65
A Multi-Level Analysis of the Demand-Control and the Demand Control Support Models of Stress in the Workplace. David L. Morrison (U. of Western Australia) Roy L. Payne (Curtin University) and Toby D. Wall (University of Sheffield)

A Multi-Level Analysis of the Demand-Control and the Demand

Embed Size (px)

DESCRIPTION

 

Citation preview

Page 1: A Multi-Level Analysis of the Demand-Control and the Demand

A Multi-Level Analysis of the Demand-Control and the Demand Control Support

Models of Stress in the Workplace.

David L. Morrison (U. of Western Australia)

Roy L. Payne (Curtin University)

and

Toby D. Wall (University of Sheffield)

Page 2: A Multi-Level Analysis of the Demand-Control and the Demand

Stress

• Stress – “The confusion created when one’s mind tries to override the body’s basic tendencies to choke the living shit out of some arsehole who desperately deserves it.” (Seen on front of a T-shirt, Perth Royal Show, 1997)

Page 3: A Multi-Level Analysis of the Demand-Control and the Demand

Stress (cont.)

• Stress – occurs when people are faced with events they perceive as endangering their physical or psychological well-being (Atkinson et al., 1996)

Page 4: A Multi-Level Analysis of the Demand-Control and the Demand

Stress

• Ability - Demand: Is a perceived imbalance between an individual’s perception of demands and the perceived ability to meet those demands

• Motive - Supply: The capacity to which a job supplies fulfillment of the individual’s motives

Page 5: A Multi-Level Analysis of the Demand-Control and the Demand

Models of Stress

Page 6: A Multi-Level Analysis of the Demand-Control and the Demand

Terminology: Stress and Strain

• Strain can be defined as “the state of where the person is experiencing physical discomfort or pain which is indicating that bodily processes are at work in an effort to return the body to its normal homeostatic state” (Fletcher & Payne, 1980)

• In short, strain is the state of being stressed• Stressors are said to be the causes of strain

Page 7: A Multi-Level Analysis of the Demand-Control and the Demand

Stress and Work

• Stress might be said to arise when there is a mis-match between the person & the job (P-E fit) (French e al., 1982)

• Eight Dimensions of Fit

Income Overtime

Length of Service Education level

Job complexity Workload

Responsibility Role ambiguity

Page 8: A Multi-Level Analysis of the Demand-Control and the Demand

Magnitude of the Problem

• 60% of absence caused by stress related disorders.

• 100 million working days lost in the UK because people cant face going to work.

• Over a $billion spent in replacing people who retire due to CHD.

Page 9: A Multi-Level Analysis of the Demand-Control and the Demand

Psychological Effects

• The Medical Research Council (Fraser, 1947) reported that

• 9% men and 13% women in light and medium engineering factories had disabling neurosis (anxiety, depression, obsessionality)

• 19% males and 23% females suffered minor neuroses

• Women exhibiting neurosis were absent for 13% of the time they could have been at work.

• In an Australian study of 516 telegraphists, Ferguson (1977), reports that 33% had a disabling neurosis (anxiety, nervous exhaustion) .

Page 10: A Multi-Level Analysis of the Demand-Control and the Demand

Who has all the stress

Page 11: A Multi-Level Analysis of the Demand-Control and the Demand

Who has all the Stress?

Page 12: A Multi-Level Analysis of the Demand-Control and the Demand

Is work totally to blame?

• Russeck (1962) "PROLONGED emotional strain associated with job responsibility... preceded the [Coronary] attack in 91% of cases as compared with an occurrence of similar strain in only 20% of normal control subjects”.

But• only 10% of men who show two or more standard risk factors (e.g.,

high blood pressure, cardiac and endocrinological reactivity or lability, serum cholesterol, triglyceride, glucose intolerance and related illnesses) developed heart disease in a 10 year study.

• Over 40% who did develop CHD did not have two or more of the risk factors.

Page 13: A Multi-Level Analysis of the Demand-Control and the Demand

Personality may play a part

Page 14: A Multi-Level Analysis of the Demand-Control and the Demand

1.64 p=.002

-1.52 p=.038

1.23, p=.24

1.20 p=.38

-2.00 p=.001

2.05 p=.003

Page 15: A Multi-Level Analysis of the Demand-Control and the Demand

Karasek’s (1979) Model: A unifying Theory

• Karasek has proposed a model by which stress in the workplace can be explained as a function of psychological demands and decision latitude of the occupation

Page 16: A Multi-Level Analysis of the Demand-Control and the Demand

Karasek’s (1979) Model: A unifying Theory

• Karasek has proposed a model by which stress in the workplace can be explained as a function of psychological demands and decision latitude of the occupation

Page 17: A Multi-Level Analysis of the Demand-Control and the Demand

Psychological Demands

• Psychological demands relate to the mental arousal or effort necessary to accomplish a task.

• Such demands can be explained as a function of task requirements such as deadlines and quotas

Page 18: A Multi-Level Analysis of the Demand-Control and the Demand

Decision Latitude

• Decision Latitude is a combination of two subcomponents:

• decision authority – the social authority over making decisions

• skill discretion – the breadth of skills usable in a job

Page 19: A Multi-Level Analysis of the Demand-Control and the Demand
Page 20: A Multi-Level Analysis of the Demand-Control and the Demand

Karasek’s (1979) Model

• Karasek proposed that the two variables combine to form four distinct occupation types

Page 21: A Multi-Level Analysis of the Demand-Control and the Demand

High-Strain Jobs

• High-Strain jobs are associated with a high level of psychological demands with a low decision latitude

• Examples of high-strain jobs include: Telephone Operator and Waiter

Page 22: A Multi-Level Analysis of the Demand-Control and the Demand

High-Strain Jobs (cont.)

• It is predicted that people with high-strain occupation would be most likely to suffer from the negative effects of stress

• This is because the tight constraints of the occupation lead to an inability of an individual to freely release tension brought upon by the high psychological demand

• This tension, instead, must be absorbed

Page 23: A Multi-Level Analysis of the Demand-Control and the Demand

Active Jobs

• Active jobs are associated with a high level of psychological demand and a high decision latitude

• Examples of active jobs include: Electrical Engineer and Teacher

• Active jobs are typically ‘high-level’ jobs

Page 24: A Multi-Level Analysis of the Demand-Control and the Demand

Active Jobs (Cont.)

• From such jobs we predict an optimistic set of psychosocial outcomes learning and growth that are conducive to productivity

• The high level of control serves to counteract the effects of a high demand

• The upper limit to individuals with active jobs is attributed to simple fatigue

Page 25: A Multi-Level Analysis of the Demand-Control and the Demand

Summary of Karasek (1979)

• Karasek’s model predicts four different job types as a function of decision latitude and psychological demands

• Using this model, it is possible to predict the levels of stress experienced by the employees of each job type

Page 26: A Multi-Level Analysis of the Demand-Control and the Demand

Summary of Karasek (cont.)

• Karasek’s model can also be used to predict levels of motivation to learn new skills

• Recall that individuals with passive jobs experienced low levels of stress, BUT also experienced very low levels of motivation

• Many such employees report such feelings as ‘not having to think anymore’

Page 27: A Multi-Level Analysis of the Demand-Control and the Demand

Other additions

• Social Support: an additional dimension to Karasek’s model proposes social support to have a positive main effect on well being and may moderate the impact of stressors on physical and psychological stress.

Page 28: A Multi-Level Analysis of the Demand-Control and the Demand

Social Support

• Overall levels of helpful social interaction available on the job from co-workers and supervisors

• Two hypotheses:– High demands with little support and control leads to

high strain

– Work motivation and learning and development opportunities occur if demands, control and support are high

Page 29: A Multi-Level Analysis of the Demand-Control and the Demand
Page 30: A Multi-Level Analysis of the Demand-Control and the Demand
Page 31: A Multi-Level Analysis of the Demand-Control and the Demand

Social Support

High Social

Support group

High Demand

High Control

High Demand

High Control

Page 32: A Multi-Level Analysis of the Demand-Control and the Demand

Four Basic Models

1. Additive Demand Control Model2. Additive Demand Control Support Model3. Muliplicative D-C Model4. Multiplicative D-C-S Model• Karasek (1989) argues the interaction is

not critical• Ganster and Fusilier argue that it is a

fundamental

Page 33: A Multi-Level Analysis of the Demand-Control and the Demand

Meta Analysis Van Der Doef and Maes (1999). Outcome measure D-C:

StrainD-C: Buffer

D-C-S: Iso-strain

D-C-S: Buffer

N of studies

           

General psychological well-being

28/41 15/31 9/19 2/5 43

Job satisfaction 18/30 10/23 8/14 2/6 31

Burnout 3/4 0/4 1/1 0/2 4

Job-related psychological well-being

7/8 1/2 1/2 1/1 8

Percentage Support 67% 43% 41% 36%

Page 34: A Multi-Level Analysis of the Demand-Control and the Demand

Summary• Roughly half the studies support all four models

for each DV. Slightly less support buffering models

• 7/8 support the direct strain (additive) D-C model• All the supportive models have a common method

problem• None of the 9 longitudinal studies support the

models. • All those that failed used objective measures of

the DV.

Page 35: A Multi-Level Analysis of the Demand-Control and the Demand

So Why Don’t We Find the Interaction?

• Problems with operationalising demands and control:

• Items are complex and reflect several characteristics

– My job demands creativity (autonomy and complexity)

– Is your work hectic? (Machine pace or impossibility of taking a break)

• Specific vs general measures – – Work method, work scheduling or work goals

– Proximal vs distal control

• Non-linear relationship between demand and control

Page 36: A Multi-Level Analysis of the Demand-Control and the Demand

Methodological Problems: Research Populations

• The effects of the many wash out the influence of the few (range restriction)

• There is too much diversity in job characteristics which adds noise and therefore reduced correlation

• The differences in health related behaviours across SES is too large and these can hardly be distinguished from job related risks

• Response bias within samples (different frames of reference)– Denial– Verbal insight

Page 37: A Multi-Level Analysis of the Demand-Control and the Demand

Non-Work Effects on measures

• It may be that external environmental features contribute to the levels of stress experienced by an individual – stress elsewhere makes the job tougher

• Non-work factors can influence mental health

Page 38: A Multi-Level Analysis of the Demand-Control and the Demand

Personality and The Model

• Stress inhibits learning: hi anxiety inhibits the capacity to accept challenge. Anxiety may accumlulate over time or some are more susceptible

• Learning inhibits stress: new learning may lead to events being perceieved as less stressful stimulating confidence

• Personality may interfere with susceptibility

LOW

1A

3

4

HIGH

Feeling of Mastery

Accumulated Anxiety Inhibits Learning Attempts

Feelings of Mastery InhibitsStrain Perception

Time

Time

Accumulated Strain

Decision Latitude

LOW HIGH

Psychological Demands

B

2

1

ActiveLearning

ResidualStrain

3

Page 39: A Multi-Level Analysis of the Demand-Control and the Demand

Statistical problems

• Power.

• Where on the continuum is the sample and does this affect the interaction?

• Missing variables - eg. Efficacy, negative affect.

• Lack of consistency in measurement.

• Level effects;

Page 40: A Multi-Level Analysis of the Demand-Control and the Demand

A level Effect?

• A proper test as described by Karasek would involve comparing jobs and symptomatology rates for the jobs

• The objective studies use the job to assign measures of demand and control but drop the job when it comes to the analysis

• This is a completely mis-leading and incorrect analytical technique.

Page 41: A Multi-Level Analysis of the Demand-Control and the Demand

Multilevel nature of the problem

• Takes into account the hierarchical structure of the data (e.g., people nested within jobs)

• Most tests of the Karasek model have violated a fundamental assumption of OLS regression that residuals are uncorrelated.

• To ignore hierarchical structure underestimates error terms which in turn over estimates statistical reliability of regression weights

• Multilevel techniques allow randoms intercepts and slopes to be modelled.

Page 42: A Multi-Level Analysis of the Demand-Control and the Demand

Inflation of alpha level in the presence of intra class correlation (Barcikowski 1981)

N .01 .05 .20

10 .06 .11 .28

25 .08 .19 .46

50 .11 .30 .59

100 .17 .43 .70

Page 43: A Multi-Level Analysis of the Demand-Control and the Demand

OLS Regression Model

Level one variation

Page 44: A Multi-Level Analysis of the Demand-Control and the Demand

Random Slopes Model

Job variation with respect to level

Job

Sat

isfa

ctio

n

Job Demand

Page 45: A Multi-Level Analysis of the Demand-Control and the Demand

Random Slopes and Intercepts Model

Level two slopes and intercept variation

Job

Sat

isfa

ctio

n

Job Demand

Page 46: A Multi-Level Analysis of the Demand-Control and the Demand

Present Study

• Karasek’s model tested mostly at the individual level and is problematic mostly due to common method variance and measurement confounding.

• Group level studies are not instructive for the following reasons:

• Buffering model not explicitly tested• Job characteristics measured at the individual level actually

ask about group level phenomena• Confounding of group proxy variables and individual level

variables• Small sample size in the one ‘acceptable’ study.

Page 47: A Multi-Level Analysis of the Demand-Control and the Demand

The present study is more comprehensive than other studies hitherto:

• Large sample size at individual and group level

• Multi-level analysis (circumvents common method variance to some extent)

• Control for quadratic effects• Separate step by step analysis of individual

and group level effects.• Comprehensive in that Demand, Control and

Support additive and buffering models examined.

Page 48: A Multi-Level Analysis of the Demand-Control and the Demand

Sample• 13,199 employees from 15 Hospital trusts representing 61-

65% response rate across trusts• 130 job titles• 9,327 completed all questions relevant to present study

• 3441 nurses 1075 doctors• 1467 admin staff 750 managers• 1409 allied medical professionals 594 ancillary staff• 591 professional and technical staff

• Excluding jobs with less than 10 people and those with missing data left a sample of 6,771 nested within 86 jobs (mean grp size =79).

Page 49: A Multi-Level Analysis of the Demand-Control and the Demand

Measures

• Demands (do not have enough time to carry out my work)

=.91

• Control (determine methods and procedures) =.88

• Support from colleagues (count on colleagues to help with a

difficult task at work) =.91

• Job Satisfaction (intrinsic and extrinsic) =.88

• GHQ12 (lost much sleep over worry) =.90

Page 50: A Multi-Level Analysis of the Demand-Control and the Demand

Data Preparation

• Scores on all measures were normalised

• Quadratic terms were prepared for each job characteristic

• Interaction terms were derived from the cross products of normalised variables

• Group level variables were calculated from the multi-level average.

Page 51: A Multi-Level Analysis of the Demand-Control and the Demand

Multi-level Regression

• Step 1 - Variance components• Step 2 - Demand and control, quadratics plus

interaction (Individual Level) • Step 3 – Support, quadratic plus interactions• Step 4 – Repeat step 2 at job level• Step 5 – Repeat step 3 at job level• Improvements in model fit determined from

the Deviance statistic (-2*loglikelihood ratio)• Variance explained for each level presented.

Page 52: A Multi-Level Analysis of the Demand-Control and the Demand

Individual & Job level correlation matrix

D C S Sat G1. Demand .40 .78 .20 .09 .14

2. Control .22 .53 .21 .17 .10

3. Support -.10 .15 .20 .07 .02

4. Satisfactn -.29 .29 .43

5. GHQ .41 -.01 -.22 -.42

Different jobs have different support networks

Page 53: A Multi-Level Analysis of the Demand-Control and the Demand

Intra class Correlations

• Jobs with the same title have common features but the social mileu is different as are the people

• Support=.027• Demands = .17• Control = .25

• These results are comparable to Van Yperen and Snidjers (2000) with .21 and .33 for demand and control.

Page 54: A Multi-Level Analysis of the Demand-Control and the Demand

Results for Job SatisfactionIndividual Level

Rsq Equn Level 1 Level2 Ch. Dev

Step 1 (Variance Components) .054*Step 2 (Demand &

Control main effects) .210 .206 .278 1570.5****

Step 4 (Demand x Control) .212 .208 .278 9.4**Step 5 (Support main effect) .324 .328 .241 106.3**Step 6 (Support Quadratic) .327 .331 .259 26.7**Step 7 (2-way Support interactions) .328 .332 .259 7.9*Job Level Variables

Step 9 (Demand and Control main effects)  .344 .322 .556 22.9***Step 10 (Quadratic Effects)  .350 .322 .685 14.2***Step 12 (Support main effect) .353 .322 .722 4.3**

Demand ß=-.352 Control ß=.257

Interaction ß=.033Support ß=.355

Support2 ß=-.048 Satisfaction lower under low and high support Job Control ß=.193

Job Demands ß=.162

Unstable

DemandxSupport ß=-.029 High Demand High Support low Satisf.

Job Support ß=-.98 unstable

Page 55: A Multi-Level Analysis of the Demand-Control and the Demand

Rsq Equation Level 1 Level2 Ch. Deviance

Individual Level Variables

Step 1 (Variance Components) .018*

Step 2 (Demand & Control main effects) .177 .163 .889 1259.2****

Step 5 (Support main effect) .203 .192  .833  30.8****

Total Level 1 Variance .206 .194 .833

Job Level Variables Step 13 (Support quadratic) .208 .194 .944 5.9*

Total Model Variance .208 .194 .944

Demand ß=.41***

Control ß =-.076**

ß =-.172***

ß=-.617 high and low levels of support have lower GHQ

GHQ12 Analysis

Page 56: A Multi-Level Analysis of the Demand-Control and the Demand

Summary Results Individual Level Analysis

• Demand-Control additive model supported for GHQ and Satisfaction

• Evidence for the active jobs hypothesis for satisfaction only

• The D-C-S model is supported for both satisfaction and GHQ

• No support for D-C-S buffer model

Page 57: A Multi-Level Analysis of the Demand-Control and the Demand

Summary Results Job Level Analysis

• Job level effects apparent only for satisfaction• Demand-Control additive model – partial support

as control but not demand supports the model. Demand acts in the opposite direction

• Demand-Control Buffer Model – no support• D-C-S Support acts in the opposite direction to

that expected• D-C-S Buffer Model – No support

Page 58: A Multi-Level Analysis of the Demand-Control and the Demand

Conclusions

• Results from the individual level seem to be inconsistent with the job level

• When job is the unit of analysis there is little consistent support for the Karasek models

• Individual level analyses should be interpreted with caution as the common method effect could be very high. The longitudinal analyses reported by Van der Doef and Maes would support this conclusion

Page 59: A Multi-Level Analysis of the Demand-Control and the Demand

Conclusions• Individual level analyses should be interpreted with caution

as the common method effect could be very high. The longitudinal analyses reported by Van der Doef and Maes would support this conclusion

• The “poor job level effects” Could be due to – Range restriction– Simpsons paradox is also evident and could account for the

demand effect. Highly satisfied people concentrated in the high demand jobs.

– The job demand-control correlation is very large at the job level (r=.78 vs r=.22)

• The latter raises the question can we ever have a fair test of the model

Page 60: A Multi-Level Analysis of the Demand-Control and the Demand

Implications

• Although differences between jobs doesn’t account for large amounts of overall variance – but large amounts of it can be accounted for with

a few variables.• Individual perceptions of job characteristics accounts for 20%

variance and yet the management of perceptions has received little attention.

• The Karasek model has little to say about the effects of jobs on well-being …

• It does have something useful to say about the perceptions of working conditions and their relationship to well-being and job attitudes.

Page 61: A Multi-Level Analysis of the Demand-Control and the Demand

Multi-level Tests of The Models

• Soderfeld et al (1997) – combined demand control measure (n=4756, g=284)

• This an additive test• Job variance = 2-8% of which 83% accounted for

by DC-S measures• Individual level variance accounted for = 28%• Shortcoming: job chars measured at the individual

level – are the effects group effects or individuals in contexts?

Page 62: A Multi-Level Analysis of the Demand-Control and the Demand

Is stress at work determined by factors at the

group pr individual level?Van Yperen and Snidjers (2000)

• n=260; g= 31• Group agreement on job chars of =.24-37• Less agreement on outcome variables =.05• Using deviation scores from the group mean for individual

level effects and group means for the group effects – • High job demands relative to the groups affected mental well-being

especially when they also had low control• More sick days taken when the group had low control and the

individual thought the demands were high.

• These results illustrate the potential importance of a multi-level analysis62

Page 63: A Multi-Level Analysis of the Demand-Control and the Demand

Shortcomings

• The effects for other dependent variables were not consistent only 2 of 8 demand and control interaction effects were significant (low n?)

• All interactions between I.V.s tested at once

• Failure to control for quadratic effects could have inflated the importance of the intereactions

• The authors fail to report where the variance was accounted for

Page 64: A Multi-Level Analysis of the Demand-Control and the Demand

De Jonge et al., (1999)Hospital Sample

• N=895; g=64• Tested demands and control and buffering effects at the individual and

group levels.• But the demand measure asked about the unit not the individual job –

what effect does this have?• Demand x control assessed at the unit and the job levels• Intra class correlations varied from .06-.12• No interaction effects at either level• Large amounts of variance accounted for at the group level (45-100%

of up to 12%).for job satisfaction but not mental health• Small amounts at the individual level (4-17% of up to 85%) for both

types of DV

Page 65: A Multi-Level Analysis of the Demand-Control and the Demand

Shortcomings

• Group dummy variables and individual level job characteristics variables added to the analysis in one hit.

• Difficult to tell how much variance the individual level variables account for

• Group dummy variables proxy work characteristics such as control and demand

• Unfair to say mental health variables not predicted by group level job characteristics variables as there was nothing left to predict.