JCQ SCALE

Journal of Occupational Health Psychology Copyright 1998 by the Educational Publishing Foundation 1998, Vol. 3, No. 4, 322-355 1076-8998/98/$3.00

The Job Content Questionnaire (JCQ): An Instrument for Internationally Comparative

Assessments of Psychosocial Job Characteristics

Robert Karasek University of Massachusetts Lowell

Chantal Brisson Laval University

Norito Kawakami Gifu University

Irene Houtman and Paulien Bongers National Institute for Work and Health

Benjamin Amick New England Medical Center

Part I discusses the Job Content Questionnaire (JCQ), designed to measure scales assessing psychological demands, decision latitude, social support, physical demands, and job insecurity. Part II describes the reliability of the JCQ scales in a cross-national context using 10,288 men and 6,313 women from 6 studies conducted in 4 countries. Substantial similarity in means, standard deviations, and correlations among the scales, and in correlations between scales and demographic variables, is found for both men and women in all studies. Reliability is good for most scales. Results suggest that psychological job characteristics are more similar across national boundaries than across occupations.

This article consists of three parts. Part I introduces the Job Content Questionnaire (JCQ) as a tool for psychosocial job assessment. First, a description of scales and their underlying theoretical concepts is presented. This is followed by a discussion of empirical issues in the development of the questionnaire and its validity. Part I concludes with a discussion of measurement issues, administrative issues, and future challenges.

Robert Karasek, Department of Work Environment, University of Massachusetts Lowell; Chantal Brisson, Department of Preventive and Social Medicine, Laval University, Quebec City, Quebec, Canada; Norito Kawakami, School of Medicine, Department of Public Health, Gifu University, Gifu, Japan; Irene Houtman and Panlien Bongers, National Institute for Work and Health, Amster- dam; Benjamin Amick, The Health Institute, New England Medical Center, Boston.

The Job Content Questionnaire (JCQ) is copyrighted. Users should request the instrument from the JCQ Center (see the JCQ Usage Policy section). The JCQ is provided with research documentation to most users free of charge, but commercial and very large research projects pay a usage fee to support comparative reliability analysis and instrument development on a nonprofit basis through the JCQ Center, University of Massachusetts Lowell.

Correspondence concerning this article should be addressed to Robert Karasek, Department of Work Environ- ment, University of Massachusetts, Lowell, Massachusetts 01854.

Part II reports the cross-national validity, for men and women, of the JCQ scales in six broadly representative populations from four advanced industrial societies: the United States, Canada, the Netherlands, and Japan. JCQ scale means, standard deviations, reliabilities, and correlations are compared. Part III reviews comparison of the intercountry and interoccupation differences in the scales, discusses specific scales issues, and discusses the implications of the study for interpretation of psychosocial job assessment questionnaires.

PART I

The Job Content Questionnaire (JCQ): Psychosocial Job Assessment Instrument

Instrument Overview

The JCQ is a self-administered instrument designed to measure social and psychological characteristics of jobs. The best-known scales--(a) decision latitude, (b) psychological demands, and (c) social support--are used to measure the high-demand/low- control/low-support model of job strain development. The demand/control model predicts, first, stress-related risk and, second, active-passive behavioral correlates of jobs. Other aspects of work demands are assessed as well:

322

SPECIAL SECTION: JOB CONTENT QUESTIONNAIRE 323

(d) physical demands and (e) job insecurity. The instrument has a recommended length of 49 questions.

All scales can be used for microlevel, job- characteristic analytic purposes, such as assessing the relative risks of individuals' exposures to different work settings to predict job-related illness development, psychological distress, coronary heart disease, musculoskeletal disease, and reproductive disorders. The scales also allow testing of hypotheses about activation, worker motivation, and job satisfaction and have been used for such studies. The conceptual framework underlying the JCQ allows its application in social policy as a measure of work quality (Karasek, 1998), in addition to the more commonly assessed work quantity issues: wages, hours, and benefits. Broader economic development issues of skill utilization as well as social costs of market-based economic development are beginning to be addressed (Karasek & Theorell, 1990) using the instrument. No personality orientation scales or measures of non-job stressors are included--two areas in which the user may want to supplement the instrument.

The JCQ has been translated into over a dozen languages. The instrument is nationally standardizable by detailed occupation in several countries, providing an occupational scoring system. An active users' group supports usage of the JCQ, and an international board of researchers decides on policy and development issues.

JCQ Scales and Their Theoretical Bases

psychological strain occur when the psychological demands are high and the worker's decision latitude is low: job strain. Low social support at work further increases risk. A second set of hypotheses, related to what might be called good stress, involves active behavior development under conditions of high demands and high decision latitude, which predict motivation, new learning behaviors, and coping pattern development (of course, the active behavior hypotheses are contingent on demands not being too high). The reverse is predicted for low demands coupled with low decision latitude: a very unmotivat- ing job setting leading to negative job learning or gradual loss of previously acquired skills.

A dynamic version of the model integrates the job strain and active behavior hypotheses with personality characteristics measuring accumulated strain and self-esteem development (Karasek & Theorell, 1990) with the goal of predicting strain development and learning over time. The model is based on measures of psychological demands of work combined with a measure of task control and skill use (decision latitude). The psychological demand dimension re- lates to "how hard workers work" (mental work load; Meshkati, Hancock, & Rahami, 1990), organization constraints on task completion, and conflicting demands. It includes subscales shown in Table 1 (Subscales 2a, 2b; see Karasek & Theorell, 1990). The "recommended version" includes additional specific measures of cognitive workload (Subscaies 2c, 2d).

The JCQ arose out of the adaptive response to serve the new empirically based areas of social epidemiology, behavioral medicine, and psychosocial job analysis, requiring a multidisciplinary theoretical model. Because the primary theoretical model upon which the JCQ is based and discussions of alternative scale formulations in its domain are extensively reviewed elsewhere (de Jonge & Kompier, 1997; Karasek, 1979, 1997; Karasek & Theorell, 1990; Kristensen, 1995, 1996; Landsbergis, SchnaU, War- ren, Schwartz, & Pickering, 1994), this article only presents outlines of those arguments by way of an introduction to the JCQ scales.

Scale 1: Decision Latitude and Scale 2: Psychological Demands

The most commonly used demand/control model hypothesis (Karasek, 1979, 1997; Karasek & Theo- rell, 1990) predicts that the most adverse reactions of

Scales la and lb: Components of Decision Latitude--Skill Discretion and Decision Authority

The worker's control over the performance of his or her own job is measured by two theoretically distinct subdimensions of decision latitude that are usually highly correlated: skill discretion and decision authority (Karasek & Theorell, 1990). Skill discretion (Subscale la) is measured by a set of questions that assess the level of skill and creativity required on the job and the flexibility permitted the worker in deciding what skills to employ (similar to variety; e.g., Hackman & Lawler, 1971). A second subdimension, decision authority (Subscale lb), assesses the organizationally mediated possibilities for workers to make decisions about their work (autonomy; e.g., Hackman & Lawler, 1971). A question on skills required by the job allows assessment of skill underutilization (Subscale lc). A

324 KARASEK El" AL.

Table 1 Scales and Numbers of Questions in the Full Recommended JCQ and the "Core QES"

Scale Core QES JCQ Full recommended JCQ

1. Decision latitude a. Skill discretion b. Decision authority c. Skill underufilization d. Work group decision authority (new) e. Formal authority (new) f. Union/representative influence (new)

2. Psychological demands and mental workload a. General psychological demands b. Role ambiguity c. Concentration (new) d. Mental work disruption (new)

3. Social support a. Socioemotional (coworker) b. Instrumental (coworker) c. Socioemotional (supervisor) d. Instrumental (supervisor) e. Hostility (coworker) (new) f. Hostility (supervisor) (new)

4. Physical demands a. General physical loading b. Isometric load (new) c. Aerobic load (new)

5. Job insecurity a. Generaijob insecurity b. Skill obsolescence (new)

Total que~ions

6 6 3 3 2 b 2 b

3 2 3

4 5 1 1

1 2

2 2 2 2 2 2 2 3

1 1

1 1 2 2

3 4 2

27 49

Note. JCQ = Job Content Questionnaire; QES = Quality of Employment Surveys. a Eight new scales/dimensions and additional items were added to make the Recommended JCQ format. also used in this scale.

b Education was

third, macrolevel component of decision latitude assesses the possibility of participatory influence on organization level issues, as well as union and work-group participation (Subscales ld, le, If).

The JCQ integrates use of both individual and occupation-based conceptions of job characteristics. The occupation-based job characteristic assessments yield an effective communication tool for interpreting the meaning of the otherwise abstract psychosocial JCQ scales in terms of specific jobs situations, and provide a source of validating information about job situations. As an example: When the two JCQ job characteristic scales---decision latitude and psychological demands--are arrayed as a four-quadrant diagram, they define the strain and active behavior hypotheses of the demand/control model. These can be used to display (see Figure 1) average job characteristics of occupations in U.S. Census occupation codes and the U.S. Quality of Employment Surveys (QES) database (Karasek & Theorell, 1990). In Figure 1, the active job quadrant (upper right) with

high demand and high control, has high-prestige occupations: public officials, physicians, engineers,

nurses, and managers of all kinds. The passive job quadrant (lower left), with low demands and low control, has clerical workers such as billing clerks,

and low-status service personnel such as janitors. The high-strain quadrant (lower right), with high demands and low control, has machine-paced operatives such as assemblers, cutting operatives, freight handlers, as

well as other low-status service operatives such as waiters. Occupations with high percentages of women are frequent (garment stitchers, waitresses, telephone operators, and other nurse's aides). Low- strain self-paced occupations (upper left) often involve significant training and self-pacing, such as repairmen, linemen, and natural scientists. Kristensen (1996) reconceptualized the four quadrants above respectively as qualified work, surveillance work, tempo work, and craftsman's work. Kristensen identified the linkages to structural and technological

I I

• Watchman

• Lineman

Decision l

L'tit:dAichite_Lc t

Engineer • Natural Programmer O Farmer

Scientist • ~]~ Teacher--H.S. WIFManager-trade • Publ c • Physician • Officials

+0.50 Bank Officer • Clerk

• Foreman Supervisor O Repairman ..~ O Nurse

Psychological / o_Machinist • Carpenter Demands

050 +0.5~ I I I il~ Fireman t

Stationary • Health • Off. Computer • EngineerA~ I Technician Operator

_ WBilling Clerk ~lhDeliveryman WSales Clerk I

• Dispatcher

1 k -

-0.5o- • Gas Station

O Janitor Attendant • Cutting tWaitress

• Miner - Operative O Nurse's Aide O • Freight handler

Construction • Telephone Laborer Operator/,f--~

" " Keypuncher / '~gh S t n t ~ ~ / ~ a i n O Garme

stitcher,-,,~.,~

O Assembler- electric/trans, mfg.

Figure I. The occupational distribution of psychological demands and decision latitude (U.S. male and female workers; N = 4,495). From "The Political Impfications of Psychosocial Work Redesign: A Model of the Psychosocial Class S~ucture" (p. 177), by R. A. Karasek, in J. V. Johnson and G. Johansson (Eds.), The Psychosocial Work Environment: Work Organization, Democratization, andHealth, 1991, Amityville, N-Y: Baywood. Copyright 1989 by Baywood Publishing Company. Reprinted with permission.

326 KARASEK ET AL.

developments in production processes, thus making it more useful in job redesign contexts.

Scale 3: Social Support

The demand/control model has been expanded by Johnson (Johnson, 1986; Johnson & Hall, 1988) with the addition of social support as a third dimension. The primary hypothesis, that jobs which are high in demands, low in control, and also low in social support at work carry the highest risk of illness, has been empirically successful in a number of chronic disease studies (Johnson, 1989; Karasek & Theorell, 1990). Karasek and Theorell (1990), Karasek, Trian- tis, and Chaudhry (1982), and Johnson and Hall (1988) discussed the differential impacts of support from coworkers and from supervisors and, within these, the separate impacts of instrumental and socioemotional support, respectively (Subscales 3a, 3b, 3c, 3d). Interpersonal hostility is also included as a measure of social support deficit (Subscales 3e, 3f). The social support addition acknowledges the need of any theory of job stress and behavior development to assess social relations at the workplace.

Scale 4: Physical Demands

The "demanding costs" of work activity are not just mental but also physical. Indeed, the more traditional concept of workloads involves physical loads. Physiological effects of stress on the cardiovascular system, the effectiveness of mental functioning, and general fatigue are shown in much research literature to depend on both mental and physical loads, and thus this measure is also included in the JCQ. Although the original QES questionnaire contains only a single item on physical exertion (Subseale 4a), the recommended form of the JCQ includes static (Subscale 4b) as well as dynamic physical loads (Subscale 4c), both shown to be important for musculoskeletal disorder development.

Scale 5: Job Insecurity

Work's psychological burden consists not only of the work of carrying out the task but also in the human costs of adapting to labor market dynamics. These have become increasingly important in the last several years, because the global economy has had job-displacing effects in many countries and increased reported job insecurity (Lohr, 1996). Measure- ment of these items poses statistical challenges (see

the Job Insecurity Scale section) because specific events of unemployment are relatively low frequency, even when the fear of job insecurity can be more widely experienced (Subscale 5a). The job insecurity effect can depend on the labor market requirements for particular skills, limiting future career development possibilities (Subscale 5b).

Theoretical Interpretation of Scales and Implications: Sociological

and Psychological Origins

The JCQ provides advantages to researchers by integrating strengths of multiple disciplines. How- ever, it also presents the JCQ users with the challenge of reconcifing multiple scientific literaanes when findings are reported. We briefly outline major linkages below.

The primary hypotheses of the JCQ, resulting from the demand/control model, are both psychological and sociological in nature and methodology. The JCQ is sociological in that it presumes existence of socially "objective" environments that systematically affect individual well-being and behavior. There is a focus on (a) major social institutional settings, (b) broad population groups, (c) covariance of measures with major demographic categories, and (d) its hypotheses, which evolved from sociological life stress and illness as well as work alienation traditions. Also, consistent with much sociological literature, associations are usually controlled for social class (for contrasting perspectives, see Ganster & Schaubroeck, 1991; Karasek & Theorell, 1990).

However, there are also differences with classic sociological approaches. The JCQ is based on a theoretical foundation that implies an alternative, psychosocial class model, which appears to predict health outcomes more effectively than the conventional class model (Karasek, 1997, 1998; Karasek & Theorell, 1990, chapters 5, 9, 10). Kristensen (1996) and Karasek and Theorell (1990) noted that the scales and methods also lend themselves to understanding of the social and technological structure of production processes. Such associations allow interpretation of JCQ findings for human capital and economic development studies.

A variety of methodological techniques from both sociology and psychometrics are used: scale reliability analysis, scale construction techniques, multilevel causal analyses, and data reduction techniques such as factor analysis. The JCQ uses sociological questionnaire assessment methods to collect valid data on social environments. The JCQ occupation


approach uses the worker's occupation as an alternative unit of analysis to the individual's own job reports, allowing linkage to other insights available for occupational experience.

The JCQ also reflects a psychological focus and is probably used more often by psychologists than by sociologists. It addresses classic psychological topics such as the behavioral basis for emotion-driven psychological distress, psychosomatic illness development, and changes in microlevel behavior related to social situations. There is a presumption that psychosocial experiences are a major determinant of health and well-being, mediated by the neurophysiological mechanisms, as well as behavioral pathways.

Psychologically, the JCQ--demand/control method reflects a stimulus approach, as opposed to a relational approach, which emphasizes personal cognitive interpretation of the person-environment relationship. The JCQ assumes that behavior is, to a significant extent, generated by social environments and their constraints outside the individual. The cognitive psychological claim that decision choices constitute the primary mental workload is contrary to the demand/control hypothesis that social demands are moderated by the behavioral degree of freedom that decision opportunities present (Karasek, 1997). The JCQ-demand/control approach also often treats emotional response as a dependent variable derived from work-related behaviorial requirements.

There are also significant congruencies with psychophysiology, as well as some differences in focus. Most previous stress theories were developed to describe reactions to "inevitable" acute stress in situations threatening biological survival (Cannon, 1914; Seyle, 1936/1976). The demand/control model was, however, developed for work environments in which stressors are chronic, not initially life threatening, and are the products of sophisticated human organizational decision making. The controllability of the stresso r was found to be important and appears to have become even more important as we develop even more complex demands and limitations on individual behavior. However, significant consistency with classic psychophysiology is demonstrated by Frankenhaeuser and Johansson's (1986) psychological research, which shows the congruence of two primary patterns of physiological response (adrena- line related and cortisol related) with the main hypotheses of the demand/control model--allowing linkage among physiological response, social situation, and emotional response patterns.

Empirical Basis of JCQ Scale Development

Origin: Stage IfPre-1984

The JCQ has developed in stages. Historically, the origin of the instrument, which predates the "recommended version JCQ" in 1985, involved analyses of broad pools of job characteristic survey data in two countries. The core questions for the JCQ scales are taken from the three nationally representative samples of the much analyzed QES gathered by the University of Michigan Survey Research Center in 1969, 1972, and 1977 for the U.S. Department of Labor. Each of the three QES surveys was eclectically designed and surveyed over 1,000 aspects of work experience, in a manner often using different questions from survey to survey. Our research group (Karasek et al., 1988; Schwartz, Pieper, & Karasek, 1988) conducted extensive statistical analyses, analyses of theoretical coherence, and analyses of individual questions predicting efficiency for these large groups of questions in the early 1980s. These confirmed that major aspects of the core content of the psychosocial work experience could be captured by the small number of QES questions. On this basis, a small subset of the questions was selected to create the Job Characteristic Linkage System (Schwartz et al., 1988). Approximately two thirds of the linkage system questions were sufficiently similar (with minor adaptations and corrections) across the years to yield common assessment of absolute scale scores: a QES-based JCQ "core" (see Table 1, column 2). The three QES survey question sets had 27 questions in common in the psychosocial area, which allowed development of a pooled sample of all 4,900 respondents, still by far the largest nationally representative U.S. data set on psychosocial job characteristics. This core serves as the source of standard score data for JCQ occupational mean scores and a basis for time-related comparisons in the scales (statistical reliability is discussed in Karasek & Theorell, 1990, Appendix 1).

The validation for the utility of such scales was also developed from a similar set of questions in longitudinal, Swedish nationally representative databases, which could be analyzed extensively for covariations with other social and individual data (Karasek, 1976, 1979; Karasek & TheoreU, 1990) and which had extensive health outcome data. These analyses, while confirming the demand/control model utility, also illuminated the importance of demographic, occupational, and social relations data, and broadly assessed work demand and hazard data

328 KARASEK ET AL.

(Karasek et al., 1981; Karasek, Schwartz, & Pieper, 1983) and implied that a future measurement instrument should not be restricted to the demand/ control task questions alone.

Stage II: JCQ Recommended Version--1985

The next stage was the development of the JCQ Recommended Version 1.1 in response to request for an instrument to assess the psychosocial hypotheses and demand/control models. This is the current version of the instrument. The design of the current JCQ was initiated by request from the U.S. National Heart, Lung and Blood Institute's request for scales for the U.S. Framingham Offspring Study. The JCQ authors were aware of the length limitations posed by research teams in national survey designs and the hesitancy of other researchers adopting question sets not of their own design, thus, the design of the new JCQ instrument in 1984 focused on a very short, efficient questionnaire that could be self-administered in 15 rain, with minimal participant guidance.

The original QES core was not theoretically precise in several areas (particularly psychological and physical demands). To add precision to the theoretical constructs of the QES core, to expand theoretical coverage of both psychological and physical demands, to expand job insecurity and social support scale coverage, and to assist in discriminant validity, we included additional newly drafted questions (see Table 1, column 1) to the QES questions, yielding the present set of scales used in the JCQ data sets. These expansions formed the Recommended JCQ Instru- ment Version (see Table 1, column 2), with 49 questions, which is the most commonly used version (14 additional new questions and 8 additional QES questions had been added beyond the original core). Additional question sets cover physical work hazards, computer interfaces, customer interaction, and psychological strain scales. An update of the recommended version in 1995 (Version 1.5) included pilot versions of a set of global economy questions and more standardizable psychological strain scales.

JCQ Focus and Scale Design Criteria

The multiple goals underlying JCQ construction introduce competing design criteria: (a) standard scale reliability assessment, (b) coverage breadth, (c) scale length economy, (d) scale number economy, and (e) specific content interpretability. A pure concept of statistical reliability means that a set of questions

should assess a single underlying theoretical con- struct. In JCQ design, true statistical reliability is sometimes balanced against a goal of a specific content interpretability. Thus, the scales are also composed of subscales with separate interpretability (see Table 1 and discussion above), a goal which competes with scale statistical reliability. The ques- t ions should also be standardizable questions. The JCQ has also had the goal of covering the most important aspects of qualitative work situations with a small enough number of scales that the interactions between the scales can be feasibly examined.

Toward "Objective "Assessment: Rating Versus Evaluation

An important goal of the JCQ is gathering "objective" data about work environments relevant for prevention-oriented goals of improving social and psychological working conditions. The Swedish Level of Living Survey (Johansson, 1971) was a methodological guide: The bias in the questionnaire response by the participant, while inevitable, was designed to be minimized. The questions are designed to report about, but not evaluate, the participant's usual or main job. Thus, the JCQ questions use simple language so there are meaningful responses possible by all employed respondents, presented in a language simple enough to be understood by participants at all education levels. The response set is designed to assess the validity of the statement about the work environment on a 4-point scale, facilitating the similar quantitative weighting of questions.

Use of participants' own questionnaire reports about their jobs, of course, automatically introduces self-perceptions--the source of the major critique of validity of instruments such as the JCQ (see the Implication for Broad Interpretability of Psychoso- cial JCQs section). In many cases, self-reports on job conditions are the only feasible information-gathering strategy about workers' detailed social working conditions. For example, it would take an outside observer much time to understand the social support situation of the worker. Frese and Zapft (1988) claimed that the risk of self-report bias depends on the degree to which the questions require a complex burden of evaluative cognition by the participant. The JCQ objective assessment goal means that the questions attempt to minimize this self-reflexive component: They report about jobs, but minimize evaluation of them. Questions of the type "lack of decision making is important for me" and "the time


pressures are too much for me" are replaced by questions that emphasize simple assessments of environmental conditions only, such as: "I have freedom to make decisions about my job" and "My job requires I work fast." Such linguistic distinctions have been considered quite significant in other psychological research contexts: for example, state- trait response differentiations based on phrasing differences such as "Today I feel angry" versus "I usually feel angry."

Sources of difficulty remain, however. Self- reflexive judgments remain in two psychological demand questions: "work hard" and "work fast" (see Appendix A). Also, the JCQ goal of broad coverage on jobs characteristics with a short set of questions means that many questions elicit summary judgments about some quality of the job (skill requirements, decision possibilities). Questions about more specific job situations could avoid this problem, but would likely make the questionnaire longer or the questions unjudgable by some respondents, and therefore the responses difficult to compare across groups.

Occupation-Based Analysis and Score Standardization

In addition to direct administration of JCQ questionnaires to workers, the JCQ system offers a second set of occupation-based methodologies. There is an extensive system of JCQ scores scales by detailed occupation and gender in several countries that is the basis of (a) the JCQ occupational score standardization system and (b) the occupational linkage system.

Detailed scoring procedures for the JCQ scales are described in the JCQ Questionnaire and User's Guide (Karasek, 1985). Most of the scales have been standardized by detailed occupation codes for several national populations (for the U.S. population: Karasek & Theorell, 1990, Appendix; Schwartz et al., 1988; with related scales standardized in Sweden [work exposure matrix: Johnson & Johansson, 1991; Johnson & Stewart, 1993]). The JCQ questions can be compared to national scale scores for detailed census code by sex and by four-digit industry code. This allows unique assessment of differences between a target group and "national norms" for psychosocial job dimensions. This allows JCQ users involved in practically oriented job analyses, small populations, or single-plant studies to compare their findings with national averages on the scales (broken down by sex, occupation, and industry).

The same occupational basis that provides the standardized scores is also the basis of an often utilized occupational score linkage system (Schwartz et al., 1988). The JCQ job characteristic scales can be linked to other databases through U.S. three-digit census occupation codes (1970) and also to four-digit U.S. Standard Industrial Classification (industry classification) codes. This database linkage system allows psychosocial job content scores to be associated with health and productivity outcomes in national or company databases already in existence (such as U.S. Census, Commerce, or U.S. National Center for Health Statistics data), for which direct questionnaire data collection would not be feasible.

Aggregate Scoring Methods for Work Groups

Bias of findings could potentially occur with self-reported psychosocial work environment and dependent variables such as depression, exhaustion, and dissatisfaction (see the Implications for Broad Interpretability of Psychosocial JCQs section). One remedy is to aggregate self-report responses by work groups with similar work situations, thus diluting individual biases (Kristensen, 1996). This is, of course, the basis of the occupation database linkage system suggested above, but systems of mixed self-report and work-group aggregated assessment have also been successfully applied (Vahtera, Pentti, & Uutela, 1996).

The alternative of expert observations is certainly theoretically desirable, but in practice it has problems. Expert observations are costly, time consuming, and in assessment of social interactions do not obviously generate more accurate measures resulting in low interrater reliability. There are also theoretical biases involved in the very concept of standard "expert" measures: It is much easier to measure the easily observed, repetitive quality of the low-status assembly- line jobs than the diverse tasks of nigh-status managers or professionals. Thus, measurement reliability for the most potent set of psychosocial job characteristics (decision latitude, skill, and decision autonomy) is probably correlated with scale level--a complex confounding of content and validity (for all methodologies, not just the JCQ).

Scale Statistical Validity

The most substantial compilation of reliability findings is presented in the following section (Part II). However, previous reliability analyses of the scales

330 KARASEK ET AL.

very similar to the JCQ scales have been published for the U.S. national populations (Q.E.S. database: Karasek & Theorell, 1990, Appendix 1; Schwartz et al., 1988). Kawakarni and Fujigaki (1996) and Kawakami, Kobayashi, Araki, Haratani, and Furui (1995) published the first studies on the reliability of recommended format JCQ scales (omitting physical demands and job insecurity). The study concluded that the JCQ is reliable for Japanese populations and found Japanese occupation scale ratings that are similar to those in the United States (see Appendix A). Brisson et al. (in press) showed JCQ scale reliabilities to be good and confirmed the scale structure (Larocque, Brisson, & Blanchet, in press) from both random population survey and a white-collar survey from Quebec, Canada. A 1993 large-scale sample in the United States (Amick, Mangione, & Wu, 1998) reported JCQ scales to be reliable, as well as scale structure confirmation, but some scales differ significantly from the recommended JCQ format. Sante Quebec (1994) showed acceptable JCQ scale reliabilities in the Netherlands, but some factor structure differences arose.

Predictive Validity

It is beyond the scope of this article to review the extensive research literature using the JCQ and JCQ-like scales to predict illness (much research is based on similar, but not exactly equivalent scales). Comprehensive reviews are presented by Marmot and Theorell (1988), Kristensen (1989), Schnall and Landbergis (1994), Kristensen (1995), Kasl (1996), and Tbeorell and Karasek (1996). However, in summary, it can be stated that the JCQ scales and JCQ-like scales demonstrate substantial predictive validity with respect to stress-related chronic disease in international and U.S. research.

Job strain and heart disease associations constitute the broadest base of empirical support for the model. JCQ scales or similar scales associate significantly with cardiovascular mortality using a wide range of methodologies. Landsbergis (augmenting his earlier review [Schnall & Landsbergis, 1994] by personal communication, December 1997) tabulated 72 published studies of cardiovascular disease (CVD) or CVD risk factors testing associations with job strain using JCQ-like scales. Of the 36 studies investigating CVD or mortality, over two thirds showed positive associations (i.e., either all significant or mixed significant positive results) with job strain, and many of these were positive cohort studies. Landsbergis's

update has cataloged 41 studies of the major coronary heart disease risk factors (blood pressure, serum cholesterol, and smoking) testing associations with job strain. In over a dozen studies of blood pressure using sophisticated ambulatory assessment technologies, all show either positive or mixed positive results. However, less-sophisticated blood pressure measurement technologies show no consistent associations, and smoking and cholesterol have mixed positive and null associations.

Consistent associations between mental strain and JCQ-like scales are also reported (see Bourbonnais, Busson, Moisan, & Vezina, 1996; Karasek & Theorell, 1990), but differential effects of job characteristics are noted. Measures of exhaustion and burnout are more consistently associated with high psychological demands, whereas depression and anxiety measures are more strongly associated with low decision latitude.

Occupational musculoskeletal injury prediction is reviewed by Bongers, de Winter, Kompier, and Hildebrandt (1993), who found support for the predictive utility of the demand/control/support model, particularly for upper extremity disorders. Many additional studies using the demand/control model and JCQ scales have been undertaken since then, including associations with pregnancy disorders (Brandt & Neilsen, 1992; Fenster et al., 1995)and immune system disfunctions (Kawakami & Fujigaki, 1996; Peters et al., 1998).

JCQ Measurement and Administrative Issues

JCQ Administration

The JCQ is designed for self-administration and has often been included as a section in other questionnaire instruments in which a short introduc- tory sentence about how to respond to the questions is included. The completion time is short, approximately 15 min for the full recommended version. Professional assistance, such as the research person reviewing the instructions, has also often occurred.

In addition to the standard JCQ questions, JCQ users are encouraged to add their own specific "umbrella questions" that refer to the measurement of specific job conditions in the surveyed work sites. Although the umbrella questions would differ between studies, they could be factor analyzed with the other JCQ questions and correlated with the standard JCQ scales used as reference points.


JCQ Breadth of Use and Scale Consistency

The instrument has been selected for large studies of job conditions in the United States, Canada, Europe, and Japan. The JCQ is being used by large studies in Europe (over 50,000 participants) and Japan (over 40,000 participants) for the study of job strain, heart disease, and absenteeism. The short length and predictive validity appear to be the major reasons for its success. Length is mentioned as a crucial instrument parameter by many users (a substantial number would like an even shorter instrument; there is no validated shorter instrument with standardizable scores at present). Because of the active international collaboration by JCQ researchers, the effort invested bymany individual JCQ projects in the past is now yielding collective benefits with the development of comparative databases that further enhance the data interpretability of each study.

JCQ Usage Study: Compliance With Recommended Format

The JCQ has retained the same recommended format (Version 1.1) since 1985 (see Appendix C). To test the consistency of scale use and to assess the utilization of the JCQ, in 1995 Robert Karasek conducted a survey of all contactable JCQ users. From the 246 recorded requests, 130 projects were estimated to have been completed by that time. The most consistently used scales are the decision latitude scales, with 85% reporting compliance with the standard version. Psychological demands has 69% using the recommended nine-question version and another 19% using the older five-question (QES core) version. Standard version social support was used 74% of the time. However, physical demands are included in only 58% of the studies and job insecurity in only 36% of the studies. The skill utilization question, important for work quality policy and productivity outcomes, is included in 42% of these studies.

Translations Into Other Languages

Authorized translations of the full JCQ instrument from English, which have been specifically approved and which are available from the JCQ Center, are French-Canadian, French-Belgium, Flemish-Bel- gium, Spanish, Swedish, Dutch, Italian, and Japanese. Other language translations are in process. The authorized process requires requesting permission, a

back translation into English submitted to the JCQ Center, and a copy of the translated instrument.

Adaptation of the JCQ for housework and home- role work, student status, and unemployment has also been undertaken but presents challenges because these less-structured social roles mean less specific questionnaire language. Adaptation of the JCQ for low-education workers and workers in less economi- cally developed countries for Spanish-speaking workers has also been undertaken.

JCQ Usage Policy

The JCQ is copyrighted and not published in the public domain; however, it is the goal of the JCQ Center to make it available to all researchers who request it with substantial supporting documentation, and to promote scientific development in the area through a users' network. The JCQ Questionnaire and User's Guide and research documentation are provided free of charge to most users. However, JCQ use by large research studies (over 750 participants) and commercial users requires payment of per-nse charges. Registration in a JCQ users' project database for the users' network for all users and a copy of the researcher's JCQ and demographic data for future reliability analyses (large studies only) are required. Contact the JCQ Center, Department of Work Environment, University of Massachusetts Lowell, Lowell, Massachusetts 01854, for details of policy, fees, and requirements.

Summary of lnstrument Description and Notes on Future Challenges

To summarize description of the instrument, we include at this point several comments that reflect information in the second, empirical section of this article.

1. The 1995 study of usage consistency demon- strates that the JCQ has provided a common set of scales used with consistency throughout almost 100 studies, including large national studies, of psychosocial factors at work. This breadth of use is a unique occurrence for measurement of psychosocial work characteristics in the United States and internationally, and it helps to overcome the major scientific deficit of such instruments: the lack of comparative assessment capability across databases (see discussion in Santer & Murphy, 1995).

2. As a result of the empirical review in the following section (Part II of this article), the JCQ

332 KARASEK ET AL.

recommended version since 1985 (and provisionally augmented in 1995) remains the "recommended version." The need to expand from the earlier "QES core questions" is validated. In particular, some of these additions are crucial for continued psychosocial research: Job insecurity is of increasing importance because of the global economy, and the physical demand scale is of increasing importance because of the increasing prevalence of musculoskeletal disorders in many countries. The "skill level required" questions (the seventh skill discretion question) is crucial for work-quality policy discussions.

The predictive validity of the scales, while not reviewed in detail in this article, is probably the major reason for the success of the instrument. In general, it can be concluded that the JCQ seems to be potentially useful in capturing important elements of psychosocial experience at work in many countries, and thus allows an internationally comparative understanding of "qualitative costs" and "benefits of work."

On the basis of the findings of the analysis of 12 separate male and female populations, we can find no compelling reasons to reject the JCQ scales on the basis of inconsistency of means and standard deviations or Cronbach alpha reliability. However, factor analyses results noted above do raise questions about inclusion of certain questions, a potential basis for future revision of the JCQ scales.

Although some of the tables in the empirical reliability test (Part II of this article) are based on restricted forms of some scales in order to ensure compatibility with several older databases, the largest of the new studies now being collected in Europe and Japan include almost complete versions of all of the recommended JCQ scales, which further bolsters the utility of using the full scales in the future.

3. Although an integrated review of the JCQ scale's shortcomings and future improvement directions is beyond the scope of this article, some preliminary observations about future challenges for the JCQ can be made, reflecting both this article and the current social context of rapidly changing working conditions.

a. The area of work quality needs to be opened up to international political and economic policy discussions, an important challenge given the accelera- tion of global economic linkages. JCQ expansions, such as the pilot versions of the JCQ global economy scales, could assist this. Assessment of socially determined possibilities for control, demand, and social integration deriving from broad global economic changes could be assessed.

b. For psychological workload and social support, improved scale performance may require linguistic tool development and local consensual validation of response meanings. Additional measurement methods beyond questionnaire use involving observations or interview methods that can be linked to the JCQ are also needed. Parts of these activities may involve establishing new languages for social policy purpose in these areas---a social advocacy task.

c. Dropout of highly "stressed" participants from studies, especially in more rigorous scientific protocols, appears to be a much larger problem than previously assumed and is growing as stresses of modern life make participation in scientific studies difficult for certain groups (temporary and transitional employees, marginal economic groups in all settings, and many populations outside of the developed industrial countries).

d. Organization-level job factors are also not studied in the JCQ, and their effects on determining job structures appear to be significant (Warren, 1998). Whether this analysis needs to be a part of an expanded JCQ or separate measurements approach, such topics should be explored for further comparative analyses.

e. New work patterns that involve computer- based communication, network communication, and mass-media communication serve as the mode of social coordination in ever larger numbers of social activities. These may require significantly altered methodologies for assessing psychosocial work experience.

f. Quantitative work aspects of work (particularly hours of work, spouse work time, income, etc.) certainly need no proof of importance but should be simultaneously gathered and used along with JCQ assessments when broad policy implications are to be understood.

g. Psychosocial experiences outside of work in the home and community, and across the life span (jointly with work experience) must be made a part of psychosocial analytic frameworks even when workplace effects are the primary scientific focus.

h. The creative behavior side of psychosocial workplace behavior needs to be given further measurement emphasis, including social interactions which make it possible. To start, more attention should be paid to existing active job hypotheses to facilitate further integration with the active coping health and well-being research themes of psychosocial research. This extension would also allow new


forms of nonmarket productive output to be better assessed, and would assist comprehensive dialogue about costs and benefits of "the new work organization." "Conducive economic policy" could serve as one basis for this expansion (Karasek, 1998; Karasek & Theorell, 1990).

i. Modified assessment methods for psychosocial working conditions in less-developed economies must also be integrated with existing JCQ scales usage, which has primarily focused on developed countries. These must assess the relationship between psychosocial work costs and benefits and more conventional economic rewards and demographic transformations.

PART 1I

International Comparison of the JCQ Scales in Four Countries

Rationale

Trends in working conditions show increased risks to well-being arising from social and psychological characteristics of work in Europe and Japan in reported national statistics (Dhondt, 1994, 1998; Paoli, 1997; Shimomitzu & Levi, 1992). Available anecdotal reports in the United States in the 1990s also suggest growing problems of work-relate.d pressures. Comparative assessment of these psychosocial exposures between countries in the global economy could open up the area of "work quality" to international political and economic policy discussions. However, significant progress requires international comparisons with standardized instruments--an area in which progress has so far been slight. The international comparisons would also provide an important reference standard for many homogeneous population studies. These challenges motivate the present investigation.

Most international comparative analyses of work focus on economic outcomes. It is well known that there are national differences in the distribution of income, with greater disparities (i.e., higher standard deviations) occurring in the United States between high and low incomes than in Canada, Europe, or Japan. Furthermore, these disparities have increased in recent years in the United States.

Similar differences might be expected for psychosocial job characteristics. For example, Orth-Gomer (1979) reported similar associations between stress

and heart disease in U.S. and Swedish samples but found different retrospectively reported ratings for work and family stressors in the two countries (with work being less important in the United States). Kasl (1996) speculated that the toll of unemployment and the nature of work in general is less important for U.S. workers than for European workers. However, this has not been tested with direct empirical data on individual work situations in broadly representative populations.

Additionally, there has been much discussion of international differences in work organizational cultures, as when Japanese just-in-time assembly methods were introduced into the United States and Europe (Berggren, 1992). One empirical investigation of work organization and company policies across seven automobile manufacturers in four countries shows that "teamwork" is very differently understood due to national (Japanese, French, Italian, and German) differences, regional, and company culture (Frieling, Freiboth, Henniges, & Saager, 1997), although the relative magnitude of intercom- party versus intercountry sources was not assessed.

Some useful internationally comparative findings exist for scales very similar to the JCQ. Typical of these findings are U.S., Swedish, and Japanese studies which show that the ranking of occupations on decision latitude scales is very similar in the United States, Japan, and Sweden (see Appendix B). Unfortunately, the utility of many studies that are potentially similar is limited by lack of truly comparable scales in each study. In general, absolute scale scores are important for JCQ research because they allow comparative analysis of national and occupational differences and facilitate comparative formulation of some "job strain" definitions. Also, although validation studies for the JCQ scales have been published for some populations, the cross- national validity of the JCQ has not been systematically assessed. Neither has the comparative reliability and validity of the JCQ scales been tested separately for male and female workers. Hall (1994) argued for the importance of examining women's job characteristics separately. International comparisons of potential gender differences could be useful because female labor participation patterns vary significantly by COuntry.

From a psychometric perspective, the process of scale validation will often examine scale reliability statistics across multiple populations, with the presumption that the scale should perform in a similar manner across populations. Differences between the

334 KARASEK ET AL.

populations themselves are not the focus. However, from a sociological perspective, the population differences or similarities are of interest. We will also follow psychometric tradition and assess the consistency of the scale performance across populations. In the case of job characteristics, the most common sociological hypothesis would be that major national differences do exist in scale characteristics, because the JCQ reflects organizational characteristics of the major social institutional framework of modem societies--frameworks that have been demonstrated to differ between countries from many perspectives.

Objectives

The main objective of the study is to compare mean values, reliability, and validity of the JCQ scales across six studies conducted in four different countries. Complementary objectives are (a) to assess the extent of similarities and differences found under relatively different national contexts (the United States, Quebec-Canada, the Netherlands, and Japan); (b) to assess potential differences by gender; and (c) to compare these differences to published findings about scale differences due to occupation.

M e t h o d

Populations

The six populations studied come from the United States (2), Canada-Quebec (2), the Netherlands, and Japan (see Table 2). The number of participants in each study ranged from 580 to 6,053 for a total of 16,601 participants (38% women and 62% men). The participation rates ranged from 65% to 93% (see Table 2). The age boundaries span the full adult working life, age 20 to 65 (retirement age), but the Japanese sample's age span is from age 20 to 60. Part-time workers are included if work time is greater than 20 hr/week except for the two Canadian samples, which include only full-time workers (->35 hr/week).

Two populations include the full occupational spectrum-- managers and professionals, clericals, line workers in white- and blue-collar occupations, and services workers----based on random samples of their geographical location work- forces: (a) the U:S. QES samples from the 1970s (three separate national samples combined; see Karasek & Theorell, 1990, Appendix 1) and (b) the Quebec province stratified random (Sante-Quebec, 1994) sample from 1990. Three other samples are broadly inclusive of the full occupational spectrum from manager to line worker: the more recent U.S. New England Medical Center (NEMC) sample, the Dutch sample, and the Japanese sample. The U.S. NEMC data include representative samples from all status levels in 16 large U.S. workplaces in seven corporations (full samples of 12 middle-sized workplaces

Table 2 Population Definition for the Studies

Participant Study country Date N % instrument Population type

U.S. QES 1970s 4,319 73 QES/JCQ National random population sample M = 2841 F = 1478

U.S. NEMC 1994 6,053 71 Base JCQ (modified) a Broadly representative: 16 large employers, blue and white collar

M = 3676 F = 2377

1990 1,232 77 Provincial, stratified random M = 707 population sample F = 525

1994 2,666 73 e Regional, 8 companies M = 1364 65 a F = 1302

1994 1,751 86 Broadly representative: 34 compa- M = 1228 hies, blue and white collar (limited F = 523 job mobility/variety)

1993 580 93 2 companies, blue and white collar M = 472 F = 108

Canada-Quebec Full JCQ (minus) b

Canada-Quebec-white Full JCQ (minus) b collar only

Netherlands Full JCQ

Japan Full JCQ (minus) b

Note. QES = Quality of Employment Surveys; JCQ = Job Content Questionnaire; NEMC = New England Medical Center; M = male; F = female. a Some signlficandy different scales. b Minus some scales. ¢ Participation for the full population who completed the psychological demands and decision latitude scales. d Participation for the population who completed also the social support scales.


and random samples of 4 very large workplaces, representative of the U.S. workforce in many respects) but omit workers in smaller companies, a lesser fraction of the workforce. The Dutch sample includes white- and blue- collar workers from 34 companies, across a range of company sizes and industry branches. It deviates from random selection in that only workers with relatively fixed work locations and a limited number of tasks were selected to facilitate physical ergonomic job assessments (however, this does not exclude professionals such as teachers, nurses, and some managers). The Japanese sample (Haratani et al., 1997) includes white- and blue-collar workers from two companies with fairly technically sophisticated output (telecommunications and power utility). The sixth sample, the second Canadian database from Quebec, is conapesed of only white-collm workers employed in eight white-collar organiza- tions engaged in semipublic, public, or private service activities.

Instrument

Our intent was to use the full recommended JCQ scales (Karasek, 1985). However, some questions were missing in some of the older studies. Therefore, the "lowest common denominator of questions" was found to maximize comparative possibilities. For psychological demands, the five- question QES version is used, because the full nine-question version was only available in four of the studies (for all correlations, separate tabulations have been made for the full nine-questions version). For the physical demands scale, although five of the studies had the one-question version, only the relatively recent Dutch sample had the full recommended five-question scale, and thus, most tabulations are based on the one-question version. Only the Canadian sample had the five-question social support scale, but all of the other samples had a QES-based four-question version, which is used in the tables. No study had the full six-question job insecurity scale, thus a special two-question version of the scale, available in three studies (U.S. QES; U.S. NEMC; and the Netherlands), was computed only for this article. The original English version of the JCQ was used, or validated tramlafions in French (Brisson, Dion, et al.; Brisson, Larocque, et al., 1998; Larocque, Brisson, & Blanchet, in press), in Dutch (Houtman, 1995; Reuvers et al., 1998), and in Japanese (Kawakami & Fujigaki, 1996; Kawakawi et al., 1995).

Data Collection

The instrument was self-administered in most studies (Study 1, 2, 4, 5, 6) or administered during a face-to-face interview (Study 3). Self-administration was usually done at the workplace. Workers were allowed by their employers to fill in the questionnaire during regular working hours. Although data on all samples had previously been collected, tabulated, and in some eases published by participating researchers, this previous material did not always follow exact JCQ guidelines or was not presented in a comparable manner. Therefore, for the parpose of the present article, one of us (RK) requested participating researchers to retabulate data following defined JCQ guidelines (as modified above) to allow meaningful comparisons. Each participating

researcher sent the tabulated data to RK, who constructed the final tables.

Analyses

All analyses were conducted separately for women and men in the six populations. Mean values and standard deviations of each scale were calculated. We used an analysis of variance (ANOVA) to evaluate statistical differences between scale means across populations. Given that the samples are large, statistically significant differences between means across populations are expected even for small and potentially nonmeaningful differences. Therefore, the proportion of variance explained by the study site was presented to quantify the magnitude of the variations between means. Reliability of the scales was assessed by the internal consistency as measured by the Cronbach's alpha coefficients. Concurrent validity was assessed by correlations between scales and subscales and by correlation of scales with age a n d education. To assess variability in correlation coefficients, we computed range and mean correlations across the six samples. A simple scale was used to quantify variability on the basis of the range of correlation coefficients among the six populations: L = low variability (i.e., a difference between the highest and the lowest correlations coefficients < .20); M = moderate variability (i.e., a difference that varies from .20 to .35); and H = high variability (i.e., a difference > .35).

Some factorial validity analyses conducted in these populations axe summarized here, but no tables are presented because of space limitations. The exact methodologies of these analyses are not always consistent across populations. Separate factor analyses of decision latitude and psychological demands; then decision latitude, psychological demands, and social support; and finally, all scales together have been reported. Sometimes the two decision latitude subscales and the psychological demand scale have been tested as two factors, sometimes as three. All analyses were conducted using standard statistical analysis programs (Study 1, 5, 6; SAS Institute, 1990; and Study 2, 3, 4; Stata Statistical Software, 1997).

Comparison of Between-Occupation Variance and Occupational Rankings

Although consistent coding of occupations across all the databases was not possible for this article, the use of interoccupation differences in job characteristics present an important reference standard for assessing the relative importance of intercountry differences. Thus, the results for the one study database that does have an interoecupation ANOVA (U.S. QES) are included here, and an occupational ANOVA for scales similar to the JCQ in Sweden is also presented (Swedish scales from Johnson and Stewart, 1993; U.S. scales from the U.S. QES study in this article; Schwartz et al., 1988). The Swedish analysis is based on a similar scale for decision latitude; however, the support scale in Sweden is only a "possibility-of" assessment (mainly instrumental) and only for coworkers. Also, the Swedish psychological demand scale is assessed more subjectively and has fewer questions than the U.S. scales. The between-occupation variance reported is a percentage of the

336 KARASEK ET AL.

Cronbach alpha reliable scale variance (this represents between .60 and .75 of the total variance for the Swedish scales). An additional person-based demographic contribution to psychosocial scales between occupation is also reported in both U.S. and Swedish studies. This addition increases the variance somewhat, of course, not because of the job alone, but because of the interaction of person and job (e.g., the additional effect of age on the decision anthority of managers). This effect is roughly corrected for in the Swedish sample by assuming the same average contributions from demographics in Sweden as in the United States, averaged across scales.

The rankings of occupations on the JCQ decision latitude scale are compared for the seven occupations reported in Kawakami et al.'s (1995) two-company study, compared with the JCQ scale rankings of the same occupations in the U.S. and Swedish nationally random samples, The U.S. and Japanese studies use the JCQ scales reported in this study, whereas a similar, non-JCQ scale is used in Sweden (Johnson, Stewart, Friediund, Hall, & Theorell, 1990).

Resul ts

Means and Standard Deviations of JCQ Scales

The means are very similar across studies (see Table 3). Although the difference between means are statistically significant among the studies for all scales, the proportions of the variance of scale scores explained by study site are small, that is, generally less than 5% for men and women for the more reliable scales. Within men samples, the proportions explained ranged from 1% to 4% for decision latitude, psychological demands, supervisor support, and coworker support (averaging 2.5%). These proportions were, respectively, 6% and 9% for physical demands and job insecurity, where only three studies have data and there are reliability limitations for the scale forms included in the study. Within women samples, the proportions explained ranged from 3% to 10% for decision latitude, psychological demands, job insecurity, supervisor support, and coworker support (averaging 4.8%), whereas it was much greater (17%) for the single-question physical demand scale.

Although differences between samples are generally small, some specific tendencies can be observed. The decision latitude means are the highest in the two Quebec samples for both men and women. The decision latitude means and more specifically the decision authority means are lower in the U.S. QES (1970s) sample than in the U.S. NEMC (1994) sample for both men and women. The decision latitude means are the lowest in Japan and U.S. QES (1970) in both men and women, although for women, the Netherlands sample is as low as Japan's.

For psychological demands (five items), the U.S. NEMC women's population is higher than others, but otherwise there are no major deviations for men or women on this scale. The physical demands scale does not show consistently higher levels for men than for women. It shows a substantially lower level in the U.S. NEMC sample than in the U.S. QES sample. The job insecurity scale (based on the specially calculated comparable subscale across populations) shows a substantially higher level in the 1990s in the United States than it did in the 1970s.

One highly consistent finding observed in all studies is the gender differences in skill discretion and decision authority. Indeed, women have consistently lower scale means for both these subscales. These gender differences tend to be smaller in the Quebec stratified random population sample than in the other samples. The gender differences average about a quarter of the population's standard deviation.

The scale standard deviations are generally higher, as would be expected in the first three populations, which have broadly representative samples, than in others in which restrictions of populations do occur, for both men and women. In three representative samples, the standard deviations are very similar on all scales.

Internal Consistency of JCQ Scales

Internal consistency of the scales tend to be similar across populations and between men and women (see Table 4). The Cronbach's alpha coefficients are generally acceptable (overall average alpha for women is .73 and for men is .74). The highest and most acceptable values of the coefficients are found for the decision latitude, physical demands, supervisor support, and coworker support scales. However, the psychological demands scale, with five questions, is only borderline (Nunnaly & Bernstein, 1994). Some minor deviations occur for the skill discretion scale, which falls to borderline levels in the Dutch study for men and women and for the decision authority scale for men. However, taken as a single scale, the decision latitude scale has quite acceptable coefficients. The skill discretion scale has a low reliability for Japanese men but is acceptable for Japanese women. Decision latitude has good reliability in Japanese women but remains low for men. The job insecurity scale also has low coefficients in two out of the three populations for which data are available.


c ~

z

~ ~ I I ~

~ ~ 1 1 ~

~ , ~ . ~ ~ ~ ' ~ ~ . ~

0

II " ~ v

~ ~ ~.~

~ °

338 KARASEK ET AL.

Table 4

Cronbach 's Alpha Reliability Coefficients of the Job Content Questionnaire Scales Among Men and Women in Six Samples

Study sample

U.S. U,S. Canada- Canada- Scale QES NEMC Quebec Quebec-W Netherlands Japan M

Men

Skill discretion .75 .79 .79 .80 .67 .59 .732 Decision authority .69 .70 .71 .70 .61 .66 .678 Decision latitude .83 .84 .86 .86 .77 .68 .807 Psychological demands (9 items) - - - - .68 .75 .74 .72 .723 Psychological demands (5 items) .63 .71 .59 .67 .57 .61 .630 Physical demands NA NA - - - - .86 b - - .860 Job insecurity .60 a .74 - - - - .49 - - .610 Supervisor support .85 .80 - - .82 .83 .89 .838 Coworker support .80 .72 - - .72 .78 .74 .752

Women

Skill discretion .71 .75 .78 .79 .65 .80 .747 Decision authority .72 .64 .70 .63 .70 .68 .678 Decision latitude .80 .81 .85 .84 .77 .84 .818 Psychological demands (9 items) - - - - .72 .72 .69 .72 .713 Psychological demands (5 items) .62 .72 .63 .64 .51 .65 .628 Physical demands NA NA - - - - .79 b - - .790 Job insecurity .47 c .76 - - - - .52 - - .583 Supervisor support .83 .84 - - .83 .83 .87 .840 Coworker support .81 .75 - - .69 .82 .76 .766

Note. QES = Quality of Employment Surveys; NEMC = New England Medical Center; Quebec-W = white-collar only. Dashes indicate no data available. a 0.53 for the three-item scale of job insecurity, b For the five-item scale of physical demands, c 0.41 for the three-item scale of job insecurity.

Correlations Between JCQ Scales and Subscales

The correlations between JCQ scales and subscales represent 26 possible correlations (decision latitude correlations with its subscales are excluded). When analyzed separately for women and for men, these produce a matrix of 52 possible correlations (see Table 5). Across study populations, 33 correlations show low variability, 12 show moderate variability, and 7 show high variability. In a number of cases, the variability occurs from extreme values observed in the Japanese population (some of the idiosyncratic Japanese findings may be due to the very small women's sample; see below). If we exclude these extreme values, 39 correlations (75%) have low variability across populations.

Review of specific associations reveals the structure of the JCQ scale relations. Decision latitude is the additive combination of skill discretion and decision authority scales, which are consistently correlated at about .55 for both men and women. The supervisor

and coworker support scales, two subcomponents of the social support scale, are correlated at .40. Generally, there is little difference in the decision latitude scale's two subcomponents, skill discretion and decision authority, in correlations with the other scales. One exception is the Japanese men for whom the skill discretion coefficients are three times as large as the decision authority coefficients for both supervisor and coworker support. The decision latitude scale correlates strongly (given moderate scale reliability) with the two social support subcomponents; it correlates moderately with psychological demands and negatively with physical work and job insecurity.

The psychological demand scale displays a low and very variable positive correlation with decision latitude for both men and women (all the high- variability associations are from the psychological demand scale). However, in the above studies, only the U.S. women's sample of the 1970s and the Dutch sample shows the negative correlation, which is associated with an increased prevalence of the "job


strain" combination. The correlation is actually strongly positive in the small sample of Japanese women. Correlations of psychological demands with decision latitude are somewhat higher for men than for women. The findings for the nine-question version of the psychological demand scale show similar variability.

The psychological demands scale varies across populations in its correlation with physical demands. It is strongly positive in two out of three populations where it is available, but very low in the third one (the U.S. NEMC sample). Psychological demands has major associational variability with supervisor support: In the U.S. QES it is strongly negatively correlated, in Japan it is uncorrelated, and in other populations it has a low correlation. Physical demands show a consistent negative association with decision latitude, and correlations are stronger for men than for women. Job insecurity shows generally consistent associations with other scales, particularly a moderately strong negative correlation with the decision latitude, supervisor, and coworker support scales.

Correlations between the nine-item version of the psychological demands scale and other JCQ scales are shown in Table 6. The correlation between the five-item and the nine-item versions is high (.88). The variability of associations with other scales remains with the nine-item version (among the four populations for which it is available). The nine-item version has a more positive overall correlation with decision latitude.

The physical demands scale displays a low and variable positive correlation with job insecurity for both men and women and a low and negative correlation with supervisor and coworker support, except for the U.S. QES women's sample and the Japanese women's sample.

Correlations of JCQ Scales With Age and Education

Correlations of JCQ scales with age and education represent 32 possible correlations (see Table 7). Of these, 16 display low variability across populations, 10 show moderate variability, and 6 show high variability. Here particularly, extreme values are observed in the Japanese sample. After excluding them, 26 correlations (81%) have low variability across populations.

The correlations with age and education are, in general, lower than those observed between the JCQ scales and subscales themselves. There are no strong

age correlations with any scale. Education has strong negative correlations with decision latitude and physical demands, but low correlations otherwise. There is a moderately strong and consistent negative association between age and education (about - .20) in all of the samples (which is much stronger for the Japanese women). These correlations may reflect intergenerational differences in education levels (lower levels for older generation workers) and the broad age range (20 to 65 years) of included participants.

The Japanese sample has substantially different correlations of decision latitude scale with age and education than the other populations. There is no association between education and either skill discretion or decision authority for the Japanese men and negative correlations for the Japanese women, but a strong positive association exists in all other samples. For the Japanese women, a high positive association between age and psychological demands is also observed. Additionally, they have a high positive correlation between age and skill discretion and a weaker negative correlation between education and skill discretion (although this correlation reduces to zero when age is adjusted; Kawakami et al., 1995). Correlations of supervisor support and decision latitude with education are also very small for the Japanese sample but are strongly positive for the other samples.

Several consistent gender-related exceptions are observed. Demands are in general more consistently related to decision latitude for men than for women: Physical demands are more highly negatively correlated with decision latitude and education for men, and psychological demand associations with decision latitude are somewhat higher for men than for women. Physical demands are more highly negatively correlated with education for men than for women, and psychological demands are negatively correlated with age for men, except for the Dutch sample, but weaker for women, except for the Japanese sample and for one Canadian sample, in which correlations for men and women are nearly the same.

Factorial Validity of JCQ Scales

The U.S. population from the 1970s shows a clear factor pattern corresponding to the JCQ scales for men (which is not surprising because this was the JCQ defining base sample) and also for women. Also in the United States, the NEMC sample from the early 1990s shows a confirming pattern for most factors for

3 4 0 K A R A S E K E T A L .

x _ .

. . ~ k .

k .

@

I

I

I I

i l I

i

I I I

~ t 4t

i I I

I I I

I I i

I I

I I I

~CYZ

I

I

I

I

I

I


o . ~ . o

. ~ o .

. ~ . I I I

• . o

I I I

I I

I I

• o

I I I

I I I

I

[....

I

I

I

o o

(

I

I

J

I L I

I I I

I I

I I I

I

I

I"

I

~ ~ g I I 1 I I

I I I

~ o o Z

I I I I

I I I I

§ ' ~ - I I I I

~ . . . o

t I t I

I I I I

I I I

• r II

342 KARASEK ET AL.

Table 6 Scale Correlations for the Nine-Item Psychological Job Demand Scale With Other Job Content Questionnaire Scales Among Men (M) and Women (W) in Four Samples: Canada-Quebec (CAN-Q), Canada-Quebec White-Collar-Only (CAN-W), the Netherlands ( Neth ), and Japan

Scale/variable

Mean Men Women

M W CAN-Q CAN-W Neth Japan CAN-Q CAN-W Neth Japan

Psychological demands (5 items) .883 .885 .88** .91"* .84** .90** .90** .90** .82* .92*

Decision latitude .293 .220 .37** .38** .10 .32** .20** .34** -.05 .39* Physical demands .270 .280 - - - - -.27** - - - - - - .28** - - Job insecurity .050 .020 - - - - .05 - - - - - - .02 - - Supervisor support -.067 -.093 - - -.01 -.18"* -.01 - - - . I0"* -.21"* .03 Coworker support .020 -.013 - - .02 .02 .02 - - - .02 .08 .06 Age -.093 .048 -.10"* -.08** .08** -.27** -.05 -.10"* .01 .33** Education .103 .120 .20** .21"* .01 -.01 .13"* .22** .06 -.07

*p < .05. **p < .06

both men and women together, although the psychological demand scale splits into two factors in this analysis where a larger number of factors than scales has been requested. In Canada, both the Quebec stratified random population sample (Larocque et al., in press) and the white-collar population (Brisson, Dion, et al., 1998) show very clear two-factor confirmations for decision latitude and psychological demands. They also show a relatively clear three- factor confirmation for decision authority, skill discretion, and psychological demands for women, although some skill discretion questions load with decision authority. For men, some psychological demand scale extension questions have unclear loadings. A Canadian white-collar four- and five- factor solution, which includes both social support scales, also shows a quite clear factor pattern and thus offers a multiscale confirmation of the JCQ structure (Brisson, Dion, et al., 1998). The Dutch study extracts a larger number of factors than JCQ scales and finds basic confirmation for almost all of the scales. However, it also obtains separate factors for a second component of psychological demands and a separate factor for repetitive work (Reuvers et al., 1998). For Japan, the published articles (Kawakami & Fujigaki, 1996; Kawakami et al., 1995) generally confirm the JCQ scales in the factor pattern, but with exceptions for several skill discretion questions, which ambigu- ously loaded with the psychological demand scale, and for repetitive work (see below). Several psychological demand questions load on skill discretion for men, but there is better confirmation for women. The social support factors are clearly distinguished.

These factor analyses raise concerns about several specific questions. The most consistently troublesome question in most studies is the "repetitive work"

question. This question has a low and inconsistent loading on the decision latitude factor. However, this question generally has a quite nonnormal distribution (and indeed a Guttman scalar relation to skill-level questions in the U.S. and Swedish databases: that is, repetitive work is much more common at the lowest skill level; Karasek, 1976). It was not specially transformed in the factor analyses and has a low communality. This probably contributes to its low and inconsistent loading on the decision latitude factor. On the psychological demand scale, the "conflicting demands" questions had low loadings for the Dutch, the Japanese women, and the Canadians; and the "wait on others" question had a low loading for the Canadian samples.

Between-Occupation Variances: International Comparison

Appendix A shows that the between-occupation variances in U.S. random samples are 45%, 7%, 4%, 10%, and 26%, respectively, for decision latitude, psychological demands, social support, physical demands, and job insecurity scales for an average between-occupation variance of the psychosocial scales of 18%. Adding the effect of demographic covafiation (for age, education, self-employment, and marital status; Schwartz et al., 1988) on the JCQ scales (the second U.S. column) increases these figures to 50%, 15%, 5%, 27%, and 29%, respectively, for an average additional variance due to personal demographics of 7% (1% to 19%). This addition, in the U.S. case, is substantial for psychological demands but not large for the other psychosocial measures. For the Swedish study, only the total variance (including the contributions of age and


.~ &

t ~ ~ ~ . . -~ ,

.=o.8 I ~

#

I I I I I

. e ~ . S. ° ~ - o ,-1. o . . I I I

• . o

I I I I

I I

° * * ~ 1 ~1

I

I

, . ~ - I I~1

I I I

* * * *

I I I I I

* * ~ * *

I I I I I I

~ $ ~ I

I

I I I

~ ~ I I

I I

I

I

. * ¢~

I I

~ 3 ,~ o._,~ s s s ~

I I

I I I

• . °

} I

0 * *

• ~ . .

I I I

~ . . °

0

* * * *

I

~ * * * * *

I

~ . ~

* * * * ~ ~ * * * * *

. ~ m ~ m ~

i

* * * * *

o v . . . . . .

* * * ~ *

I

* * * * * 0 * * * * *

~ Z

# ~ * * .

. ~ ¢!. ¢!. ~. ~ . I I I I I I I

0

~ ~ . I I I' I" I" I" I"

o

" 0

I=

. .4

A

e ,,~

ee

d

I ~ ¢~ °

~ V

II

344 KARASEK ET AL.

experience) along with occupation is reported: The average Swedish total occupation, plus demographic, covariance is 34%. Correcting the Swedish figures to eliminate the demographic component (assuming similar contributions in the United States and Sweden), the Swedish occupation component would be 26%. This yields a roughly estimated U.S.- Swedish average of 22% between occupation variance across multiple scales that is due to occupation alone.

As a "quantitative scale" benchmark for between- occupation variances, income from job is reported for the United States. It shows a lower between- occupation variance (20%) than many of the psychosocial job characteristics.

Appendix B shows that the occupational rankings for six occupations on decision latitude from the United States, Sweden, and Japan are very similar: Only one occupation differs in the rankings, and that is from the more limited Japanese sample. The U.S. and Swedish samples have the same rankings.

After making an estimated correction for the demographic component in the Swedish tabulation to obtain the occupation-only-based variances, the interstudy/intercountry variances from Table 5 are a fraction of the interoccupational variances from Appendix A. The average is 22% of the variance due to occupation for the psychosocial scales included, compared with an average of 5% due to interstudy differences. These are, respectively, one tenth (decision latitude), one fifth (psychological demands), one half (physical demands and job insecurity), and two thirds for (social support), which has very low interoccupational variance.

PART III

Intercountry Comparison of JCQ Scales

General Summary

Across the six populations, the means and standard deviations of the JCQ scales tend to be similar. These six populations come from four different national work cultures from both Eastern and Western hemispheres, and in one country across a two-decade time span. Five of the six populations studied include wide occupational spectrums: blue collar, upper and lower status white collar, and include male and female workers who are separately tabulated. The results are consistent with the differences in working conditions in modern industrial countries are more similar across national boundaries than they are across occupational boundaries. The interstudy/

country variances are less than a quarter of the interoccupational variances from similar samples, and less than a tenth for the decision latitude scale. Three of the studies are based on random population samples or many-company samples and have almost unrestricted occupational distributions. The standard deviations are very consistent across these samples. Three other studies have sample limits that would reduce expected job variance to some degree. The Dutch sample included workers with "relatively fixed, less mobile" jobs (albeit including many professionals), the second Canadian sample was restricted to white-collar workers (mainly in service industries), and the Japanese sample included only two companies with fairly technical product outputs and a very small women's populations. The standard deviations of job characteristics in the latter samples are smaller than in the former samples, as would be predicted.

In addition to predictable and consistent means, the JCQ scales--with the exception of the highly variable psychological demand scales--have relatively consistent scale intercorrelations. This suggests that their "meanings" are similar across populations, in terms of interrelationships among working condition indicators.

There is a potentially important exception for Japan, however, where decision latitude has a much different correlation with age and education than in the other studies. The Japanese difference with respect to this important set of social status determi- nants is consistent with expectation of national differences that exists between Japan and the United States, Canada, and Europe. However, the small size of the Japanese two-company population is a reason for caution on a general interpretation of this result. Furthermore, two thirds of the Japanese variations are in the women's subpopulation, which is by far the smallest and most restricted sample in this article. With a sample size of 109, it is less than a tenth the size of the average sample included here and is gathered from two companies dominated by male employment, and thus is possibly a special group. A broader Japanese sample (now being constructed; Kawakami et ai., 1997) could test the hypothesis of greater homogeneity of job circumstances or different social status associations in Japan.

International Similarity of JCQ Scale Means and Standard Deviations

There is little support found in general for the initial presumption that major national differences in


psychosocial job characteristics exist between countries. There are some limitations in overall conclusions about the psychosocial characterization of work: Not all measures of working conditions are included; working hours are omitted. Nevertheless, our findings imply that although differences in social conditions of work certainly can exist, they are smaller in relative magnitude than might have been expected between the countries in our sample: small by comparison to occupational differences in the same countries (as has been studied in the United States, Sweden, and Japan). One possible conclusion is that at the macrosocial level, differences remain, reflecting the impact of national differences, but at the "reductionist" microlevel, job characteristics tend to be homogeneous.

It cannot be said that small differences between study means is evidence that the JCQ lacks ability to discriminate "objective" difference: The JCQ has a strong and consistent ability to discriminate detailed occupation. Indeed, some of its scales discriminate occupation better than income (see Appendix A; also Schwartz et al., 1988), certainly a "hard empirical" measure. It has clearly been shown in several samples (see Appendix B) that consistent interoccupational differences exist on the JCQ decision latitude scales, which are shown to be impoaant for the social organization of work from many other points of reference.

A full ANOVA (not included in this article), across countries and occupations, would need to properly test the hypothesis that variation in psychosocial working conditions in industrialized countries today is primarily "within occupation" and that smaller fraction remains between countries. It was unfortunately not possible to include direct comparison of occupational means by study as a result of different occupational coding schemes and incomplete coding in some databases (future publications will attempt to compare these results). However, the other published data included in this article allow a preliminary assessment. When we compare the 45% (U.S.) between-occupation variance for the most reliable scale, decision latitude, it can be seen that the between-study, "national" variations are less than one tenth of between-occupation variance for this scale, an order of magnitude. For psychological demand, the "national" variance is roughly a fifth of the occupational variance reported in the other studies. For job insecurity and physical demands, where our group of studies provides a less certain comparative base, the fraction is roughly one half, and for the social support scales, which discriminate

poorly by occupation, it is three quarters. The magnitude of person-based demographic differences, assessed in a covariariance test (which is one component of individually based variance in job assessments), can be clearly estimated (in the United States). They add about another third on average to the occupational variances (however, these contributions are as large as the occupational differences for psychological demands and job insecurity scales).

The findings, together with the occupational variance findings, suggest small "workplace cross- national" differences and much larger and consistent interoccupational differences. That is, the differences between managers and assembly workers are "relatively" similar in every country. The consistency of the standard deviations of the samples suggests that similar interoccupational differences exist in those samples as well (i.e., manager and line worker differences).

This study does not assess working hours, commut- ing time, or spouse's work time, all of which could contribute to the psychosocial costs of work, and which may differ between the United States, Canada, Europe, and Japan. Japan has had the most clearly expressed concern; "death by overwork" (karoshi; Uehata, 1991) is a publicly discussed issue, and working hours appear to be significantly in excess of reported statistics (Shimomitzu & Levi, 1993). Increased pressure to work hours in excess of the traditional 40 hours is also noted in the United States (Schor, 1991). Because official statistics can underreport working hours, it would be desirable to assess these in the future in the JCQ. Physical work hazards are also omitted (see Paoli, 1997, for a European comparison of relative magnitudes of these effects). Thus, the overall scope of working burdens is underestimated by looking only at this study's results. However, this study does assess the previously unassessed (internationally) and "soft," but increasingly important, social and psychological character of work. It finds that the dimensional structure of work appears similar across countries.

With these limitations in mind, a broad conclusion of this study seems justifiable: for workers' jobs, the global economy is apparently already here. The global economy in advanced industrial societies has created many consistent conditions of employment-- overcoming the centuries-old national bases for different work content and replacing it with an emerging set of psychosocial, "work quality" international norms for working behavior. Because many of these new measures are outside of the currently studied "quantitative" job attributes such as wages,

346 KARASEK ET AL.

hours, and benefits, it is even more vital that they become well understood as quickly as possible. In the early 20th century, major international borrowing of social relationship-determining work organization practices might have seemed unthinkable. However, over the last several decades, global workplace homogenization appears to have occurred---driven by global economic market pressures---to the extent that its consequences are now an eminently researchable topic.

Job Insecurity

The job insecurity scale shows a potentially important difference in means between studies, but the small number of studies makes it difficult to draw rigorous conclusions. The job insecurity scale (based on the specially calculated comparable subscale across populations) shows a substantially higher level in the 1990s in the United States than it did in the 1970s (by over 1.5 standard deviations--a major difference). This difference is consistent with public perceptions of high job insecurity in the 1990s, with up to 50% of the working population feeling insecure (Lohr, 1996). Also, the job insecurity scale level in the United States in the 1990s is substantially higher than in the Netherlands. The sample in the Netherlands only includes employed workers in stable companies (but this is true also of the U.S. NEMC sample in the 1990s). The finding stands in contrast to the conventional "quantitative" international comparisons, which usually rate the United States more favorable on job security because of the substantially lower national rates of unemployment. However, the U.S. "safety net" for the unemployed is generally less secure than that in the Netherlands. Fear of job loss can apparently be high even when the national unemployment rate is low--and indeed the U.S. press recorded such fears due to downsizing in the early 1990s in the United States because of perceptions of a weakening social contrast in the country to support worker entitlements (Gleckman, 1995).

Discussion of Correlations and Meaning

The consistent relationship between the scales is an important confirmation of the consistent meaning of the JCQ scales in different national work cultures and by gender (albeit with the notable exceptions mentioned earlier). The correlations between the

scales help define their meanings. For example, the relationship between decision latitude and social support, and each of their relationships to age and education, help define the meaning of both concepts: how much of decision authority is related to good relationship with supervisors, how much to age or to educational background. The correlation of psychological demands with the decision latitude subscales assesses the degree to which "responsibility" is related to authority, or how much psychological demands is related to skill level. The correlation between psychological demands and physical demands shows the inseparability in some work settings of these two types of demands in the work process. Strong connections in physiological effects also exist.

International Gender Comparisons

The primary gender differences found are systematically lower skill discretion and decision authority for women in all of the studies: about a quarter of the overall population standard deviation between male and female workers for each scale. In the United States, the difference for decision latitude is larger, and in the Quebec regional sample, the difference is smaller than in the other samples. In general, this difference reflects the enduring sex differences in both authority and opportunities to use and develop skills in the workplace across all the countries studied: a deficit of good psychosocial working conditions for women. Although this may not be as large as gender-based wage differences, it would be expected to contribute to reduced well-being for women in general, because stressor-strain relationships are often found to be generally similar for men and women in later adulthood (Frankenhaeuser & Johansson, 1986; Karasek, 1990).

Otherwise, the gender differences found in scale means in this four-country comparison are not large. Scales reliabilities and most scale correlations are also similar. However, both psychological and physical demands are more positively correlated to decision latitude, and sometimes with age and education for men than for women, implying clearer patterning of work role definitions by age and education for men. Finally, although the level of psychological demands are similar for men and women across studies, the psychological demand variable seems to have even more variable associations with other scales for women than it does for men.


Specific JCQ Scale Results

Psychological Demand Scale

The psychological demand scale has been criti- cized by researchers who have used the demand/ control framework for several deficiencies: sensitivity to bias related to health status (Kristensen, 1996) and its less consistent ability to predict outcomes such as heart disease in simultaneous association with decision latitude (Johnson, Stewart, Hall, Fredlund, & Theorell, 1996; Karasek & Theorell, 1990; Theorell et al., 1998).

Inconsistency of Meaning

The psychological demand scale provides some assurance of its reliability across diverse populations. Nevertheless, the psychological demand scale has the most variable of all correlations in the JCQ scale correlation matrix: In fact a majority of the variability for the entire set of JCQ scales occurs in conjunction with this scale. The variability in the association of the psychological demands scale across samples supports the interpretation that its meaning may differ by population group. This variability is congruent with the lack of consistent between-occupational discrimination of psychological demands in U.S. samples (Schwartz et al., 1988) and higher, but still inferior, discrimination in Swedish samples (Johnson & Stewart, 1993).

One possible reason for this instability of meaning could be that work's qualitative issues are simply historically lagging behind in awareness, and thus in contemporary language formation. By contrast, social discourse in modern industrial societies has a good language for quantitative occupational differences-- certainly for physical demands and wages, but now even for skill use and social authority. "Social bargains" around these quantitative issues have, over centuries, been incorporated into the different boundaries of existing occupational categories (butcher, baker, carpenter, etc.), yielding almost quantitative precision of meaning. However, negotiations about psychological pressures and social support have local, person-related validity--at least in the early stages of social discourse. Perhaps the "linguistic bargaining" (Habermas, 1984) in this area has not yet generated a sufficiently consistent vocabulary across diverse groups and situations to leave an imprint on institutional or occupational structures (which would, in turn, reinforce the consistency of meaning). Thus, the inability of a set of questions to assess psycho-

logical demand phenomena consistently could then be a reflection of the relatively poor precision of general current public discourse in these areas: not an "unimportant" problem but just a "new" problem. Aggregate occupation scoring system methods (Frese & Zapft, 1988; Karasek et al., 1988) or work group-based assessments (Vahtera et al., 1996) can reduce relative error variance in scales by averaging out reporting discrepancies in multiple persons' reports of a single work setting, but they cannot resupply variance lost as a result of imprecision in the original assessments.

Other job assessment methods might be developed that can overcome such "language deficits" by proactively assisting workers in generating reliable, consensus reports in this area. One example of such language development methods for job assessment is the "conducive dialogue" in which mapping of social relationships in production activity provides the basis for development of new languages about job alterna- fives (Karasek, 1990). Unfortunately, these methods are much more time consuming than survey methods.

Collinearity With Decision Latitude

The correlation between psychological demands and decision latitude is an important issue for testing the demand/control model. A positive correlation can indicate collinearity difficulties and be a source of weak tests of association because the high-strain combination (high demands, low control) would be an infrequent occurrence under such conditions. Because decision latitude is the substantially more statistically reliable of the two scales (particularly between occupations), their common variance with a dependent variable is likely to be attributed to decision latitude in hierarchical linear regression models. The correlation observed in the above studies is on average positive and stronger than previously observed (Karasek & Theorell, 1990). Only the U.S. women's sample of the 1970s and the Dutch sample show the negative correlations (associated with frequent high-demand/low-control conditions) that were reported as a major contributor to the work- related psychosocial stress risk for women (Karasek & Theorell, 1990). The positive correlation gives rise to another form of the same problem: The low-job- strain population can be small--in fact too small to test reliably. The correlation is also positive in several major study databases in heart disease research (.31; S. Stansfield, personal communication, June 28, 1998) in the Whitehall database (Marmot, Bosma,

348 KARASEK ET AL.

Hemingway, Brnnner, & Stansfeld, 1997) and the occupational linkage-based tests (Karasek et al., 1988), which could partly explain the relatively poorer showing of psychological demands in prediction of coronary heart disease in those studies (Bosma, Peter, Siegrist, & Marmot, 1998). This is a problem in other studies in which the demand/control model is tested with dichotomous or trichomous scales in which the high-demand/low-decision latitude combination is the exposure condition. The problem requires enlargement of the exposure criteria and thus diminished sensitivity.

Both the multicoUinearity and the disappearing exposure group problem are compounded when the associations with age are included. Table 7 shows a consistent negative association between age and psychological demands (and even a small positive correlation with age and,°decision latitude), meaning that the frequency of "high-strain" job holders is less in older age cohorts. Of course, older populations are often at higher risk (for coronary heart disease). This effect~s a three-way interaction in which psychological demands is the lowest reliability measure of the three. Similarly, the higher correlations of education with decision latitude and the still positive correlation of education and psychological demands mean there will be relatively few high-status, high-strain jobs. Thus, social class, education, or decision latitude could claim common variance between psychological demands and the dependent variable in hierarchical linear regressions.

Another problem, related both to psychological demands and to job strain in general, is underreport- ing in high-strain jobs. High-strain job incumbents, time pressured by practical situations almost by definition (often with multiple jobs, uncertain jobs, or temporary jobs), consistently report being too harried to participate in complicated research projects. Although population studies of job characteristics with very simple questionnaires might find only a moderate dropout of high-strain participants, the dropout rate increases differentially for high-stress groups with the more difficult study protocols (Costas & R. A. Karasek, personal communication, March 30, 1997). Such protocols are important for confirming illness etiology and for long-term follow-ups. These are the cornerstones of strong scientific confirmation of causal associations. The magnitude of this source of bias is not clear.

in social interaction (Hochschild, 1983; see, alterna- tively, Marshall, Barnett, & Sayer, 1997) and cognitive workloads. Emotional contact remains excluded in the present JCQ. However, there is evidence that the nine-question recommended JCQ scale version is an improvement for cognitive workloads, and the correlation is higher with skill discretion, which also assesses cognitive engagement. This version may be a more "white-collar sensitive" psychological demand formulation. The implication of the higher physical demands scale correlation of the shorter five-question version of the scale (only the Dutch sample allows this direct comparison) implies that the five-question version--perhaps explicitly the "work hard" and "work fast" questions--assesses physical as well as psychological loads. For most illness-causation hypotheses, this is not a substantive disadvantage: Most of the psychophysiological costs of work occur during the high levels of autonomous nervous system arousal (sympathetic or fight-or-flight response) to which physical demands certainly also contribute physiologically. This can, however, lead to interpretative ambiguities.

Job Insecurity Scale

In the broad area of job insecurity, the JCQ scale includes two types of information: (a) overall assessments of job insecurity and future career prospects and (b) specific data about layoff and work instability history. Although a statistically more homogeneous scale could be achieved by dropping some aspects of the JCQ measure (indeed the alpha scores of the two questions included for the specially calculated comparative version of the scale have higher reliability than the more comprehensive version; see Karasek & Theorell, 1990, Statistical Appendix), the robustness of the scale's interpretability would suffer. Unfortunately, both of the versions included in this study fall short of the full JCQ recommended scale version, which has not been reported here. Whether its greater inclusiveness leads to a higher overall alpha or a lower one cannot be judged at this time.

Implication for Broad Interpretability of Psychosocial JCQs:

The "Self-Report Bias" Issue

Insufficient Coverage

The psychological demand scale has been criti- cized for failure to adequately cover emotional labor

The most discussed aspect of questionnaire-based instruments for workplace research is the issue of objective validity of self-report questionnaires. The findings above shed additional light on the ongoing


debate about self-report bias in the use of questionnaire-based job assessment. The issue is whether self-reported variations on the scales correspond to the "objective reality" that other knowledgeable observers would also report, or whether there is something idiosyncratic about the observer himself or herself that make such reports unreliable as an assessment of the external environment--wbether this be due to demographic status (e.g., aging or gender), national culture, or enduring characteristics of the personality. As an estimate of these person- based effects, demographic status adds an average of 7% of the variance in the job scales (as covariates to occupation in U.S. national studies; Schwartz et al., 1988). In Table 3 we showed interstudy differences that could be presumed to reflect national differences in culture with respect to work explain an average of 5% of the scale variance. The suggestion of the present study, then, further narrows down the alternative, person-based explanations to the prima facie claim that the job scales measure the work environment. The question now remains whether the vast remaining portion of the variance in the scales exists due either to the job or to pure personality. There is no doubt much random error variance in the scales (assessed in Table 4), but this is not of consequence to this debate because it will not contribute to causal associations in either direction. Because the debate seems to have come down on the side of psychological characteristics attributed to persons as the source of such scale variance, we now reassess the self-report bias critique.

The critique of self-reports begins with cautions about methodological weaknesses of questionnaire reports. Two problems in particular limit causal interpretations in this area: (a) individual differences in perceptions of stressors exist and (b) common- method variance could inflate associations between job measures and self-reported well-being measures. This later problem varies by scale, however: potentially significant for psychological demand scales (where Hackman and Lawler, 1971; Frese and Zapft, 1988; and Kirmeyer and Dougherty, 1988, show worker-observer reliabilities of .32, .35, and .35, respectively), and probably insignificant for decision latitude scales (where Hackman and Lawler, 1971; Frese and Zapft, 1988; and Caiffm, 1983, show worker-observer reliabilities of .71, .54, and .65 to .75, respectively). This limitation may relect the magnitude of cognitive assessment required by the questionnaire format (Frese & Zapft, 1988), and also the degree of worker familiarity with or the

availability of vocabulary about precise category boundaries in each area of job experience.

However, the critique of potential bias in the scales (Brief, Burke, George, Robinson, & Webster, 1988; Ganster & Schaubroek, 1991) can also easily be overdone, leading to Type II statistical errors. The current form of this critical position claims that JCQ-like job assessment methodologies really tap a personality-based "negative affect factor" that accounts for both respondents' negative descriptions about their job and their negative emotional state.

Take, for example, a common but potentially misleading application of the "triviality trap." In this trap, the typical associations between the job characteristics and mood states, when assessed by questionnaires, are spuriously inflated by common questionnaire response behavior for both sets of variables, which then accounts for the observed associations. The recommended solution proposed in Brief et al. (1988) and Watson and Pennebaker (1989) and endorsed by Ganster and Schanbroeck (1991) is that these associations should be controlled for person-based measures of negative affect. This cure is worse than the problem itself. It potentially creates a problem by overcontrolling the associations. Claim- ing the negligible remaining relationship is clear evidence of a no-job/well-being association and is a "triviality turnaronnd."

Consider application of this trap with specific questions used in JCQ research: "My job requires working fast" and "There is little decision freedom on my job" in the job measure category; and, in the dependent variable category, psychological strain and mood reports such as "The future looks hopeless" and "I have sleeping difficulties" (of course, the alternative outcome for much of the above-cited research is quite objective coronary heart disease diagnoses).

The negative affect control solution is often based on other questionnaires that assess trait-based negative affect (often trait anxiety has been used) that have very much in common with mood state measurement instruments, which are the studies' primary dependent variable. In fact, they are often based on the very same questionnaire items with different formats and instructions (Spielberger, Gorsuch, Lushene, Vagg, & Jacobs, 1983; Spielberger & Sydeman, 1994, p. 296). For illustrative purposes, consider the trait measure of anger: I am [generally feel] quick tempered" and "feel infuriated when I do a good job and get a poor evaluation." Suppose the state anger questions: "I am furious [right now]" and "I am mad" measure the dependent variable. In the negative-affect adjustment

350 KARASEK ET AL.

solution, job/well-being and state associations are controlled with trait measures of this type. In this example, an association between "My job requires working fast" and "[Right now], I am furious," controlled by "[Generally] I feel mad" used as a measure of negative affect. Finding that most of the association disappears, researchers in this tradition have often concluded that most of the variance can be accounted for by negative affectivity--a personality trait--and not by job conditions. However useful such trait measures may be as approximations in other research contexts, for the close cause-and-effect simultaneous control of this research context, we would not endorse the validity of claims to "pure state" and "pure trait" measurement status, even when other research has confirmed state-trait differen- tiation, as in the case of Spielberger et al.'s scales. It is certainly not clear that they have been endorsed by their originators for such purposes. The trait measures can easily include state components that overcontrol the associations and cause "I3rpe II errors. The result is an underestimation of job effects, an understatement of "valid" disease state prevalence, and a false attribution of disease state to person characteristics.

Indeed, recent research (Dollard & Winefield, 1998) that explicitly test for the possibility of such overcontrol with negative affectivity, using job experience cohorts to test whether negative affectivity is itself associated with duration of exposure to stressful job circumstances, finds that it is. This type of finding even further undermines the interpretation that controlling for such measures should be used as a basis for rejecting associations between job and well-being: The presumably "pure" traits are not pure methodologically but are also impure in that short- term measurements can be partly the long-term results of environmental experience at work (longitudinal effects noted in Johnson et al., 1996; Karasek & Theorell, 1990). Certainly, many of these personality trait measures are at least as contaminated by common-method variance problems as the job characteristics they attempt to purify. They are also usually assessed in samples in which there is limited variance in environmental stressors, which also undermines associations with job measures. For good reasons, ethical committees ensure that laboratory stressors are far from the potency of truly difficult work situations, which also limits the likelihood of discovery of social environment effects here.

The negative affect critique of job and well-being associations, of course, implies its own causal hypothesis: Personality causes the negative emotional

states and reporting pattern. Given the systematic and objectively linked associations between job characteristics and occupation around the world in findings such as those reviewed above, and negative well- being associations in literature of many of the same countries cited in Part I, this critique necessarily takes on further implication that may represent a heavy, unexpected burden of proof for its advocates. The implication is that the industrial world's low-decision- latitude populations are generally afflicted with the consistent, social-class-based personality deficit of negative affectivity--which compels such people to systematically complain--but who have "objectively" nothing to complain about--in their similar jobs in the United States, in Japan, in Canada, in Sweden, in the Netherlands, and so on.

The critiques of self-report job analysis scales may be having the unfortunate effect of taking the heat off the world's business leaders to humanize work environments. Humane-sounding work-design buzzwords were commonly used in many "reengineer- ing" job change programs in the United States in the early 1990s. However, these changes appear to have had very negative impacts on broad groups of workers, particularly because of unexamined psychosocial consequences. The discussion of these topics, sometimes emanating from U.S. business schools, which in particular should know the dangers, seems instead to be preoccupied with the above critique of psychosocial research.

In conclusion, psychosocial researchers have admit- ted to limitations of these measures. It is now time for the detractors to reassess them in light of broad workplace reality that is presently evolving. Research- ers in this area need to do as much as they can to validate a social and political discussion, as well as scientific discussion of these topics, which are affecting so many people, now on a global scale.

References

Amick, B., Mangione, T., & Wu, V. (1998). Work organization and drinking: A test o f two psychosocial exposure models. Manuscript submitted for publication.

Berggren, C. (1992). Alternatives to lean production: Work organization in the Swedish auto industry. Ithaca, NY: ILR Press.

Bongers, P. M., de Winter, C. R., Kompier, M. A. J., & Hildebrandt, V. H. (1993): Psychosocial factors at work and musculoskeletal disease. Scandinavian Journal of Work, Environment and Health, 19, 297-312.

Bosma, H., Peter, R., Siegrist, J., & Marmot, M. (1998). Two alternative job stress models and the risk of coronary heart disease. American Journal of Public Health, 88, 68-74.


Bourbonnals, R., Brisson, C., Moisan, J., & Vezina, M. (1996). Job sla'ain and psychological distress in white- collar workers. Scandinavian Journal of Work Environ- ment and Health, 22, 139-145.

Brandt, L. P., & Neilsen, C. V. (1992). Job stress and adverse outcome of pregnancy: A causal link or recall bias? American Journal of Epidemiology, 135, 302-311.

Brief, A., Burke, M., George, J., Robinson, B., & Webster, J. (1988). Should negative affectivity remain an unmea- sured variable in the study of job stress? Journal of Applied Psychology, 73, 207-214.

Bfisson, C., Dion, G., Blanchet, C., Moisan, J., Guinont, C., & Vezina, M. (1998). Reliability and validity of the French version of the 16-item psychological demand and decision latitude scale of the Karasek Job Content Questionnaire in white-collar workers. Manuscript submitted for publication.

Brisson, C., Laroque, B., Bourbonnais, R., Dion, G., Moisan, J., Blanchet, C., Guimont, C., Vinet, A., Vezina, M., Dagenais, G., & Laflamme, N. (1998, August). The Job Content Questionnaire (JCQ) in the Canadian samples. Paper presented at the First International Conference on Psychosocial Factors at Work, Copenha- gen, Denmark.

Cannon, W. B. (1914). The emergency function of the adrenal medulla in pain and the major emotions. American Journal of Physiology, 33, 356-372.

de Jonge, J., & Kompier, M. A. J. (1997). A critical examination of the demand-control-support model from a work psychological perspective. International Journal of Stress Management, 4, 235-258.

Dhondt, S. (1994). Monitoring occupational health and safety in Europe: Time constraints and its implications. Leiden, The Netherlands: TNO-Gezondheidsonderzoek (Netherlands Institute for Health and Prevention).

Dhondt, S. (1998). Time constraints and autonomy at work in the European Union. Dublin, Republic of Ireland: European Foundation for the Improvement of Living and Working Conditions.

Dollard, M. E, & Winefield, A. H. (1998). A test of the demand-control/support model of work stress in correc- tional officers. Journal of Occupational Health Psychol- ogy, 3, 243-264.

Fenster, L., Schaefer, C., Mathur, A., Hiatt, R. A., Pieper, C., & Hubbard, A. E. (1995). Psychologic stress in the workplace and spontaneous abortion. American Journal of Epidemiology, 142, 117-183.

Fox, M. L., Dwyer, D. J., & Ganster, D. G. (1993). Effects of stressful job demands and control on physiological and attitudinal outcomes in a hospital setting. Academy of Management Journal, 36, 289-318.

Frankenhaeuser, M., & Johansson, G. (1986). Stress at work: Psychobiological and psychosocial aspects. International Review of Applied Psychology, 35, 287-299.

Frese, M., & Zapft, D. (1988). Methodological issues in the study of work stress: Objective vs. subjective measurement of work stress and the question of longitudinal studies. In C. L. Cooper & R. Payne (Eds.), Causes, coping and consequences of stress at work (pp. 375-411). New York: Wiley.

Frieling, E., Freiboth, M., Henniges, D., & Saager, C. (1997). Effects of team work on the working conditions of short cycled track work: A case study from the European

automobile industry. International Journal of Industrial Ergonomics, 20, 371-388.

Ganster, D. G., & Schaubroeck, J. (1991). Work stress and employee health. Journal of Management, 17, 235-271.

Gleckman, H. (1995, November 20). Rewriting the social contract. Business Week, p. 70.

Griffen, R. W. (1983). Objective and social sources of information in task redesign: A field experiment. Administrative Science Quarterly, 28, 184-200.

Habermas, J. (1984). Theory of communicative active action (Vol. 1). Cambridge, England: Polity Press.

Hackman, J. R., & Lawler, E. E. (1971). Employee reactions to job characteristics. Journal of Applied Psychology, 55, 259-286.

Hall, E. M. (1994). Women, work, and health: Employment as a risk factor for coronary heart disease. Journal of Preventive Cardiology, 4, 365-384.

Haratani, T., Kawakami, N., Kobayashi, E, Ishizaki, M., Hayashi, T., Masumoto, T., Hiro, H., Aizawa, Y., Hashimoto, S., & Araki, S. (1997, August). Work stress, depression, accidents and sick absence: Preliminary findings from the Japan Work Stress Cohort Study. Paper presented at the 15th Asian Conference on Occupational Health, Kuala Lumpur, Malaysia.

Hochschild, A. (1983). The managed heart: Commercializa- tion of human feeling. Berkeley: University of California Press.

Houtman, I. (1995, May). Reliability and validity of the Dutch version of the Karasek Job Content Questionnaire. Paper presented at the NIOSH/APA Conference on Stress, Work and Health, Washington, DC.

Johansson, S. (1971). Om LevnadsnivdandersOkningen [On the national survey of level of living]. Stockholm: Allmamma F6flaget.

Johnson, J. V. (1986). The impact of workplace social support, job demands and work control upon cardiovascular disease in Sweden. Unpublished doctoral dissertation, Johns Hopkins University.

Johnson, J. V. (1989). Collective control: Strategies for survival in the workplace. International Journal of Health Services, 19, 469--480.

Johnson, J. V., & Hall, E. M. (1988). Job strain, work place social support, and cardiovascular disease: A cross- sectional study of a random sample of the Swedish working population. American Journal of Public Health, 78, 1336-1342.

Johnson, J. V., & Johansson, G. (Eds.). (1991). The psychosocial work environment: Work organization, democratization and health. In Essays in memory of Bertil Gardell. Amityville, NY: Baywood.

Johnson, J. V., & Stewart, W. (1993). Measuring work organization exposure over the life course with a job-exposure matrix. Scandinavian Journal of Work, Environment and Health, 19, 21-28.

Johnson, J. V., Stewart, W., Friedlund, P., Hall, E. M., & Theorell, T. (1990). Psychosocial job exposure matrix: An occupationally aggregated attribution system for work environment exposure characteristics (Stress Re- search Report No. 221). Stockholm: National Institute for Psychosocial Factors and Health.

Johnson, J. V., Stewart, W., Hall, E. M., Fredlund, P., & Theorell, T. (1996). Long-term psychosocial work environment and cardiovascular mortality among Swed-

352 KARASEK ET AL.

ish men. American Journal of Public Health, 86, 324-331.

Karasek, R. A. (1976). The impact of the work environment on life outside the job. Unpublished doctoral dissertation, Massachusetts Institute of Technology.

Karasek, R. A. (1979). Job demands, job decision latitude, and mental strain: Implications for job redesign. Adminis- trative Science Quarterly, 24, 285-308.

Karasek, R. A. (1985). Job Content Questionnaire and user's guide. Lowell: University of Massachusetts Lowell, Department of Work Environment.

Karasek, R. A. (1990). Lower health risk with increased job control among white-collar workers. Journal of Occupa- tional Behaviour, 11, 171-185.

Karasek, R. A. (1991). The political implications of psychosocial work redesign: A model of the psychosocial class structure. In J. V. Johnson & G. Johansson (Eds.), The psychosocial work environment: Work organization, democratization, and health (pp. 163-190). Amityville, NY: Baywood.

Karasek, R. A. (1997). Demand/control model: A social, emotional, and physiological approach to stress risk and active behavior development. In ILO encyclopedia of occupational health and safety. Geneva: ILO.

Karasek, R. A. (1998). The new work organization, conducive production, and work quality policy. In M. Marmot (Ed.), Labor market changes and job insecurity: A challenge for social weUfare and health promotion (pp. 78-105). Copenhagen: WHO/Europe.

Karasek, R. A., Baker, D., Marxer, F., Ahlbom, A., & Theorell, T. (1981). Job decision latitude, job demands, and cardiovascular disease: A prospective study of Swedish men. American Journal of Public Health, 71, 694-705.

Karasek, R. A., Schwartz, J., & Pieper, C. (1983). Validation of a survey instrument for job-related cardiovascular illness. New York: Columbia University, Department of Industrial Engineering and Operations Research.

Karasek, R. A., & TheoreU, T. (1990). Healthy work: Stress, productivity, and the reconstruction of working life. New York: Basic Books.

Karasek, R. A., Theorell, T., Schwartz, J., Schnall, P., Pieper, C., & Michela, J. (1988). Job characteristics in relation to the prevalence of myocardial infarction in the U.S. HES and HANES. American Journal of Public Health, 78, 910-918.

Karasek, R. A., Triantis, K., & Chaudhry, S. (1982). Co-worker and supervisor support as moderators of associations between task characteristics and mental strain. Journal of Occupational Behavior, 3, 147-160.

Kasl, S. V. (1989). An epidemiological perspective on the role of job control in health. In S. Sauter, J. J. Hun~ll, & C. L. Cooper (Eds.), Job control and worker health (pp. 161-189). New York: Wiley.

Kasl, S. V. (1996). The influence of the work environment on cardiovascular health: A historical, conceptual, and methodological perspective. Journal of Occupational Health Psychology, 1, 42-56.

Kawakami, N., & Fujigaki, Y. (1996). Reliability and validity of the Japanese version of Job Content Question- naire: Replication and extension in computer company employees. Industrial Health, 34, 295-306.

Kawakami, N., Kobayashi, E, Araki, S., Haratani, T., & Furui, H. (1995). Assessment of job stress dimensions based on the job demands-control model of employees of

telecommunication and electric power companies in Japan: Reliability and validity of the Japanese version of Job Content Questionnaire. International Journal of Behavioral Medicine, 2, 358-375.

Kawakami, N., Tanigawa, T., Araki, S., Nakata, A., Sakurai, S., Yokoyama, K., & Morita, Y. 0997). Effects of job strain on helper-inducer (CD4 +CD29 + ) and suppressor- inducer (CD4+CIMSRA+) T cells in Japanese blue- collar workers. Psychotherapy and Psychosomatics, 66, 192-198.

Kirmeyer, S. L., & Dougherty, T. W. (1988). Workload, tension, and coping: Moderating effects of supervision support. Personnel Psychology, 41, 125-139.

Kristensen, T. S. (1989). Cardiovascular diseases and the work environment: A critical review of the epidemiologic literature on chemical factors. Scandinavian Journal of Work, Environment and Health, 15, 245-264.

Kristensen, T. S. (1995). The demand-control-support model: Methodological challenges for future research. Stress Medicine, 11, 17-26.

Kristensen, T. S. (1996). Job stress and cardiovascular disease: A theoretical critical review. Journal of Occupa- tional Health Psychology, 1, 246-260.

Landsbergis, P. A., Schnall, P. L., Warren, K., Schwartz, J. E., & Picketing, T. (1994). Association between ambulatory blood pressure and alternative formulations of job strain. Scandanavian Journal of Work Environment and Health, 20, 349-363.

Larocque, B., Brisson, C., & Blancbet, C. (in press). Coherence interne, validite factorielle et validite discrimi- nante de la traduction francaise des EcheUes de demande psychologique et de latitude decisionnelle du "Job Content Questionnaire" de Karasek [Internal consistency, factorial validity, and discriminant validity of the French version of the Karasek Job Content Questionnaire]. Revue d'Epidemiologie et de Sante Publique.

Lohr, S. (1996, December 29). Though upbeat on the economy, people still fear for their jobs. New York I~mes, p. 26.

Marmot, M. G., Bosma, H., Hemingway, H., Brunner, E., & Stansfeld, S. (1997). Contribution of job control and other risk factors to social variations in coronary heart disease incidence. Lancet, 350, 235-239.

Marmot, M. G., & Tbeorell, T. (1988). Social class and cardiovascular disease: The contribution of work. Interna- tional Journal of Health Services, 18, 659--674.

Marshall, N., Barnett, R., & Sayer, A. (1997). The changing workforce, job stress, and psychological distress. Journal of Occupational Health Psychology, 2, 99-107.

Meshkati, N., Hancock, P., & Rahami, M. (1990). Tech- niques in mental workload assessment. In Ergonomic task analysis (pp. 65-93). London: Taylor & Francis.

Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric theory (3rd ed.). New York: McGraw-Hill.

Orth-Gomer, K. (1979). Iscbemic heart disease and psychological stress in Stockholm and New York. Journal of Psychosomatic Research, 23, 165-173.

Paoli, E (1997). Second European survey on working conditions, 1996. Dublin, Republic of Ireland: European Foundation for the Improvement of Living and Working Conditions.

Peters, M., Godaert, G. C. R., Ballieux, R. E., Brosschot, I. E, Sweep, C. G. J., Swinkels, L. M. J. W., Vliet, M., Heijman, C. J. (1998), Immune responses to experimental


stress: Effects of mental effort and uncontrollability. Manuscript submitted for publication.

Reuvers, M., de Jonge, J., Houtman, I., Bongers, P., & Kompier, M. (1998). Linear or nonlinear effects of job characteristics on subjective outcomes and sickness absence? Manuscript submitted for publication.

Sante-Quebec. (1994). Et votre ca va? Rapport d'enquire que coise sur la sante cardiovasculaire [What about your heart? Report of the Quebec Cardiovascular Health Survey]. Vieneuil, France: Govemement du Quebec.

SAS Institute. (1990). SAS/Stat. user's guide (Version 6, 4th ed). Cary, NC: Author.

Sauter, S. L., & Murphy, L. M. (1995). The changing face of work and stress. In S. L. Sauter & L. M. Murphy (Eds.), Organizational risk factors for job stress (pp. 1--6). Washington, DC: American Psychological Association.

Schnall, P. L., & Landsbergis, P. A. (1994). Job strain and cardiovascular disease. Annual Review of Public Health, 15, 381-411.

Schor, J. B. (1991). The overworked American: The unexpected decline of leisure. New York: Basic Books.

Schwartz, J., Pieper, C., & Karasek, R. A. (1988). A procedure for linking job characteristics to health surveys. American Journal of Public Health, 78, 904-909.

Seyle, H. (1976). The stress of life. New York: McGraw-Hill. (Original work published 1936)

Shimomitzu, T., & Levi, L. (1992). Recent working life changes in Japan. European Journal of Public Health, 2, 76-86.

Siegrist, J. A. (1996). Adverse health effects of high-effort/ low-reward conditions. Journal of Occupational Health Psychology, 1, 27-41.

Spielberger, C., Gorsuch, R., Lushene, R., Vagg, P., &

Jacobs, G. (1983). Manual for the state-trait anxiety inventory. Palo Alto, CA: Consulting Psychologists Press.

Spielberger, D., & Sydeman, S. (1994). State-trait anxiety inventory and state--trait anger expression inventory. In M. Maruish (Ed.), The use of psychological tests for treatment planning and outcome assessment (pp. 292- 321). Hillsdale, NJ: Erlbaum.

Stata statistical software: Release 5.0 [computer software]. (1997). College Station, TX: Stata

Theorell, T., & Karasek, R. A. (1996). Current issues relating to psychosocial job strain and cardiovascular disease research. Journal of Occupational Health Psychology, 1, 9-26.

Theorell, T., Tsutsumi, A., Hallquist, J., Reuterwall, C., Hogstedt, J., Fredlund, P., Emlund, N., & Johnson, J. V. (1998). Decision latitude, job strain, and myocardial infarction: A study of working men in Stockholm. American Journal of Public Health, 88, 382-388.

Uehata, T. (1991). Long working hours and occupational stress-related cardiovascular attacks among middle-aged workers in Japan. Journal of Human Ergology, 20, 147-153.

Vahtera, J., Pentti, J., & Uutela, A. (1996). The effect of objective job demands on registered sickness absence spells: Do personal, social and job-related resources act as moderators? Work & Stress, 10, 286-308.

Warren, (1998). Psychosocial, biomechanical, and organizational contributors to development of musculoskeletal disorders in the Dutch Monitor Survey. Unpublished doctoral dissertation, University of Massachusetts Low- ell, Department of Work Environment.

Watson, D., & Pennebaker, J. (1989). Health complaints, stress, and disease: Exploring the central role of negative affectivity. Psychological Review, 96, 234-254.

(Appendix follows on next page)

354 KARASEK ET AL.

A p p e n d i x A

B e t w e e n - O c c u p a t i o n R e l i a b l e V a r i a n c e (RV) in t h e J o b C o n t e n t Q u e s t i o n n a i r e

S c a l e s f o r M a l e W o r k e r s

United States Sweden

% RV with % RV with Measure % RV demographic covariance demographic covariance

Decision latitude 45 50 64 Psychological demands 7 15 29 Social support 4 5 - - Physical demands 26 27 29 Job insecurity 10 29 - - Income from job 20 35 - -

Note. Dashes indicate no data available.

A p p e n d i x B

O c c u p a t i o n a l R a n k i n g s o n D e c i s i o n L a t i t u d e in t he U n i t e d S ta tes , S w e d e n , a n d J a p a n

Men

United States Sweden Japan

1. Managers 1. Managers 1. Managers 2. Clerical super- 2. Clerical super- 2. Clerical super-

visors visors visors 3. Linemen b 3. Linemen ~ 3. Foremen

4. Foremen 4. c 4. Clericals 5. Clericals 5. Clericals 5. Linemen b

6. Operators 6. Operators 6. Operators

Women

United States Sweden Japan

1. Managers 1. Managers a 1. Managers 2. Clerical super- 2. Clerical super- 2. Nurse b

visors visors a 3. Nurse b 3. Nurse b 3. Clerical super-

visors 4. Clericals 4. Clericals 4. Clericals 5. Telephone 5. Telephone 5. Telephone

operators operators operators

a Women general managers are a very small category with very large variation, b Tie ranking in most-experience categories with clerical managers, c Foremen cannot be separately ranked in the Swedish system.

SPECIAL SECTION: JOB CONTENT QUESTIONNAIRE

Append ix C

Job Conten t Quest ionnaire (JCQ) R e c o m m e n d e d Version (Version 1.11, unchanged since 1985; Abbrevia ted Wordings)

355

la. Skill Discretion "learn new things"; "repetitive work"; "requires creativity"; "high skill level"; "variety"; "develop own abilities"

lb. Decision Authority "allows own decisions"; "little decision freedom"; "a lot of say"

lc. Skill Utilization "education required by job" (also requires education)

1. Decision Latitude = a weighted sum of la and lb

2. Psychological Job Demands "work fast"; "work hard"; "no excessive work"; "enough time"; "conflicting demands"; "intense concentration"#;

"tasks interrupted"#; "hectic job"#; "wait on others"# 3a. Supervisor Social Support

"supervisor concerned"; "supervisor pays attention"; "hostile supervisor"#; "helpful supervisor"; "supervisor good organizer"

3b. Coworker Social Support "coworkers competent"; "coworkers interested in me"; "hostile coworkers"#; "friendly coworkers"; "coworkers work

together"#; "coworkers helpful" 4. Physical Job Demands

"much physical effort": "lift heavy loads"#; "rapid physical activity"#; "awkward body position"#; "awkward ann positions"#

5. Job Insecurity "steady work"; "job security"; "recent layofl~"#; "future layoff"; "career possibilities"#; "skills valuable"#

Note. The symbol # indicates questions were added in 1985 to create the recommended version. For scale scoring, see the Job Content Questionnaire and User's Guide (Karasek, 1985). The macrodecision scales are not included here because of lack of broad use. Additional recommended "global economy" questions (5) were added in 1995 (September 1995, revision 1.5), but these are still informal recommendations, because pilot data have not been reviewed.

Received March 12, 1998 Revision received June 3, 1998

Accepted July 16, 1998 •

Acknowledgment of Reviewers for Volume 3

The editor gratefully acknowledges the assistance of the fo l lowing persons who

served as ad hoc reviewers of manuscr ipts submit ted to the Journal of Occupational Health Psychology in lt598,

Ryan Amacher Mark Eakin Kenneth Price Charles Bond Cathy Henney Jody Reinhartz Nancy Burns Scott Keller Arie Shirom Carolyn Cason Jeff McGee Kai Spratt Thomas Dougherty Gary McMahan

JCQ SCALE

Education

coronary heart

unpublished

england medical

decision latitude

job content

recommended

psychological

occupational