôle sur la planification e présente comme un médiateur partiel entre le contrôle sur les ressources et les autres dimensions de contrôle. Le modèle suggère e considérer le contrôle de l’avenir comme une variable exogène. Enfin, les effets directs mis en évidence pour quatre dimensions de estionnaire” (WOCCQ): bjective stress� ité de travail (WOCCQ) : 254 I. Hansez / Revue européenne de psychologie appliquée 58 (2008) 253–262 factors is hypothesized to influence the generation of stress. High job control also has an impact on health and well-being, i.e. fewer somatic co Smith et al Several on the well specifically mechanism different m tested the m did not con comes. She role (endog variable. Jo allow indiv their desire this mecha ception of reducing th This underl makes it po ception of Consequen of the job s stress. In th insofar as o work envir The job the referen epidemiolo in an interv Even if this in the scie be demons authors cri measure of (1979) refl authors (e. et al., 1997 confused in utilization, However sk (Wall et al authors rec and job dem 1.1. From multifacete From a m important c 1979, p. 29 scales (e.g 1995). De Croo have used t Karasek (1 ing one-dimensional scales, we could observe an evolution in the way in which control scales are conceived. Admittedly, cer- thor y ar s of ify K ake r ay, dis tive tain hree diag ale o f Sm not auth l to unat ed sc on th onse ds to ted, sibil re d ct o c to . Di it dif d, on the a ssar ntrol r of is a p soci re n inter ly de d in espo ale th of gene proa prop ntio soci rom ns unti ls p ling. rwar mplaints and higher satisfaction (e.g. Spector, 1986; ., 1997). models exist to explain how job control has an impact -being of workers (Frese, 1989; Karasek, 1979). More , according to Frese (1989), it is possible, as far as s of control are concerned, to distinguish between oderating and direct effects. Carayon (1993) has oderating effect of control assumption. Her results firm a moderating effect of job control on stress out- argues that “job control does not play a mediating enous variable) but rather functions as an exogenous b control could be a structuring factor that would iduals to adjust job demands and other job elements to d level” (Carayon, 1993, p.474). Frese (1989) called nism “stressor reduction”. This means that the per- control has an indirect effect on stress reactions by e impact or intensity of the demands of the situation. ines the importance of a job control measurement that ssible to assess whether or not a worker has a per- control over all job factors included in his/her tasks. tly, control dimensions focusing on specific aspects hould be negatively correlated with the perception of is sense, perceived control is even more important nly perceived control leads the person to change the onment situation (Frese, 1989). demand–control model (Karasek, 1979), which is ce in the field of job control research, is useful for gical studies but insufficient for diagnostic purposes entionist perspective (de Jonge and Kompier, 1997). model is intuitively attractive and largely recognized ntific community, its empirical validity still has to trated. From a methodological point of view, many ticize the use of the decision latitude construct as a job control. The decision latitude items of Karasek ect decision authority and skill discretion. But many g. De Croon et al., 2000; Kristensen, 1995; Smith ) mention that the skill discretion items are quite that they are closer to job characteristics such as skill job complexity and job variety than to job control. ill discretion is not necessarily linked to job control ., 1995; Van der Doef and Maes, 1999). All these ommend more specific scales to measure job control ands. unidimensional scales of job control to d job control scales ethodological point of view, we have to deal with an riticism of studies dealing with job control (Karasek, 0): the measurement of job control through global . Jones and Fletcher, 1996; McKnight and Glass, n et al. (2000) observe that up until now many studies he general decision latitude construct, as defined by 979). During the 1980s, following criticisms regard- tain au but the aspect to ver still m Kellow The resenta who ob from t Moos tion sc items o It is certain paralle Unfort tifacet agree little c of fiel neglec respon measu prospe specifi structs makes notice allow is nece job co numbe scales psycho therefo of the ficient targete In r trol sc of view which our ap ology interve psycho 1.2. F questio Up trol too model put fo s have created scales that are still one-dimensional, e based on items which make reference to various the work situation. But most recent studies, aiming arasek’s model or the moderating role of control, eference to one-dimensional scales (e.g. Barling and 1996). parity of the scales is quite obvious. The most rep- example of the problem is probably Carayon (1993) s four different response formats for nine items taken different scales: the autonomy scale of the Insel and nosis (1974, cited by Carayon, 1993), the participa- f Caplan et al. (1975, cited by Carayon, 1993) and ith et al. (1981, cited by Carayon, 1993). until the end of the 1980s and thereafter, that we see ors propose scales with several control dimensions, the elaboration of different taxonomies (Table 1). ely, numerous criticisms can be levelled at these mul- ales. First of all, it is said that the authors seem to e content of the different facets but, unfortunately, nsus is to be found regarding the optimal number be considered. Certain aspects are also relatively in particular control over the physical environment, ities and the future. Moreover, different indicators ifferent aspects, and this means that there is little f obtaining sound databases. The scales are often the job analyzed and do not incorporate validity con- fferent indicators measure different aspects, which ficult to look for reliable data. As Jackson et al. (1993) ly standardised, generally applicable measures will ccumulation of comparative and normative data that y to make more systematic judgments about whether is at an optimal level. Finally, the relatively small items contained in the existing multidimensional roblem when making a differentiated diagnosis of the al risks in a working environment. The fields should ot only be pertinent for the purposes of adjustment vention after the diagnosis, but should also be suf- tailed to allow more specific control problems to be each field (Kristensen, 1995). nse to these criticisms, we have developed a con- at relates to different fields of work, from the point the stressor reduction mechanism (Frese, 1989), and rally applies to any working environment. Above all, ch aims to be practical in the sense that the method- osed should be useful in terms of the potential for n by any professional concerned by the problem of al risks in the workplace. a new multifaceted job control scale to research l now, none of the studies that have highlighted con- ertaining to different aspects of work have attempted The authors involved in this field of study have d multidimensional control scales (e.g. Breaugh, I. Hansez / Revue européenne de psychologie appliquée 58 (2008) 253–262 255 Table 1 Overview of job control multidimensional scales in the scientific literature. Author(s) N items Response format α Cronbach Breaugh y 3 Agreement 0.91 nomy 3 7 points 0.81 y 3 0.83 McLaney & H cal en Jackson et al. Jimmieson & De Jonge, Lan tion n Widerszal-Ba & Zolnierczy e l level icipat Sargent & Ter 1985; Jack Laney, 198 Zolnierczy the constru selves on t by Karasek Karasek m 1997), in t based on se Jonge et al Terry an approach to control ove of a high l mean a hig observes th linked to th point out th (e.g. work p more impo control. Sta more impo Baker et al ent levels a effects. Th al., 2000; H sioned cou particularly importance Going b job control Do all aspe same way? than others systematica ontr ine leve r? In ns. ral m ntrol ss in thod ater Mul l’s ( Year of publication Dimensions 1985 Work method autonom Work scheduling auto Work criteria autonom urrell 1988 Task control Decision control Control over the physi Resource control 1993 Timing control Method control Terry 1993 Task control Decision control deweerd & van Breukelen 1994 Control within a situa Control over a situatio zyl 1995 Basic control k Control over time fram Control related to skil Control related to part ry 1998 Task control Decision control Scheduling control son et al., 1993; Wall et al., 1995; Hurrell and Mc 9; Sargent and Terry, 1998; Widerszal-Bazyl and k, 1995) but their efforts have focused chiefly on ction of scales. In order to do so, they base them- he criticisms of the decision latitude scale proposed (1979). Their main objective remains linked to the odel (e.g. Elsass and Veiga, 1997; Mullarkey et al., hat they sometimes consider a global control score veral subscales that are brought to the fore (e.g. de ., 1996). d Jimmieson (1999) recommend a multidimensional the extent that workers are able to perceive personal r different facets of their work, and the perception evel of control over one facet does not necessarily h level of control over other facets. Frese (1989) also Is the c determ a high anothe questio structu the co progre 2. Me 2.1. M 2.1.1. Kas at areas of control can vary depending on how they are e workers’ daily activities. Sargent and Terry (1998) at the sources of control which are relevant to the task ace, organization of the task, control of planning) are rtant stress moderators than the peripheral sources of rting from the idea that certain types of control are rtant than others (e.g. Israel and Schurman, 1990), . (1996) measure several types of control at differ- nd show that various types of control have different erefore, according to several studies (e.g. Zijlstra et urrell and Lindstrom, 1992), the type of control envi- ld have different effects on subjective stress. More , direct sources of task control could be of primordial . eyond the unidimensional quality of the concept of thus allows us to ask some interesting questions. cts of job control influence subjective stress in the Are some aspects of job control more important in the prediction of stress? Do areas of control vary lly according to the type of profession envisioned? empirical r elaborate t (WOCCQ) elucidated Alpha coef 2001) inclu subscales: involved in the tasks, jo to oneself a time manag (11 items). of the situa without be about the w as I want”. 1 The comp www.woccq.b 7 Intensity 0.85 4 5 points 0.74 vironment 2 0.79 2 0.82 4 Importance 0.85/0.79 6 5 points 0.77/0.80 6 Importance 0.81 5 7 points 0.83 4 Opportunities 0.75 6 5 points 0.82 3 Not communicated 0.62 3 0.85 3 0.69 ion 3 0.77 6 Importance 0.84 3 5 points 0.76 5 0.80 ol of different areas of work accumulated in order to a general sense of control over the job activity? Can l in one area of control compensate for weakness in this paper, we will focus on the first two research The objective is to work out and test a theoretical odel that will make it possible to incorporate all subscales for explaining stress and, from there, to the understanding of the stress process. ial tifaceted job control scale 1989) recommendations and suggestions from other esearch (e.g. Jones and Fletcher, 1996) were used to he “working conditions and control questionnaire” . After a pilot test, some items were reformulated, and even eliminated according to the Cronbach’s ficients. The final version of the WOCCQ (Hansez, des 77 items. The items were grouped together in six control over resources needed to perform the tasks the job (eight items); control over tasks (definition of b role and procedures) (16 items); control over risks nd to others (16 items), planning control (12 items); ement control (14 items); and control over the future Each item makes reference to a person’s perception tion at his/her work, such as “I see my work piling up ing able to resolve latencies”, “I can say something ay work should be done”, “I can adapt my work pace 1 The items were not formulated directly in terms of lete questionnaire is available on request from the author or from e. 256 I. Hansez / Revue européenne de psychologie appliquée 58 (2008) 253–262 control but they could all be easily coded in terms of control. The WOCCQ response format was a four-point frequency Lik- ert scale as 2 = sometim my job, 4 = valence of control. Th means of th a Rasch an use of quan the Rasch p on an inter liminary an tests) in Pr ering the 7 for the item When cons coefficients ability coef Low subje resources a the interind whole, the and outfit transforme is compatib 2.1.2. Psyc The “Ps orated at th Tessier and mensional conception conception pathologica observed d symptoms results of t versions PS 1990). We am straine mouth feel I am short through Ra results yiel scale as res the items a On the who (Hansez, 2 2.2. Subjec The stud in daily co vey: the co trainers, w population workers, was tested. Four hundred and seventeen counselors returned the questionnaire, or a response rate of 79 %; 316 train- ticip uals n, th 5 me he s m. roce t, tra dlem ires ssure ation nd t esti ining s, re ary. ults relim t the axim riabl that ated. gativ egre l exp merl all jo atura corr .30). lann ubsc hav esult bsca rtial subs mila n an subs way ral m the s “a situa cessa follows: 1 = rarely or never applicable to my job, es applicable to my job, 3 = regularly applicable to almost always or always applicable to my job. The the items was balanced. High scores reflect high job e validity of the WOCCQ has been determined by e combined use of the item response theory through alysis, a study of the construct validity and the joint titative and qualitative data (Hansez, 2001). We used arameters which are continuous variables measured val scale. All variables were normalised after a pre- alysis of these parameters (skewness and kurtosis elis 2 (Jöreskog and Sörbom, 1999). When consid- 7 items altogether, the Rasch reliability coefficients s and for the subjects are 0.93 and 0.89 respectively. idering the job control subscales, the Rasch reliability for the items are in a margin of 0.90–0.95. The reli- ficients for the subjects vary between 0.56 and 0.79. ct reliability coefficients apply mainly to control of nd risk control but are probably justified in part by ividual variability about workers’ perception. On the means and standard deviations of the infit (weighted) (unweighted) fit statistics in their mean square and d (t) forms are acceptable which means that the data le with the model (Hansez, 2001). hological state of stress measure ychological state of stress measure” has been elab- e University of Laval (Canada), by the team of Dr. Dr. Lemyre. Their aim was to construct a unidi- measurement that would be representative of the of stress as an indicator of the adaptive tension, a which separated stressors (sources of stress) from l symptoms (illnesses). The state of stress is self- irectly by the person and is not derived from clinical as is done with some existing questionnaires. The his research is the PSSM (49 items) and two short SM-A and PSSM-B (25 items each) (Lemyre et al., used the short version A. Examples of items are: “I d or nervous”, “I have a lump in my throat or my s dry”, “I feel pressed for time, caught up with time, of time”. The item response theory was also used sch analysis to validate this instrument. The final ded a version of 20 items with a five-point-Likert ponse format. The Rasch reliability coefficients for nd for the subjects are 0.92 and 0.85 respectively. le, the 20 items had high indices of fit to the model 001). ts y took place in a public employment agency. Agents ntact with users of the services took part in the sur- unselors, who offer support and job orientation, and ho teach job skills to the unemployed. The entire of counselors and trainers, corresponding to 1100 ers par individ functio and 32 der). T Belgiu 2.3. P Firs as mid tionna was a inform lems a The qu the tra trainer volunt 3. Res 3.1. P Firs and m the va seems associ and ne 3.2. R contro For which This s lating R2 = 0 and “p other s future) these r trol su the pa ning” find si Carayo ning” In this structu On work i of the the ne ated in the study, or a response rate of 59 %. Since 82 returned the questionnaire without specifying their e total population is 816 subjects. In all, 399 women n took part in the study (92 agents did not specify gen- ubjects worked in 12 different locations throughout dure ining correspondents for each agency were chosen en to win workers’ trust and to collect the ques- anonymously in ballot boxes, so that confidentiality d. The training correspondents participated in an al meeting to raise awareness about the stress prob- he procedure to follow when collecting the data. onnaire was then distributed in the 12 agencies by correspondents. Every agent, both counselors and ceived a questionnaire. The agents’ participation was inary results means (M), standard deviations (S.D.), minimum um values, and zero-order Pearson correlations of es were calculated (Table 2). On the face of it, it all job control facets are significantly and strongly Additionally, each job control facet was significantly ely correlated to subjective stress. ssion analysis to define a theoretical model of job laining stress y, scientific authors have implicitly used a model in b control subscales influence stress at the same level. ted model corresponds to a linear regression, trans- elations in terms ß (F(6, 765) = 55,464, P < 0.0001, In this regression, the t values for the “resources” ing” subscales are not significant (Table 3). The ales (task management, risks, time management and e significant coefficients. In an attempt to explain s, we can admit an important overlapping of job con- les which are intercorrelated (Table 2). In this sense, correlation between the “resources” and the “plan- cales and stress is low (Table 3). Moreover, we can r results for the control of resources dimension in d Zijlstra (1999). The “resources” and the “plan- cales are important from a conceptual point of view. , we postulate that they have a special position in our odel. basis of the definition according to which stress at response by the employee in the face of the demands tion for which the said employee doubts that he has ry resources, and which he feels he has to face up I. Hansez / Revue européenne de psychologie appliquée 58 (2008) 253–262 257 Table 2 Descriptive statistics and intercorrelations of job control facets and subjective stress (Rasch parameters). 2 3 4 5 6 Resources – – – – – Task managem – – – – – Risks 0.38** – – – – Planning 0.61** 0.35** – – – Time manage 0.30** 0.28** 0.54** – – Future 0.43** 0.26** 0.30** 0.08* – Subjective Str −0.50** −0.39** −0.41** −0.38** −0.37** ** P < 0.01; * Table 3 Linear regress Model Correlations VIFa (Constant) Resources (re Task managem Risks (ris) Planning (pla Time manage Future (ave) Dependent va a Collineari to” (De Ke controlling imbalance others, agr prerequisit we will alw variable of Basing o vation of t example, “j teleworking a central r present tim on resourc the job. Pla onomies (e control sca et al., 1993 (1993) go defining se according leads one subscale. Accordi resources i trol over p resources agement o future on t influence th subs ting llow n the N Mean S.D. Min Max 1 783 0.64 0.73 −1.60 3.14 – ent 786 0.72 0.76 −1.62 3.31 0.59** 786 0.66 0.54 −1.20 3.64 0.31** 786 0.32 0.61 −1.56 2.76 0.47** ment 786 0.44 0.74 −1.46 4.14 0.29** 785 0.44 0.82 −1.76 3.78 0.32** ess 776 −0.84 1.12 −10.0 2.07 −0.39** P < 0.05. ion of job control facets on subjective stress. Std. Coef. (Beta) t Sig. −1.73 0.08 s) −0.04 −1.17 0.24 ent (exi) −0.20 −4.44 0.00 −0.16 −4.87 0.00 ) −0.03 −0.66 0.51 ment (tem) −0.17 −4.76 0.00 −0.20 −5.85 0.00 riable: subjective stress. ty statistic VIF: variance inflation factor (ideal value = 1; worrying if > 18). yser and Hansez, 1996, p. 133), the perception of the resources is the first stage in the perception of an and, therefore, of stress. Frese (1989, p. 111), among ees with the idea that the “resources” subscale is a e in his proposal of job control subscales. In this way, ning” stress. Tes thus a stress i ays consider this “resources subscale” as the first job control-stress models we intend to test. urselves, among other things, on an empirical obser- he new forms of work organisation, such as, for ust-in-time management”, virtual enterprises or even , it is clear that control over work planning plays ole in the perception of working conditions at the e. This work planning should also depend directly es that the worker has at his disposal to perform nning control is very often cited in job control tax- .g. Zapf, 1993) and in existing multidimensional job les (Breaugh, 1985; de Jonge et al., 1994; Jackson ; Sargent and Terry, 1998). Frese (1989) and Zapf beyond the idea of the importance of planning in veral decision or control opportunities, particularly to the sequence of actions and to planning. This to suppose a central role for the planning control ng to our theoretical model (Fig. 1), control over s considered as an initiator of the process. The con- lanning is a mediator between the control over the on the one hand, and the control over the man- f the task, the time constraints, the risks and the he other. These four job control subscales directly e generation of stress. The “Resources” and “Plan- analysis of relative im stress (seco 3.3. Statist Covaria in order t considerati reliability i define the Zero-order Partial Part −0.34 −0.04 −0.03 1.64 −0.44 −0.16 −0.13 2.17 −0.36 −0.17 −0.15 1.24 −0.38 −0.02 −0.02 2.10 −0.32 −0.17 −0.14 1.47 −0.36 −0.21 −0.17 1.27 cales would therefore have an indirect effect on the fit of this theoretical model to a data sample should us to know if all job control facets will influence same way (first research question). A more detailed this model should also give us indications about the portance of job control aspects in the prediction of nd research question). ical analyses nce matrices were performed using Lisrel 8.30 o analyse the structural models. Two statistical ons have to be formulated about the variables. The ndices of the Rasch parameters were considered to latent variables (Table 4) (Anderson and Gerbing, Fig. 1. Theoretical model of job control and stress. 258 I. Hansez / Revue européenne de psychologie appliquée 58 (2008) 253–262 Table 4 Observed variables and latent variables in the structural models. Rasch parameter (“Observed variable”) Variance error (δ) Latent variable λ parameter Control of resources RASRES 0.22 Res 0.77 Task management control RASEXI 0.12 Exi 0.89 Risk control RASRIS 0.11 Ris 0.79 Planning control RASPLA 0.12 Pla 0.82 Time management control RASTEM 0.13 Tem 0.87 Future control RASAVE 0.16 Ave 0.88 Subjective stress RASMSP 0.32 Msp 0.86 Table 5 Goodness of fit statistics. 2 2 S Model 1 0 Model 2 0 Model 3 0 Model 1 is th l 3 m an exogenous dardis adjusted good alidat 1988, p. 41 of each of equal to (1 858). If th latent varia indicator a Moreov have defini tant to defi variable. T that latent of 0 and va in Lisrel) o equal to 1) have only o the first tec 3.4. Testin In all, t statistics an tively. In the fi (Tables 5 an values of th Table 6 R-square for l Variable Task managem Risks Planning Time manage Future Subjective str s are se to 5). eyon reve f 2.5 f 9.9 orat del 2 indic ance on heor bsca 2). M istics xcep χ df P value χ /df RMSEA Test for close fit 198.50 12 0.000 16.54 0.14 0.00 8.36 4 0.07 3.32 0.038 0.66 11.23 8 0.18 1.4 0.023 0.94 e basic theoretical model. Model 2 authorizes some error correlations. Mode variable. RMSEA: root mean square error of approximation; Std. RMR: stan ness of fit index; CFI: Bentler’s comparative fit index; ECVI: expected cross-v 5). In Lisrel 8.30, this means fixing the error variance the observed variables, if the error variance value is -reliability)*variance (Farkas and Tetrick, 1989, p. e error variance is determined, the relation of each ble to its indicator is equal to the square root of the lpha coefficient. er, latent variables, which are not observed, do not te scales. To define the model properly, it is impor- ne the origin and the unit of measure of each latent wo different techniques can be used: (a) to postulate variables are standardised, so that they have a mean riances equal to 1 in the population (standard method r (b) to fix a coefficient different from 0 (normally for the relation with the observed indicator. As we ne indicator for each latent variable, we have chosen hnique. t value not clo (Table value b matrix value o value o to elab Mo some Covari relaxed in the t trol su Table fit stat cant, e g the theoretical structural model hree models have been tested. The goodness of fit d the R-squares are shown in Tables 5 and 6 respec- rst stage, the initial model (model 1) has been tested d 6). The analysis of the parameters shows that all the e structural coefficients are superior to 0.05 and that atent variables in the three models. R-square Model 1 Model 2 Model 3 ent 0.80 0.60 0.86 0.35 0.66 0.35 0.79 0.60 0.67 ment 0.36 0.26 0.81 0.26 0.13 – ess 0.51 0.48 0.50 stress, whic is rather low statistics ar nificant. St and further Model 3 in Fig. 2. It between va model 2, w only does t but on task 2 The follo control/future ning control/r (pla/exi). Som management control/future trol (RASRIS td. RMR GFI AGFI CFI ECVI AIC .053 0.93 0.84 0.90 0.30 230.50 .012 1 0.98 1 0.07 56.36 .011 1 0.99 1 0.06 51.23 odifies the basic model mainly by considering future control as ed root mean square residual; GFI: goodness of fit index; AGFI: ion index; AIC: Akaike’s information criterion. all significant. Standard errors are all acceptable and 0. But the goodness of fit statistics are not acceptable The chi-square is significant and the RMSEA has a d the acceptable margins. The standardised residuals als that some residuals go largely beyond the critical 8, with a minimum value of −7.05 and a maximum 2. The modification indices were taken into account e the second model. is close to the basic theoretical model but allows ators and latent variables to have correlated errors. errors between some pairs of variables have been the basis of modification indices.2 This modification etical model is justified if we remember that job con- les are strongly and significantly intercorrelated (cf. odel 2 is acceptable (Table 5) as far as goodness of are concerned. Structural coefficients are all signifi- t the t value of the risk control subscale on subjective h is inferior to 1.96. The R-square of future control (Table 6) in other respects. General goodness of fit e acceptable even if the chi-square is not totally sig- andardised residuals are all in the acceptable margins modification indices are not proposed. is illustrated in the form of standardised coefficients proposes more in-depth modifications in the relations riables. As future control was explained slightly in e consider it as an exogenous variable in model 3. Not his variable have a direct effect on subjective stress, management control and risks as well. wing pairs of variables have been relaxed: task management control (exi/ave); planning control/future control (pla/ave); plan- isk control (pla/ris); planning control/task management control e error covariances of Rasch parameters are also relaxed: time control/planning control (RASTEM/RASPLA); time management control (RASTEM/RASAVE); risk control/task management con- /RASEXI); risk control/planning control (RASRIS/RASPLA). I. Hansez / Revue européenne de psychologie appliquée 58 (2008) 253–262 259 Task ma manageme any covaria as (and eve nificant par and standar In Fig. 2 The more (ß = 0.82) a ning plays manageme agement co more task m becomes d future cont (ß = 0.19) a have a dire be all the m (ß =−0.25 (ß =−0.25 50% of the age is in a in the scien and Terry, 1998). The the mediator. R tations. Th manageme direct effec (t(771) = 4. tion of a to between co agement co control of r ment (ß =− P = NS and sider a glob only the re agement co mediator in The goo sented in T the conclusion that model 3 is the only model that fits the data. In general, models 2 and 3 seem to fit the data better odel 3 in diffe t nes he o t is sligh valu cal p ation ess. ind l. Fu s con rrors pro rs. cuss aim ode nna e W ta sa ce s e ind s in e ob in the t for s tha spec trol direc con fou eme l. Ho ers. ces her trol Fig. 2. Final model of job control facets and stress. nagement depends on available resources, and time nt results from task management. In this third model, nce error is correlated. Results (Table 5) are at least n more) interesting as those found for model 2: sig- ameters, acceptable general goodness of fit statistics dised residuals in the acceptable margins. , the first exogenous variable is control of resources. resources the worker has, the higher are planning nd task management (ß = 0.54). As expected, plan- a central role: it has a positive direct effect on task nt (ß = 0.33), risk control (ß = 0.49) and time man- ntrol (ß = 1.48). Another important result is that the anagement is controlled, the more time management ifficult (ß =−0.83). The second exogenous variable, rol, has a significant direct effect on risk perception nd task management (ß = 0.17). Four control facets ct effect on subjective stress. Subjective stress will ore low since task management (ß =−0.23), risks ), time management (ß =−0.22) and future control ) increase. This model makes it possible to explain variance of subjective stress (Table 6). This percent- ccordance with, and even superior to, results found tific literature (e.g. de Jonge et al., 1999; Jimmieson 1993; Sargent and Terry, 1998; Schreurs and Taris, oretical model presented planning control as a total esults in model 3 do not totally confirm our expec- e fact that control of resources influences task nt control through planning control does not cancel a t of control of resources on task management control 07, p < 0.001). We have also tested the assump- than m model square are no sider t close fi AGFI ECVI empiri expect tive str and an contro way a ance e it will worke 4. Dis The stress m questio (i.e. th our da influen to hav aspect First, w stress accoun appear other a the con The in is then 3. Only manag contro the oth influen The ot job con tal mediator by considering separately the relation ntrol of resources and risk control and time man- ntrol respectively. The structural relations between esources and risk control (ß = 0.08) and time manage- 0.56) are not significant, respectively t(771) = 0.51, t(771) =−0.48, P = NS. Moreover, when we con- al model including these three additional relations, lation between control of resources and task man- ntrol is significant. Planning control is then a partial this model. dness of fit statistics for the three models are pre- able 5. Strictly considering the chi-square leads to on subjecti almost the conclude t stress is ide practically It is also us to better This is an i ward in ou The transac that the co the process 1. There are few differences between model 2 and goodness of fit statistics. Unfortunately, the chi- rence cannot be tested statistically since our models ted models. Model 3 seems slightly better if we con- ther goodness of fit indices: the P value for test of higher and nearly perfect, the RMSEA is lower, the tly higher and the parsimony index (AIC) and the e are in favour of model 3. As far as theoretical and oints of views are concerned, model 3 meets our s more. The job control subscales influence subjec- There is a direct effect for four job control subscales irect effect for control of resources and planning ture control is an exogenous variable, in the same trol of resources. The last point concerns covari- . Since model 3 does not allow covariance errors, bably be easier to replicate it in other samples of ion of this paper was to develop a theoretical job control- l with the development of a multifaceted job control ire covering important job aspects concerning stress OCCQ). Testing the fit of this theoretical model to mple could enable us to know if all job control facets tress in the same way (first research question) and ications about the relative importance of job control the prediction of stress (second research question). serve that not all of the facets of job control influence same way. A linear regression seems insufficient to the richness of our job control subscales. Clearly, it t the resources available are important and influence ts of work. Planning seems to be a partial mediator for of time management and the risks inherent in work. t effect of control of resources and planning control firmed whatever model we use, i.e. model 2 or model r facets of the WOCCQ directly influence stress: task nt; time management; risk control; and, finally, future wever, this latter facet intervenes independently of Since it is related to the concept of job insecurity, it the direct control that workers have within their jobs. question concerns the extent of the influence of the facets on stress. As far as direct effects of job control ve stress are concerned, structural coefficients are same and vary between −0.22 and −0.25. We can hat the effect of the four job control subscales on ntical since their percentage explanation of stress is the same. important to emphasize that our final model enables explain stress than those proposed in the literature. mportant result, which should allow us to move for- r understanding of the phenomenon being studied. tional definition of stress is reinforced to the extent ntrol of resources is actually the starting point of . The model also shows that planning, judged to be 260 I. Hansez / Revue européenne de psychologie appliquée 58 (2008) 253–262 important in the contemporary context of firms, plays a major role in the stress process. In additi three main model we p considered These resu siders futu Future con cept of job authors gen to maintain objectives is directly also mentio studied as ies illustrat the explana important r influences Our model has a direc working co future has a and on the m interest tha current con regarding h order to ma also the ris Subsequ as a total m other job c mediator fo control of r result is no ning in actu resources is himself to and to man requires. O the worker and materi Finally, relationship But this ca level of the daily basis, sharing are dependent be that the are concer manageme low contro in a definit But we construct v correlations between the six control subscales of the WOCCQ and the job control subscales of Karasek (1979), i.e. authority n an er (1 pact e m thori inat is tim e jo se t of K easu erth re. se. T nship aints task The t. M ely a th onst s and onst onsi uctio trol c 989 elate racti con chara of scale rk w t is of ips b it i ctive g th n pr h, e cas searc nclu con ions d to urth ting fy th on to the answers of the two basic research questions, observations emerged from the analysis of the final ropose in this paper. First of all, future control was as an exogenous variable in the final structural model. lts support Carayon’s (1993) assumption which con- re and career dimensions as exogenous variables. trol can be understood as a variable close to the con- insecurity as defined by Hellgren et al. (1999). These erally define job insecurity as a feeling of incapacity sufficient motivation in order to attain the defined in an uncertain work environment. This definition linked to the concept of control. De Witte (1999) ns that this concept of job insecurity has been rarely a chronic stressor. Nevertheless, in the three stud- ed by the author, this concept has a strong effect in tion of psychological stress. Our study confirms the ole of this concept because in our model this facet subjective stress in the same way as the other facets. also allows us to prove that this control subscale also t effect on the way in which workers perceive their nditions. It is noted that the scale of control over the n influence on the control over risks inherent in work anagement of tasks on a day-to-day basis, hence the t this concept has for intervention, especially in the text of uncertainty. The more the worker feels secure is professional future, the better he will perform in nage not only all his tasks on a day-to-day basis, but ks and responsibilities within his job. ently, we notice that planning control does not act ediator between the control of resources and the ontrol subscales. Planning control is only a partial r risk control and time management control because esources has a direct effect on task management. This t surprising as we know the importance of work plan- al organizational functioning. Having the necessary not sufficient. The worker needs to plan his activities some extent in order to decrease the risks in his job age all of the time constraints that the organisation n the other hand, to fulfil all his tasks on a daily basis, needs to directly control the informational, relational al resources at his disposal. the last observation concerns the unexpected negative between task management and time management. n be partly explained if we refer to the hierarchical task. If a worker has a high control over his tasks on a insofar as work objectives, task procedures and work clearly defined, this means that he or she is probably on a higher hierarchical level. In this case, it could worker has less autonomy as far as time constraints ned. For instance, in the Taylorism approach, task nt is clearly defined to the worker but he or she has a l on the required work pace and standard production e working time. can offer another interpretation of this result. In the alidity analyses of the WOCCQ (Hansez, 2001), the decisio Spreitz and im that tim the au determ that th sent th are clo model to a m Nev the sco decrea relatio constr for the ments. relevan positiv the ide time c straint time c case, c sor red of con In 1 lems r of inte of job other results ate a of wo case, i notion tionsh or not perspe siderin questio researc In any tive re stress. 5. Co We tradict respon now. F interes to veri d skill discretion and the empowerment subscales of 995), i.e. meaning, self-determination, competence have been calculated respectively. We have noticed anagement control was not strongly correlated with ty decision subscale of Karasek (1979) and the self- ion subscale of Spreitzer (1995). We can conclude e management subscale did not adequately repre- b control concept. Since we know that these items o the items of demands in the job control-demand arasek (1979), these items have certain similarities re of demands. eless, it is necessary to be careful when interpreting The higher the score, the more the time constraints herefore, according to our model 3, the negative between the task management control and the time should be interpreted as follows: the higher the score management control, the greater the time require- explanation linked to the hierarchical level remains odel 3 also supposes that the control of planning is correlated to time constraints. This would suppose at “the more I plan my work myself, the lower my raints.” Finally, the relationship between time con- stress could be expressed as follows: “the lower my raints, the less I am stressed out.” But even in this dering control according to the Frese’s (1989) stres- n mechanism remains plausible even if other models an be considered. , Ganster and Fusilier revealed a number of prob- d to research on job control, especially in terms ve effects (Ganster and Fusilier, 1989). The effects trol, it seemed, were difficult to separate from cteristics of the work situation. It is true that the this study prove just how difficult it is to cre- of control that deals with the different aspects ithout mixing up closely-related concepts. In this a matter of the concepts of job insecurity and the job demands. Given the complexity of the rela- etween job control subscales, we can ask whether s a good strategy to position ourselves within the of stressor reduction (Frese, 1989) when con- e different facets of control. The answer to this obably presupposes the necessity of complementary specially at the level of possible interactive effects. e, this underscores the interesting side of quantita- h for a more theoretical reflection on the process of sion clude that the causal model contains very few con- in relation to the existing literature. It allows us to certain questions, which have been left open until ermore, it also brings up questions that would be to test in the future. It would be beneficial to be able e extent to which the theoretical model brought to I. Hansez / Revue européenne de psychologie appliquée 58 (2008) 253–262 261 light by the data from the public institution, can be applied to other work environments. A multigroup analysis in Lisrel should allow us to Another which has, trol. Accor seems to b analysis an Elovainio e more to co rounding s surroundin ceptual stu such as jo interesting. individual o or more su job charact stress. One im same-sourc the study. are also ne Apart fr explaining whereby jo an importa combating sions of the in this resp ing job con the worker his work. T possibility or supervis underlining the worker sufficiently cient, espec of planning straints im work. Fina worker wit context of dom are o involves co made in th supervisor. Beyond the interven tion of the users of the the form an of the item next stage study offer interventio fact of being informed about the results of the diagnosis and secondly that the results are clearly diffused. Only in the third n is t the e co t of of t WOC proc imp sis, e ignifi ion o nce n, J.C view a 423. n, G. er, S.L y, Ch ., Isra nal s –115 J., K of wo , J.A 51–5 , P., 1 477. , P., Z strain 8. n, M . Job ers: A ss Me e, J. and-c equa e, J., rol-su nal of e, J. tse A strich n). G e, J of d cts on –116 er, V. ravail ail 33 e, H., ature k and io, M l and ultilev lth Ps .M., V nal of respond to this question. study should focus on other variables besides stress, nevertheless, only a limited relationship to con- ding to the literature, job satisfaction, for example, e a good path to explore. Spector’s (1986) meta- d more recent studies (e.g. Sargent and Terry, 1998; t al., 2000) clearly show that this variable is linked ntrol than the stress variable. But the confusion sur- tudies of job satisfaction is just as important as that g studies of stress. Therefore, a preliminary con- dy would be indispensable. Organizational variables b challenge and work commitment would also be The upcoming studies could also take into account bjective characteristics such as the hierarchical level, bjective ones such as negative affectivity. Specific eristics could also contribute to the explanation of portant limitation of our study has to do with the e self-report data and the cross-sectional nature of A longitudinal study and objective stress measures eded. om the recognition of the concept of job control in the well-being of workers, the study of the process b control influences well-being seemed to us to be nt step in understanding how to orient the action for stress. The structural model on the six control dimen- WOCCQ enables us to formulate recommendations ect. Three conditions seem to be essential in help- trol have a positive effect on well-being. First of all, has to control the resources he needs to carry out his implies having the necessary information, the to make decisions, the potential help of colleagues ors, or benefiting from sufficient cognitive resources the important role of continuous training. Secondly, must have the possibility to plan his/her work in a autonomous manner in order for him/her to be effi- ially in changing work environments. This autonomy should help him/her to manage the many time con- posed on him/her and the risks inherent to his/her lly, the management should ideally also provide the h job security regarding his/her future. But in this changing work environments, the margins of free- bviously narrow. Prevention at this level probably ntinuous information being given about the changes e company and adequate social support from the these practical recommendations about the content of tion, we have also evaluated the potential of interven- WOCCQ. In a preliminary study of 24 interviews of WOCCQ diagnosis, the results are positive regarding d the structure of the questionnaire, the readability s and their fit to reality. The real problem lies in the when one goes from diagnosis to intervention. The s some evidence about the brakes and incentives of n. Among the stimulating factors, the first one is the positio cited in tion ar the fac instead as the in the to give diagno is as s ticipat work. Refere Anderso A re 411– Aronsso Saut Wile Baker, E patio 1145 Barling, role Breaugh 38, 5 Carayon 463– Carayon and 32–4 De Croo 2000 driv Stre de Jong dem tural de Jong cont Jour de Jong trich Maa datio de Jong tion Effe 1149 De Keys au t Trav De Witt liter Wor Elovain tiona A m Hea Elsass, P Jour he participation of the workers, a success factor often scientific literature. As far as the brakes on interven- ncerned, we found organizational changes but also focusing the intervention on a specific department he whole company. In conclusion, a diagnosis such CQ must be considered as the necessary first stage ess of preventing stress at work. It is also necessary ortance to the communication of the results of the specially in a context of organizational change. This cant, or even more significant, than the active par- f the workers in the process of preventing stress at s ., Gerbing, D.W., 1988. Structural equation modeling in practice: nd recommended two-step approach. Psychological Bulletin 103, , 1989. Swedish research on job control, stress and health. In: ., Hurrell, J.J., Cooper, C.L. (Eds.), Job control and worker health. ichester, pp. 75–90. el, B., Schurman, S., 1996. Role of control and support in occu- tress: An integrated model. Social Science and Medicine 43, 9. elloway, E.K., 1996. Job insecurity and health: The moderating rkplace control. Stress Medicine 12, 253–259. ., 1985. The measurement of work autonomy. Human Relations 70. 993. Job design and job stress in office workers. Ergonomics 36, ijlstra, F., 1999. Relationship between job control, work pressure : Studies in the USA and in the Netherlands. Work and Stress 13, ., Van Der Beek, A.J., Blonk, R.W.B., Frings-Dresen, M.H.W., stress and psychosomatic health complaints among Dutch truck re-evaluation of Karasek’s interactive job demand-control model. dicine 16, 101–107. , Janssen, P.P.M., van Breukelen, G.J.P., 1996. Testing the ontrol-support model among health-care professionals: A struc- tion model. Work and Stress 10, 209–224. Kompier, M.A.J., 1997. A critical examination of the demand- pport model from a work psychological perspective. International Stress Management 4, 235–258. , Landeweerd, J.A., van Breukelen, G.J.P., 1994. De Maas- utonomielijst: Achtergrond, constructie en validering (The t Autonomy Questionnaire: Background, construction and vali- edrag en Organisatie 7, 27–41. ., Mulder, M.J.G.P., Nijhuis, F.J.N., 1999. The incorpora- ifferent demand concepts in the job demand-control model: health care professionals. Social Science and Medicine 48, 0. , Hansez, I., 1996. Vers une perspective transactionnelle du stress : Pistes d’évaluation méthodologiques. Cahiers de Medecine du (3), 133–144. 1999. Job insecurity and psychological well-being: Review of the and exploration of some unresolved issues. European Journal of Organizational Psychology 8 (2), 155–177. ., Kivimaki, M., Steen, N., Kalliomaeki, L.T., 2000. Organiza- individual factors affecting mental health and job satisfaction: el analysis of job control and personality. Journal of Occupational ychology 5 (2), 269–277. eiga, J.S., 1997. Job control and job strain: A test of three models. Occupational Health Psychology 2, 195–211. 262 I. Hansez / Revue européenne de psychologie appliquée 58 (2008) 253–262 Farkas, A.J., Tetrick, L.E., 1989. A three-wave longitudinal analysis of the causal ordering of satisfaction and commitment on turnover decisions. Journal of Applied Psychology 74, 855–868. Frese, M., 1989. Theoretical models of control and health. In: Sauter, S.L., Hurrell, J.J., Cooper, C.L. (Eds.), Job control and worker health. Wiley, Chichester, UK, pp. 107–128. Ganster, D.C., Fusilier, M.R., 1989. Control in the workplace. In: Cooper, C.L., Robertson, I. (Eds.), International Review of Industrial and Organizational Psychology. Wiley, Chichester, UK, pp. 235–280. Hansez, I. (2001). La validation du WOCCQ: Vers un modèle structural du stress et du contrôle de l’activité de travail. Thèse de doctorat en psychologie non publiée, Université de Liège, Liège. Hellgren, J., Sverke, M., Isaksson, K., 1999. A two-dimensional approach to job insecurity: Consequences for employee attitudes and well- being. European Journal of Work and Organizational Psychology 8 (2), 179–195. Hurrell, J.J., Lindstrom, K., 1992. Comparison of job demands, control, and psy- chosomatic complaints at different career stages of managers in Finland and the United States. Scandinavian Journal of Work, Environment and Health 18, 11–13. Hurrell, J.J., Mc Laney, M.A., 1989. Control, job demands and job satisfaction. In: Sauter, S.L., Hurrell, J.J., Cooper, C.L. (Eds.), Job control and worker health. Wiley, Chichester, UK, pp. 97–103. Israel, B.A., Schurman, S.J., 1990. Social support, control and the stress process. In: Glanz, K., Lewis, F.M., Rimer, B. (Eds.), Health behavior and health education: Theory, research and practice. Jossey-Bass, San Francisco, CA, pp. 187–215. Jackson, P.R., Wall, T.D., Martin, R., Davids, K., 1993. New measures of job control, co Psycholog Jimmieson, N work stre Psycholog Jones, F., Fle Winnubst chology. W Jöreskog, K.G Software Karasek, R., 1 cations fo Kasl, S.V., 19 In: Sauter health. W Kristensen, T.S., 1995. The demand-control-support model: Methodological challenges for future research. Stress Medicine 11, 17–26. Lemyre, L., Tessier, R., Fillion, M., 1990. Mesure de Stress Psychologique (MSP) : Manuel d’utilisation. Éditions Behaviora, Québec. McKnight, J.D., Glass, D.C., 1995. Perceptions of control, burnout and depres- sive symptomatology: A replication and extension. Journal of Consulting and Clinical Psychology 63, 490–494. McLaney, M.A., Hurrell, J.J., 1988. Control, stress and job satisfaction in Cana- dian nurses. Work and Stress 2, 217–224. Mullarkey, S., Jackson, P.R., Wall, T.D., Wilson, J.R., Grey-Taylor, S.M., 1997. The impact of technology characteristics and job control on worker mental health. Journal of Organizational Behavior 18, 471–489. Sargent, L.D., Terry, D.J., 1998. The effects of work control and job demands on employee adjustment and work performance. Journal of Occupational and Organizational Psychology 71, 219–236. Schreurs, P.J.G., Taris, T.W., 1998. Construct validity of the demand- control model: A double cross-validation approach. Work and Stress 12, 66–84. Smith, C.S., Tisak, J., Hahn, S.E., Schmieder, R.A., 1997. The measurement of job control. Journal of Organizational Behavior 18, 225–237. Spector, P.E., 1986. Perceived control by employees: A meta-analysis of stud- ies concerning autonomy and participation at work. Human Relations 39, 1005–1016. Spreitzer, G.M., 1995. Psychological empowerment in the workplace: Dimen- sions, measurement and validation. Academy of Management Journal 38, 1442–1465. Terry, D.J., Jimmieson, N.L., 1999. Work control and employee well-being: A decade review. In: Cooper, C.L., Robertson, I.T. (Eds.), International iew o 5–14 Doef, holog Stress D., Ja sures nal of al-Ba in ts. In 1. F.R.H beter anisat gnitive demand, and production responsibility. Journal of Applied y 78, 753–762. .L., Terry, D.J., 1993. An experimental study of the effects of ss, work control and task information on adjustment. Applied y: An International Review 47, 343–369. tcher, B.C., 1996. Job control and health. In: Schabracq, M.J., , J.A.M., Cooper, C.L. (Eds.), Handbook of work and health psy- iley, Chichester, UK, pp. 33–50. ., Sörbom, D., 1999. PRELIS 2. User’s Reference Guide. Scientific International, Chicago. 979. Job demands, job decision latitude, and mental strain: Impli- r job redesign. Administrative Science Quarterly 24, 285–308. 89. An epidemiological perspective on the role of control in health. , S.L., Hurrell, J.J., Cooper, C.L. (Eds.), Job control and worker iley, Chichester, UK, pp. 161–190. Rev pp. 9 Van der psyc and Wall, T. mea Jour Widersz trol effec 29–4 Zijlstra, hoe Org f Industrial and Organizational Psychology. Wiley, London, UK, 8. M., Maes, S., 1999. The job demand-control(-support) model and ical well-being: A review of 20 years of empirical research. Work 13, 87–114. ckson, P.R., Mullarkey, S., 1995. Further evidence on some new of job control, cognitive demand and production responsibility. Organizational Behavior 16, 431–455. zyl, M., Zolnierczyk, D., 1995. Dimensions of job con- computerized and traditional office work and its health ternational Journal of Occupational Safety and Ergonomics 1, ., Den Hoedt, M.C.J., de Vries, R.E., 2000. Job control, hoe meer ? Een studie naar de relatie van control met stress. Gedrag en ie 13, 1–12. The "Working conditions and control questionnaire" (WOCCQ): Towards a structural model of subjective stress Introduction From unidimensional scales of job control to multifaceted job control scales From a new multifaceted job control scale to research questions Method Material Multifaceted job control scale Psychological state of stress measure Subjects Procedure Results Preliminary results Regression analysis to define a theoretical model of job control explaining stress Statistical analyses Testing the theoretical structural model Discussion Conclusion References