Clinical definitions determining the size of bullied workers versus data driven estimation with latent cluster analysis
Guy Notelaers1, Jeroen Vermunt,2, Stale Einarsen3 & Hans De Witte1
1: Katholieke Universiteit Leuven, Belgium; 3: Bergen University, Norway; 2:Tilburg Univeristy, Netherlands
ConclusionBulliying measured with the NAQ is a multidimensional construct when the answers to the negative acts are not recoded into ‘yes’ or ‘no’. Nonetheless there is proof for a multidimensional approach, the latent factor solutions are inferior to the latent cluster solutions.
This latent cluster solution is multidimensional by nature. It captures six classes of respondents of which only one is to be considered as a victims cluster. Therefore we conclude that 3% of the population is victim of bullying at work. However there are other cluster who are clearly exposed to bullying but the occurence of bullying is not high enough to cause serious (z<-1) health consequences. These clusters are referred to as latent bullying and workrelated bullying. Three clusters are not really related to bullying. For riskassessment these cluster can be interesting in that two clusters can be dedected where preventive measures can avoid bullied persons to become victims.
Compared to other ways of categorizing respondents into classes i.e. operational criteria it is clear that the latent cluster approach has the most discriminatory power and captures best the self-judgement of respondents (subjective method).
Maybe the latent cluster approach suits a phase model that has been suggested troughout the years (Bjorkvist, Einarsen, Zapf, Leymann). Ph ase I : no actsPhase II : at a very low level workrelated acts emerge. Phase III : the level of workrelated acts remains constant but some personal directed acts ermerge at a very low level. Phase IV : intensification (towards now and then and sometimes between now and then and once a month). Phase V : intensification of work related acts (once a month). Phase VI : overall intensification (once a month-week)
contact : [email protected]
ReferencesEinarsen, Raknes, Matthiesen & Hellesøy, (1994). The negative acts questionnaire .Mikkelsen, E. Einarsen, S. (2001) Bullying in Danish work-life: Prevalence and Health
correlates. European Journal of Work and Organizational Psychology, 10, 4, 393 – 413.Salin, D. (2001) Prevalence and forms of bullying among business professionals: a
comparison of two different strategies of measuring bullying. European Journal of Work and Organizational Psychology, 10, 4, 425-411.
Vermunt, J.K. Magidson, J. (2002) Latent Class Cluster Analysis. In : Hagenaars, J.&McCutcheon (eds), Applied Latent Class Analysis, Cambridge : Cambridge University Press, 89-106.
Vermunt, J.K. Magidson, J. (2003) Latent Gold : Users GuideLatent Gold : Users Guide. Statistical Innovations.
Zapf, D. Knorz, C. Kulla, M. (1996) On the Relationship between Mobbing Factors and Job Content, Social Work Environment, and Health Outcomes. European Journal of Work and Organizational Psychology, 5, 2, 215-237.
Zapf, D. Einarsen, S. Hoel, H. Vartia, M. (2003) Empirical findings on bullying in the workplace. In: Einarsen, S. Hoel, H. Zapf, D & Cooper, C. (Eds) Bullying and Emotional Abuse in the Workplace. Taylor & Francis, London, 103-124.
THE 4th INTERNATIONAL CONFERENCE ON BULLYING AND HARASSMENT IN THE WORKPLACE (Bergen, Norway, 2004)
IntroductionResearch shows that bullying at work is a widespread phenomenem. In some research bullying at work takes the form of a desease...
-Scandinavian reserach (Einarsen & Skogstad, 1996; Leymann, 1996): 4% - British studies 10% - Netherlands (Hubert & van Veldhoven, 2001) 2 tot 3% - In Belgium prevailence rates are between 10 (Notelaers & De Witte, 2003a; Notelaers & De Witte, 2003b) and 16% (Opdebeeck, et. al, 2002).
These rates differ a lot. That questions how we measure bullying and how we decide who is bullied and who is not.This is a major problem for organisations who are meeting the EC directive about riskanalysis. Following the control cycle (Cox & Gonzales, 2000) it is vital to make an inventory of the psychosocial risks and to make a link between the risks and its healh consequences in order to
- Eliminate the risk (primary prevention)- Prevent the risk (secondary secondary)- ‘Cure’ the victims (tertiary prevention)
Thus it is essential that our categorization methods can capture or discriminate between who is a victim of bullying and who is not.
Measuring bullying at workSubjective
Until now different definitions were used but nowaday in Europe many use Einarsen & Skogstad ‘s (1996) definition. But many have used different responscategories and different time – reference (e.g. 6 and 12 months)
ObjectiveBy means of questionnaire without referring directly to ‘bullying’.
LIPT (46 items)NAQ (22 items, 29 items, 32 items, 17 items) and
NAQ-r => different prevailence rates of bullying
Categorizing victims/non-victims Different decisions to attribute victims and non victims are in fact questioning the dimensionality of bullying concept. Applying the operational definition of one or two acts during the last six months is assuming a one dimensional construct.
But the literature reveals : - 5 dimensions according the effect
they cause (Leymann, 1996)
- 7 dimensions by Knorz&Zapf (1996)
- 5 dimensions by Einarsen & Raknes (1997)
- 2 dimensions by Einarsen & Hoel (2001)
- 4 dimensions in Dick & Rayner’s instrument (2004)
How to categorize victims / non – victims when dimensions are related/ non related? Which dimension? How to weight? How are dimensions related?
For riskanalysis it is a problem to decidewhether someone is bullied or not with a multidimensional concept.
Aim of this researchTo inspect wheter bullying, as it has been measured by the NAQ without dichotomizing the responscategories into yes and no, is a multidimensional measurement. And if so to to inspect whether a latent cluster approach (Vermunt&Magidson, 2002) can help to classify respondents in homogenous groups according to their exposure to bullying.
Gedruckt im Rechenzentrum der Universität Leipzig
Sample6175 observations stem from two kinds of research :research to inventarise wellbeing (14 studies) and research with focus on mobbing (4 studies) (see the other poster (Notelaers, et. al, 2004). MethodSubjective methods ‘Are you being bullied at work?’ (‘never’, ‘sometimes’, ‘often’ and ‘always’) ‘Are you being bullied at work during the last six months?’ (‘never’, ‘sometimes’, ‘often’ and ‘always’ )‘Are you being bullied at work the last six months according to our definition’(no, never’, ‘yes, seldom’, ‘yes, sometimes’, ‘yes, weekly’ and ‘yes, daily’) Objective method : Belgian NAQ 17 items
Results
cond. prob. answ. 'never'
0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
0,8
0,9
1
information jokes under level taking job gossip exclusion priv. life Insultes quit job mistakes silence no value no opinion funny supr. misusew ork tohard
not bullied at all not bullied nor / nor
latent victim work related bullying severe victim
cond. prob. ans. 'now and then'
0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
0,8
0,9
information jokes under level taking job gossip exclusion priv. life Insultes quit job mistakes silence no value no opinion funny supr. misusework tohard
not bullied at all not bullied nor / nor
latent victim work related bullying severe victim
cond. prob. answ. 'once a month'
0
0,05
0,1
0,15
0,2
0,25
0,3
0,35
0,4
information jokes under level taking job gossip exclusion priv. life Insultes quit job mistakes silence no value no opinion funny supr. misusework
not bullied at all not bullied nor / nor
latent victim work related bullying severe victim cond. prob. answ. 'once a week or more
0
0,1
0,2
0,3
0,4
0,5
0,6
information jokes under level taking job gossip exclusion priv. life Insultes quit job mistakes silence no value no opinion funny supr. misusework tohard
not bullied at all not bullied nor / nor
latent victim work related bullying severe victim
Discriminatory power in z-values
- 0,61- 0,370,11- 1,01- 0,378- 0,458- 0,110,060,29Quality of sleep
- 0,49- 0,330,09- 0,72- 0,387- 0,35706- 060,23worrying
- 0,56- 0,360,11- 0,88- 0,455- 0,4450,014- 040,31Recovery need
- 0,6- 0,40,12- 0,593- 0,563- 0,240,042- 0,0620,28Involve-ment
- 0,79- 0,530,16- 1,03- 0,62- 0,370,0710,0210,32Pleasure at work
2 acts
1 act0 actsevere victim
work related bullying
latent victim
nor / nor
not bullied
not bullied at all
z- values
operational cutt off’s
Latent Cluster ModelCluster model – operational
definitions in %
1003,227,618,715,428,936,24column total
9,282,94,081,20,250,880two acts
90,70,323,537,515,12836,24no / one act
2 acts or more
20,63,065,973,32,145,091,0251 act or more
79,40,161,645,413,223,835,21no act
1 act or more criterium
row totalsevere victimwork related bullying
latent victimnor / nor
not bullied
not bullied at all
Latent Cluster Model
Classifications methods –subjective measure bullying
,51,38,38Clustermodel
,38,28,26Mikkelsen & Einarsen
,36,26,24Leymann
Legal definition of bullying
Were you bullied at work during the last six months
Are you being bullied at work
Spearman correlation
Six clustermodel for bullying at work
Evaluation of clustermodel
Checking First Hypothesis : LCA-apprach
0.0040927,2110118689,34-Factor-correlated
0,1660,050,38837276346112464,26- cluster direct lang. effects and 3 local dependencies (see : Notelaers, et. al, 2004)
0,0144597,57104122307,84-Factor – uncorrelated(Einarsen & Rakness, 1997)
0,110,0260,3436802,441031145042-Factor correlated
0,150,010,3039361,56101117045,82-Factor uncorrelated(Einarsen & Hoel, 2001)
0,100,3437037,72102114730,62-Factor correlated
0,21010.0060,3337164,84100114840,51-Factor-5L
0,150,040,4033710,043911139018-Cluster
0,140,060,3934023,52342113790,97-Cluster
0,130,0260,3834567,42293113911,36-Cluster
0,110,0020,3735323,54244114243,95-Cluster
0056344,9481335711-Cluster-base model
Class.Err.bootstrap p-value
prop red errorL²NparBIC(LL)
0,320,1070,030,0090,0130,002
once a week or more
0,2170,1170,080,0170,0240,004
once a month
0,310,2250,580,3360,2460,067
once a week
0,1520,5510,310,6380,7180,927
never
654321order of latent clusters
0,0320,0830,090,1650,2770,353
proportion (size)
victimwork related bullying
latent bullying
nor / nor
not bullied
not bullied at all
2,41,621,71,41,271,08you work to hard
1,81,241,31,061,061,01misuse of work
2,41,311,61,241,011,01‘funny suprises’
32,692,21,681,841,09opinion does not count
3,12,512,21,591,761,07job and efforts not valued
31,821,91,251,211,01opening encounters hostile or silence
2,41,531,71,331,181,02rep. remarks about mistakes
2,31,331,41,081,031remarks that you should quit
2,61,241,61,311Insultes
31,431,91,561,061,02rep. remarks private life
2,61,491,71,211,161,01exclusion from group
3,32,042,21,921,361,15gossip
2,52,071,81,141,431,04taking job / competences
2,72,5521,511,91,29work under level
3,21,652,21,671,11,07jokes
3,12,652,21,632,11,35witholding information
severe victim
work related bullying
latent victim
nor / nor
not bullied
not bullied at all
1: never2 : now and then3: once a month4: once a week or more
Discussion : a phase model for bullying at work?