Friday, November 30, 2012

Lama juga tidak menulis, terlalu fokus dengan ulasan-ulasan berita yang diterbitkan melalui media alternatif. Isu sekarang ialah bila PRU13 akan berlangsung. Bukan setakat rakyat yang menunggu malahan ia menjadi makin penting buat ramalan pelaburan jangka masa panjang berkaitan dasar pengurusan negara. Jadi tidak menjadi satu yang menghairankan jika terdapat penangguhan kemasukan pelaburan ke dalam negara berpunca daripada ketidaktentuan kesinambungan dasar-dasar pengurusan negara. Satu segi, Pakatan Rakyat (PR) menawarkan urus tadbir negara yang diistilahkan sebagai "good governor" atau urus tadbir yang baik dengan usaha-usaha yang jelas untuk mengekang perilaku rasuah melalui kaedah pengisytiharkan harta para pemimpinnya, kaedah tender terbuka dan usaha-usaha untuk mengekang manipulasi kroni dalam projek-projek kerajaan. Penstrukturan semula subsidi dengan mengekang monopoli satu pihak dan memberi fokus bantuan untuk mengurangkan beban rakyat melalui janji penurunan harga minyak, kajian semula subsidi kepada pembekal tenaga dengan menyemak semula kandungan perjanjian antara kerajaan dengan pembekal. Meningkatkan kadar royalti minyak kepada 20% untuk negeri-negeri yang terlibat dengan pengeluaran minyak negara. Penghapusan pinjaman PTPTN, penurunan harga kereta, pemberian tambahan sebanyak RM500 kepada golongan pendidik dan cabaran kepada rakyat untuk menolak PR sekiranya gagal memenuhi aspirasi rakyat pada PRU14. Dalam sekian kalinya, enakmen undang-undang Islam negeri Kelantan akan dikuatkuasakan dan tidak mustahil perkara ini akan menjadi punca pergolakan antara PAS dengan DAP. Sehubungan itu, satu mekanisme yang jelas perlu ada dalam pemikiran para pemimpin PAS untuk menghadapi DAP yang telah jelas menentang pelaksanaan undang-undang Islam. Selain bergantung kepada hidayah Allah ke atas pemimpin-pemimpin DAP, PAS perlu giat memberi kefahaman kepada DAP tentang undang-undang Islam. Selain PAS perlu memiliki kekuatan suara melalui kerusi-kerusi parlimen yang dimenangi dengan bilangan yang lebih besar berbanding sedia ada. Isu penguatkuasaan kaedah-kaedah Islam ke atas salon-salon yang dimiliki oleh bukan Islam di Kelantan perlu diberi perhatian yang serius dan dihadapi dengan penuh hikmah sebagai ujian dalam melaksanakan hukum Allah. Kegagalan pemimpin-pemimpin PAS Kelantan menghadapi kes ini dengan penuh hemah akan memberi bayangan kaedah menyelesaikan masalah seandainya peraturan Islam dilaksanakan kelak. 22 tahun rasanya sudah cukup tarbiyah diberikan kepada masyarakat Kelantan tentang Islam sebagai "the way of life". Dengan wujudnya Dewan Himpunan Penyokong-Penyokong Pas memberi gambaran pelaksanaan motto membangun bersama Islam di Kelantan mampu diterima oleh masyarakat Kelantan. Kesanggupan sami-sami buddha di Kelantan menyumbang kewangan bagi projek lebuh raya rakyat memberi gambaran penerimaan Islam sebagai medium penyatuan harmoni antara rakyat berlainan agama. Dasar kerjasama antara PAS, DAP dan PKR sebenarnya banyak berkisar untuk menjatuhkan BN dan merampas Putrajaya bukan kerjasama untuk membina negara berasaskan Islam, cumanya DAP menerima akhlak-akhlak baik yang dianjurkan oleh Islam dalam soal urus tadbir negara. Jika ditakdirkan Pakatan Rakyat diberi kuasa oleh rakyat, maka ideologi asal perjuangan PAS dan DAP akan menjadi satu persoalan. DAP menekankan soal Malaysia untuk rakyat Malaysia tanpa mengira bangsa, hal ini telah terkandung dalam Islam dengan Islam untuk semua tanpa mengira bangsa. Malah lebih menyeluruh.  Persoalannya, setakat mana penerimaan peraturan Islam oleh DAP jika dilaksanakan. Dalam hal ini PKR perlu memainkan peranan sebagai moderator untuk menjadikan suasana lebih positif. Pohon dimudahkan oleh Allah.

Saturday, February 11, 2012

How Can the Five Job Characteristics Help Motivate Teachers?

How Can the Five Job Characteristics Help Motivate Teachers?thumbnail
The Five Core Job Characteristics can motivate teachers.
Just like anybody else, teachers want to feel respected by their employers and to believe that the work they do makes a difference. In any university's teacher training program, young, hopeful educators will discuss their vision of making a difference with students as one of their main motivating factors. Teachers who have been on the job for years also are motivated by these characteristics even if they do not use these exact terms when they discuss their job satisfaction.

By exploring the comments teachers make, the job characteristics that motivate them can become clear to administrators and others concerned with teacher motivation.
Related Searches:
  1. What are the Five Core Job Characteristics?

    • In 1976, motivational researchers J. Richard Hackman and Greg R Oldham defined five characteristics in jobs that people felt highly motivated performing had in common. If defined in terms of teaching, they look like this:
      Skill Variety: This means there is a perceived variety and complexity of skills and talents required to perform the job.
      Task Identity: Which means the teacher perceives her work's place in the district's larger plan.
      Task Significance : The job is perceived to affect the well-being of others.
      Autonomy: The teacher perceives an opportunity to employ personal initiative in order to do the work.
      Feedback From the Job: The teacher feels that he gets accurate information about his job performance.

    Skill Variety

    • A teacher motivated by Skill Variety may say something like this: "People don't understand what I do. They think I just grade homework, give tests, and enter grades."
      This disgruntled teacher will go on to explain how he creates assignments and lesson plans designed to teach students a skill while meeting state learning standards for students at his grade level.
      This teacher perceives and values his position's skill variety. He keeps striving even though he thinks others do not notice his position's complexity. His perception of his administration's seeming lack of awareness can wear on his motivation over time. This can become critical when an administration makes hiring and firing decisions based on salary level and number of years of service alone.
      To combat this, a school district administration and a community that applauds its teachers for all their "hard work" can make this teacher feel more valued. Recognizing individual teachers for their successes and for the times they do more than grade homework and give tests helps teachers stay motivated based on the perception of skill variety in their position.

    Task Identity

    • A teacher motivated by task identity may make comments like this: "I work in a good school district. Our students graduate with a solid education."
      This teacher sees her place in the big picture and will be motivated by her contribution to the district's larger goals. When each grade level clearly feeds into the next and teachers do not feel completely separated and unable to communicate between levels, teachers will work together to make sure that one grade's curriculum leads into the next.
      Communication between grades and buildings creates a big-picture view. This can be enhanced when administration allows time for teachers to meet and discuss needs and perceptions between grades and between schools. The elementary school should meet and discuss student preparedness with the middle school, and the middle school should do the same with the high school. When this occurs, each teacher sees how the work he performs fits into the entire task of educating the student, creating meaning for that teacher's struggles.

    Task Significance

    • Teachers know that their work is important, but it doesn't hurt to have others agree occasionally. The teacher with the "If you can read this, thank a teacher," bumper sticker on his car is motivated by Task Significance.
      Some teachers are driven by this characteristic. They have an eye to the future. This factor is probably the one that keeps teachers in the classroom. When students repeatedly complain about how "stupid" the assigned tasks are, teachers' unconscious perceptions of task significance become worn. If that same teacher has an administration that ignores the workload created by a teacher's classroom duties, cutting back on materials and planning time, that teacher begins to believe that her efforts in the classroom do not count, and task significance is lost.
      To help a teacher stay motivated by task significance, a school district or parent/teacher organization can try to find ways to remind teachers that they have an important impact on their students' lives. Making planning time and materials a priority would help. Some schools have used a yearly letter writing campaign from students to teachers who have had a positive influence on them. Keeping track of alumni and reporting what those students have achieved in their adult lives also helps teachers remember the significance of their task.

    Autonomy

    • In an era when laws, standards, and political agendas dictate what needs to happen in the classroom, teachers feel less control over what they can do. The teacher motivated by autonomy may exclaim, "I am a professional. I know what it takes to do my job." This teacher wants to be given the assignment, the time, and the material knowing that he will be held accountable for accomplishing the district's clearly stated goal.
      When school administration feels the need to manage every aspect of a teacher's planning period or the way that teacher spends all time allotted to instructional development, he will lose sight of the feeling that the administration respects his professionalism. Parents who second guess teachers and undermine the teacher's decisions also diminish a teacher's feeling of autonomy.
      An administration that can responsibly release power to teachers to perform their jobs, respecting their ability to recognize situations and respond to them accordingly, creates that feeling of autonomy teachers need to feel like respected professionals.

    Feedback From the Job

    • Everyone likes a job well done. A teacher who is motivated by feedback will talk about student performance saying things like, "Look at how well my students are doing. Tommy is better at writing papers than he was when he arrived in my class, and everybody really seems to be understanding what we're doing." Teaching comes ready made to deliver feedback from the job. It comes in the form of test scores, homework grades, student attitudes, and parent and administrator reactions. When most or all of these are clear, so is the motivation engendered by them.
      When students are unmotivated despite a teacher's best efforts to modify instruction and reach them, the year can become long and hard. If teachers only hear negative feedback from the administration and "no news is good news," a teacher's perception of feedback from the job is uncertain and most likely negative.
      While no one can completely control a classroom's motivation to learn or make students study, parents and administration can strive to give teachers at least as much positive feedback as negative feedback. People who work with children are advised to give three positive comments for every negative one. This goal works with school employees as well. Be honest, but upbeat, and teachers will remain motivated by feedback from their job.

    Conclusion

    • When all these characteristics have been met, the teaching climate should be quite comfortable and teachers should be very motivated and should be experiencing the meaningfulness of their work, feeling responsibility for what they do, and knowing the results of their work as well. All this will result in satisfied teachers and better educated students.

Related Searches






References

Resources

  • Photo Credit teacher & students image by Luisafer from Fotolia.com

Comments

You May Also Like

Helpful?
Yes
No

Thursday, February 9, 2012

Does Job design influence motivation?

by V S Rama Rao on July 29, 2010
Job characteristics model (JCM): Hackman and Oldham’s job description model. The five core job dimensions are skill significance autonomy and feedback.
What differentiates one job from another? We know that a traveling salesperson’s job is different from that of an emergency room nurse. And we know that both of those jobs have little in common with the job of an editor in a newsroom or that of a component assembler on a production line. But what is it that allows us to draw these distinctions? We can answer these questions through the job characteristics model (JCM) developed by J Richard Hackman and Greg R Oldham.
1) Skill variety: The degree to which the job requires a variety of activities so the worker can use a number of different skills and talents.
2) Task identity: The degree to which the job requires completion of a whole and identifiable piece of work.
3) Task significance: The degree to which the job affects the lives or work of other people
4) Autonomy: The degree to which the job provides freedom, independence and discretion to the individual in scheduling the work and in determining the procedures to be used in carrying it out.
5) Feedback: The degree to which carrying out the work activities required by the job results in the individual’s obtaining direct and clear information about the effectiveness of his or her performance.
Exhibit below presents the model.
The Job characteristics model
Core Job dimensions  Critical psychological states  Personal and work outcomes
Skill variety Task identity Task significance  Experienced meaningfulness of the work  High Internal work motivation
Autonomy  Experienced responsibility for outcomes of the work
Feedback  Knowledge of the actual results of the work activities  Low absenteeism and turnover
Feedback  Employee growth  Low absenteeism.
Notice how the first three dimensions –skill variety, task identity and task significance – combine to create meaningful work. What we mean is that if these three characteristics exist in a job, we can predict that the person will view his or her job as being important, valuable and worthwhile. Notice too that jobs that possess autonomy give the job incumbent a feeling of personal responsibility for the results and that if a job provides feedback, the employee will know how effectively he or she is performing.
From a motivational point of view, the JCM suggest that internal reward are obtained when an employee learns (knowledge of results trough feedback) that he or she personally (experienced responsibility through autonomy of work) has performed well on task that he or she acres about (experienced meaningfulness through skill variety, task identity and /or task significance). The more these three conditions characterize a job, the greater the employee’s motivation performance and satisfaction and the lower his or her absenteeism and the likelihood of resigning. As the model shows the links between the job dimensions and the outcomes are moderated by the strength of the individuals are moderated by the strength of the individual’s growth need (the person’s desire for self esteem and self actualization). Individuals are more likely to experience the critical psychological states and respond positively when their jobs include the core dimensions than are individuals with a low growth need. This distinction may explain the mixed results with job enrichment (vertical expansion of a job by adding planning and evaluation responsibilities): Individuals with low growth need don’t tend to achieve high performance or satisfaction by having their jobs enriched.
The JCM provides significant guidance for job redesign for both individuals and teams (Exhibit below). The suggestions in Exhibit which are based on the JCM, specify the types of changes in jobs that are most likely to improve in each of the five core job dimensions.
Guidelines for job Redesign
Suggested action Core job Dimensions
Combine tasks  Skill variety  Task identity
Form natural work units  Task identity  Task significance
Established client relationship  Skill variety  feedback
Expand jobs vertically  Autonomy
Open feedback channels  Feedback



more at http://www.citeman.com/9802-does-job-design-influence-motivation.html#ixzz1lqha7Y1P

CURRENT RESEARCH IN SOCIAL PSYCHOLOGY


Volume 5, Number 12
Submitted: March 22, 2000
Resubmitted: May 11, 2000
Accepted: May 22, 2000
Publication date: May 22, 2000

THE IMPORTANCE OF THE CRITICAL PSYCHOLOGICAL STATES IN THE JOB CHARACTERISTICS MODEL: A META-ANALYTIC AND STRUCTURAL EQUATIONS MODELING EXAMINATION

Scott J. Behson
Fairleigh Dickinson University
Erik R. Eddy
The Group for Organizational Effectiveness, Inc.
Steven J. Lorenzet
Rider University
ABSTRACT
Hackman and Oldham (1976) originally proposed their Job Characteristics Theory as a three-stage model, in which a set of core job characteristics impact a number critical psychological states, which, in turn, influence a set of affective and motivational outcomes (see Figure 1). Interestingly, most subsequent research has omitted the critical psychological states, focusing, instead, on the direct impact of the core job characteristics on the outcomes (i.e., a two-stage model). Meta-analytic data from the thirteen studies that have investigated the full, three-stage Job Characteristics Model was used as input into a structural equations modeling analysis (Viswesvaran & Ones, 1995) to examine competing versions of the Job Characteristics Model and to determine the importance of the critical psychological states. Results suggest that, while the two-stage model demonstrates adequate fit to the data, information on the critical psychological states is important for both theoretical and practical reasons.
[170]
---------------
[171]
Figure 1. Hackman & Oldham’s (1976) Job Characteristics Model
RESEARCH ON THE JOB CHARACTERISTICS MODEL
Hackman and Oldham’s (1975, 1976, 1980) Job Characteristics Model (JCM) is one of the most influential theories ever presented in the field of organizational psychology. It has served as the basis for scores of studies and job redesign interventions over the past two decades, and this research has been extensively reviewed (Fried & Ferris 1987; Loher, Noe, Moeller & Fitzgerald, 1985; Taber & Taylor, 1990). The majority of research has supported the validity of the JCM, although critiques and modifications have been offered (Roberts & Glick, 1981; Salancik & Pfeffer, 1978).
[171]
---------------
[172]
Interestingly, an evaluation of the research that has been conducted on the JCM suggests that few researchers have tested the model the way in which it was originally proposed. According to Hackman and Oldham (1976, 1980), the critical psychological states (CPS) make up the "causal core of the model" and should fully mediate the effects of the core job characteristics (CJC) on relevant individual outcomes. Hackman and Oldham developed the model by identifying psychological states important for job satisfaction and motivation, and then worked backwards to identify job characteristics that could elicit these psychological states. Thus, the model is centered around the critical psychological states, and "the core job characteristics were identified to serve the critical psychological states, not the other way around" (Johns, et al., 1992, p. 658).
Although much of the earliest research into the validity of the JCM (e.g., Arnold & House, 1980; Wall, Clegg, & Jackson, 1978) explicitly examined all of the linkages within the JCM, most subsequent investigations have omitted the CPS, and have instead investigated only the direct relationships between the CJC and a number of outcomes. "One of the most critical gaps in JCM research involves how infrequently the total model has been tested . . . the rarity of studies that incorporate the mediating psychological states is remarkable" (Johns, et al., 1992, p. 658). Further, "since few studies have included the CPS, one could question whether the motivational underpinnings of this theory have been adequately examined or represented in JCM evaluations" (Renn & Vandenberg, 1995, p. 280).
The omission of the CPS from JCM investigations could be warranted if there were theoretical or practical rationale for this practice. However, "virtually no empirical evidence has accumulated supporting the practice of excluding the CPS from tests of the theory. The practice of excluding the mediating role appears to have occurred without empirical or theoretical justification" (Renn & Vandenberg, 1995, p. 280; see also Fried & Ferris, 1987; Hogan & Martel, 1987).
[172]
---------------
[173]
Most importantly, the omission of the CPS from empirical investigations of the JCM could lead to erroneous predictions (Fox & Feldman, 1988). For example, the fact that skill variety has been found to be positively correlated with job satisfaction could lead practicing managers to believe that satisfaction can be improved simply by increasing this CJC. However, according to the JCM, skill variety should only lead to positive outcomes to the extent that this increase results in a corresponding increase in experienced meaningfulness of the work. If an increase in variety does not result in increased feelings of meaningfulness, it is reasonable to hypothesize that this would result in a negative or non-significant change in satisfaction. The increased variety might only reflect more boring, meaningless things to do. In short, without measuring the CPS, our understanding of how CJC affect work outcomes can be incomplete or misleading. Due to the prominence of the JCM, the lack of data regarding the relationships between the CPS and the other elements of the JCM can have far-reaching consequences.
Further, this lack of available data has prevented the major meta-analytic reviews of the JCM from making definitive statements about the CPS. While Fried and Ferris (1987) included 76 studies in their meta-analysis of the JCM, they could find only eight studies that examined the entire JCM (i.e., including the CPS) and only three that tested the mediating effects of the CPS. Thus, Fried and Ferris (1987) were unable to make definitive conclusions as to the validity or importance of the CPS, although they stated in their qualitative discussion that there was suggestive evidence that the CPS are critical to the model. The Loher et al. (1985) meta-analysis did not address the critical psychological states at all. Rather, it focused solely on the relationships between the CJC and satisfaction. Thus, despite over two decades of active research on the JCM, the there has yet to be a comprehensive statement made concerning the role of the CPS in the JCM, and there has yet to be a quantitative review of the JCM examining all the relationships within the JCM.
[173]
---------------
[174]
Recently, however, several researchers have called for, and conducted research on, the full JCM model, with particular emphasis on the CPS. In general, these more recent studies have utilized sophisticated analytic techniques such as structural equations modeling, as opposed to bivariate correlation analysis. While the results and conclusions of these investigations have varied, there is general consensus that (a) the original JCM represents an adequate, but imperfect model, (b) the inclusion of the CPS in the investigation of the JCM explains additioanl variance in the outcome measures, and (c) that the CPS may represent partial, not complete, mediators of the CJC-outcome relationships. Due to the renewed interest in examining the CPS, we feel that there are a sufficient number of studies to warrant a summary analysis. Thus, the goals of this paper are to: (a) quantitatively summarize the findings of all existing studies which have examined the complete JCM, (b) test the adequacy of the original Hackman and Oldham model against the more commonly researched two-stage model, and (c) provide evidence to judge the importance of the CPS to the JCM.
The two competing models tested in this study are: (1) The original Job Characteristics Model, as proposed by Hackman and Oldham (1976) and (2) A modified JCM in which the critical psychological states are omitted. The original model will be tested to provide a test of the adequacy of the original model among the studies that have measured the JCM in its entirety. It is expected that the original model will provide an adequate fit for the data. The modified model represents the vast majority of studies that have measured the links between CJC and outcomes, while omitting the intervening CPS. It is expected that this model will not be as adequate as the models that encompass all three stages of the JCM (Renn & Vandenberg, 1995; Hogan & Martel, 1987). Please note that moderator variables, such as Growth Need Strength, were not incorporated into the tested models. This decision is discussed later in the paper.
The present study utilizes both meta-analytic and structural equation modeling techniques (see Viswesvaran & Ones, 1995) to provide a comprehensive test of the JCM based on the collected results of past research. "Another need for future research is to continue to utilize structural equation modeling to analyze data already collected. Numerous JCM data sets have been analyzed with less sophisticated techniques; such data could be re-analyzed using causal modeling. . . . The resulting group of analyses, taken as a whole, might then be subjected to meta-analysis" (Hogan & Martel, 1987; p. 261-2).
[174]
---------------
[175]
This approach for studying the JCM seems appropriate for several reasons (Hogan & Martel, 1987). First, structural equations modeling is appropriate for testing competing interpretations of the same model. Second, structural equations modeling can handle the simultaneous and multiple-stage nature of the mediated job characteristics model better than traditional regression analytic techniques. Further, the use of meta-analytic data also helps us avoid problems such as small sample size, low power, and homogeneous samples of jobs and organizations.
In addition, our analysis has been able to avoid the most common concerns that have been expressed regarding the use of the procedures as laid out in Viswesvaran and Ones (1995). First, the use of meta-analytic input could lead to vastly different sample sizes for each cell in the input matrix. This does not appear to be a problem for the current analysis because all values were gathered from meta-analytic samples ranging from 8,016 to 8,964 individual subjects.
Second, some are concerned that widely discrepant operationalizations could be combined as indicators of the same latent variable. All of the studies included in the meta-analysis used the measurement scales from the Job Diagnostic Survey (JDS) (see Hackman & Oldham, 1975, 1980), obviating this concern. Finally, some researchers caution that the use of these procedures could result in a correlation matrix in which there are missing values. In this analysis, there are no missing values in the meta-analytic correlation matrix.
METHOD
Relevant studies were gathered through a variety of sources: (a) a computer-based search of JCM keywords using Psychlit and Dissertation Abstracts dating back to 1976, (b) a reference list search of found articles and existing JCM meta-analyses, and (c) a hand search of five prominent organizational psychology/management journals (Academy of Management Journal, Journal of Applied Psychology, Journal of Management, Organizational Behavior and Human Decision Processes/Human Performance, and Personnel Psychology), from 1976 to 1998. The literature search yielded a total of thirteen independent studies appropriate for inclusion in the meta-analysis. Inclusion criteria for studies were (a) the study must contain information regarding the full JCM, including the CPS, and (b) the study must report correlations between CPS and CJC and/or outcome measures.
[175]
---------------
[176]
Studies were divided among the three authors and coded independently. To insure reliability, articles were divided again and re-coded by a different author. Disagreements were resolved by discussion. Table 1 provides a list of all the studies included in the meta-analysis, their sample size, sample, measure used, and whether the study supports the importance of the CPS in the JCM. Please note that no study that explicitly examined the CPS found them to be entirely unimportant to the JCM model.
Table 1. Characteristics of Studies Included in the Meta-Analysis
  N Samples Measures Support for CPS?
Arnold & House (1980) 120 Engineers JDS Did Not Test
Barnabe & Burns (1994) 247 Teachers JDS Yes
Becherer, Morgan, & Lawrence (1982) 211 Sales JDS Yes
Champoux (1991) 247 State Agency JDS Partial
Fox & Feldman (1988) 119 Variety of Jobs JDS/JDI Partial
Griffeth (1985) 76 Work Study JDS Did Not Tex
Hackman & Oldham (1975) 658 Variety of Jobs JDS Partial
Hogan & Martell (1987) 208 NAVY-Variety of Jobs JDS Yes
Johns, Xie, & Fang (1992) 300 Managers JDS Yes
Kiggundu (1980) 138 Financial Company JDS Did Not Test
Renn & Vandenberg (1995) 188 Variety of Jobs JDS/JDS-R Yes
Tiegs, Tetrick, & Fried (1992) 6405 Variety of Jobs JDS Did Not Test
Wall, Clegg, & Jackson (1978) 47 Sales JDS Partial
Studies were coded for three potential moderator variables: sample type (white collar, blue collar, mixed), research design (experiment, quasi-experiment, non-experiment), and instrument used (JDS, JDS-Revised, other). The analyses for type of sample revealed no consistent pattern of differences. Analyses were not conducted for the other two variables, due to the lack of variation among primary studies.
[176]
---------------
[177]
The meta-analytic correlations between each of the elements are displayed in Table 2. Each of the effect sizes were based upon between nine and thirteen independent samples and upon between 8,016 and 8,964 participants. The mean sample size of each of the studies included in the meta-analysis was 690 and the median sample size was 208. Effect sizes were not corrected for unreliability at this stage of the analysis. This correlation matrix was transformed into a covariance matrix using the standard deviations calculated by Oldham, Hackman, and Stepina (1979), which are based on 6,930 respondents from 876 different jobs in 56 organizations and were previously used to represent population parameters by Arnold and House (1980), Fried and Ferris (1987) and Hackman and Oldham (1980). The reader should note that the standard deviations used in this analysis are based on normative data, and were not meta-analytically derived from the included studies.
Table 2. Meta-Analytic Correlations and Mean Reliabilities
  SD 1 2 3 4 5 6 7 8 9 10 11
1. Skill Variety 1.57 .70                    
2. Task Significance 1.25 .41 .59                  
3. Task Identity 1.44 .22 .20 .65                
4. Autonomy 1.39 .43 .32 .32 .67              
5. Feedback 1.34 .35 .34 .26 .39 .71            
6. Experienced Meaningfulness 1.14 .46 .45 .24 .42 .38 .75          
7. Experienced Responsibility 0.96 .34 .33 .27 .39 .34 .59 .71        
8. Knowledge of Results 1.14 .16 .23 .28 .29 .49 .40 .34 .72      
9. Satisfaction 1.07 .35 .29 .22 .42 .36 .65 .49 .42 .80    
10. Growth 1.15 .50 .38 .26 .54 .44 .65 .51 .40 .69 .81  
11. Internal Satisfaction 0.77 .35 .33 .17 .30 .42 .57 .59 .25 .43 .50 .69
Note. Mean reliabilities are reported on the diagonal.
Note. All 95% confidence intervals did not include zero.
Note. Standard deviations from Oldham, Hackman & Stepina (1979)
[177]
---------------
[178]
Next, the procedures outlined by Viswesvaran and Ones (1995) for using meta-analysis to create a covariance matrix to be used as input to a structural equations analysis were employed. The seven-step process is shown in Table 3. Similar procedures have been employed by Carson, Carson, and Rowe (1993), Horn, Caranikas-Walker, Prussia and Griffeth, (1992), and Premack and Hunter (1988), among others. Our meta-analysis is consistent with these procedures, except that (a) a LISREL 8.0 analysis was performed instead of traditional path analysis and (b) the correlations used in the analysis were not corrected for attenuation due to unreliability. This decision will be discussed later in the paper.
Table 3. Steps for Combining Psychometric Meta-Analysis and Structural Equations Modeling
Measurement Model
  1. Identify important constructs and relationships.
  2. Identify different measures used to operationalize each construct.
  3. Obtain all studies reporting either (a) correlations between conceptually distinct operational measures or (b) artifact information on any of the conceptually distinct operational measures (identified in step 2).
  4. Conduct psychometric meta-analyses and estimate true score correlations between the measures (identified in step 2).
  5. Use factor analysis to test the measurement model.
Causal Model
  1. Estimate the correlations between the constructs (forming composites for the different operationalizations of the same construct).
  2. Use LISREL with the estimated true score correlations to test proposed theory.
Note. Adapted from framework presented by Viswesvaran and Ones (1995).
[178]
---------------
[179]
For the LISREL 8.0 analyses, the parameter estimates were based on a sample covariance matrix and a maximum likelihood solution. The median sample size, 208, was used in this stage of the analysis because the X2 statistic is biased against large sample sizes (Jaccard & Wan, 1996).
The fit of the data to the model was assessed using several indices, including: the X2 statistic, the Goodness of Fit Index (GFI), the Root Mean Square Error of Approximation (RMSEA), and the Comparative Fit Index (CFI). The X2 statistic, and the GFI are indices of absolute fit which measure how far the model deviates from a model of perfect fit. The CFI is an index of comparative fit that measures how far a model deviates from a model of good fit. The RMSEA is a test of parsimony that takes the number of paths into account when determining fit. Model adequacy is also assessed by examining the amount of variance explained in the outcome measures and the ratio of predicted to significant paths.
The GFI, CFI, and RMSEA statistics are useful for assessing the fit of the individual models; however, they cannot be used to compare across models. The X2 statistic can be used to compare the relative fit of competing models, but only if these models are nested within each other. However, the two models being compared in this study are not nested. Therefore, two commonly used statistical indices, the Akaike Information Criterion (AIC) and the CIAC (an extension of the AIC, which more strongly penalizes models for lack of parsimony), were used to compare these two non-nested models on a common metric. These statistics are seen as most appropriate when comparing two non-nested models (see Lin & Dayton, 1997).
RESULTS
First, the original JCM model (Model 1) was tested (see Table 4). The fit indices for this model were: X2 (25) = 124.25, p < .05, GFI = .91, RMSEA = .14, and CFI = .89. The CFI and GFI indicate acceptable levels of model fit, while the RMSEA and the X2 value are less supportive of good model fit. However, X2 is influenced by sample size, and the RMSEA index penalizes models for lack of parsimony. Therefore, these findings are not unexpected.
[179]
---------------
[180]
Table 4. Results of Tests of Goodness of Fit for the Various Models
Statistic
c2 df Ratio of explained paths RMSEA GFI CFI Explained variance in DV
Model AIC
Model CIAC
Rules of Thumb for "Good Fit"
ns - - <.08 >.90 >.90 - - -
1. Original Job Characteristics Model
124.25* 25 12/14 .14 .91 .89
.42 sat.
.42 growth
.38 mot.
294.48 446.29
2. Normally Tested JCM (excluding CPS)
12.09* 3 7/15 .16 .99 .98
.37 sat
.43 growth
.32 mot
80.09 227.56
Note. * indicates result was statistically significant at p < .05
Figure 2 shows the estimates of the structural coefficients for Model 1. Standardized estimates appear on each path. Twelve of the fourteen paths in this model were statistically significant, and the variables in the model were able to account for approximately 42% of the variance in satisfaction, 42% of the variance in growth satisfaction, and 38% of the variance in motivation.
Figure 2. SEM of the Original JCM
[180]
---------------
[181]
Next, the two-stage model normally tested in the literature was explored (Model 2). The results of the goodness of fit indices were: X2 (3) = 12.09, p < .05, GFI = .99, RMSEA = .16, and CFI = .98. All of these values, except for the RMSEA, indicate good model fit. Seven of the fifteen paths were statistically significant in this model (see Figure 3). The model was able to account for approximately 37% of the variance in satisfaction, 43% of the variance in growth satisfaction, and 32% of the variance in motivation.
Figure 3. SEM of the JCM Normally Tested in the Literature
[181]
---------------
[182]
In short, the original JCM can be seen to (a) explain more variance in the dependent variables, and (b) have a greater percentage of statistically significant causal pathways than the abridged version of the JCM. The two-stage JCM, however, attained greater model fit, as indicated by the GFI, CFI and chi-squared indices. Neither model showed an acceptable level of parsimony according to the RMSEA index.
Finally, in order to compare the models with a common metric, the AIC and CIAC statistics were used. When comparing two or more models, the model of best fit is the one with the lowest values (Lin & Dayton, 1997). Both the AIC and the CAIC indicate that the normally tested two-stage model demonstrates superior fit (see Table 4).
DISCUSSION
The quantitative results of this analysis suggest that the two-stage model normally tested in the literature may provide a better fit to the available data than the three-stage model originally proposed by Hackman and Oldham (1976). However, adequate comparison among competing models requires more than comparing fit ratios. The reasonableness of values contained in a model and a model’s correspondence with relevant theory are equally, if not more, important. Thus, while the two stage model may result in more adequate model fit, a closer examination of the two models support, rather than refute, the contention that the CPS are indeed critical to the JCM.
Several path coefficients in Model 2 run counter to well-established theory regarding the design of work. In particular, eight of the nine paths between skill variety, task significance, and task identity and the three outcome variables are not statistically significant (see Figure 2). In comparing these path coefficients with those of Model 1, the importance of the CPS to the JCM becomes clear. In Model 1, both skill variety and task significance demonstrate statistically significantly positive indirect relationships with the outcome variables, as mediated by experienced meaningfulness. These relationships provide evidence that, while skill variety and task significance may not be directly related to job affect and motivation, they can be important in eliciting experienced meaningfulness of the work. It is this psychological state, however, that is crucial for the beneficial outcomes of job redesign. Thus, the comparison between the path coefficients in these two competing models accentuates the importance of the CPS to job redesign. The non-significant paths in Model 2 provide evidence that increasing job characteristics may have little or no impact if the employee does not experience the CPS. This underscores the importance of the CPS as the "causal core of the model" (Hackman & Oldham, 1976, p. 255).
[182]
---------------
[183]
Our results also lead to several other interesting observations. For instance, in both of the competing models, autonomy is the CJC with the strongest relationships with outcome variables. This finding is consistent with several recent streams of research into work motivation, including Ajzen’s (1991) Theory of Planned Behavior and Deci and Ryan’s Cognitive Evaluation Theory (e.g., Deci & Ryan, 1991), which stress the importance of autonomy and self-determination. Further, recent practitioner-oriented research on organizational development and change has established that allowing personal control is a key to successful change in employee attitudes, behaviors, and value orientation (e.g., Parker, Wall & Jackson, 1997).
In addition, it should be noted that neither model tested in this study demonstrated exceptional fit to the data. It was certainly expected that the JCM, in either form, would not be particularly parsimonious. However, this study does provide some suggestions for avenues of future research. In particular, research aimed at trimming the model and balancing parsimony and variance explanation concerns is clearly warranted. Again, autonomy is seen as a particularly crucial construct for this purpose.
The limitations of the present study also warrant discussion. First, the meta-analytic data was derived from only 13 primary studies, and some have argued that this relatively low k could lead to unstable meta-analytic results (Oswald & Johnson, 1998). However, this number of primary studies is not uncommonly low, given recent publications (e.g., Donovan & Radosevich, 1998). Further, our data was derived from a large number of subjects (n varied from 8,016 to 8,964) across a wide variety of occupations and job settings. Thus, one can be reasonably confident in the external validity of our results.
Another potential criticism of this research is that a large proportion of our sample was derived from one primary study (Tiegs, et al., 1992). To address this concern, we ran our analyses both with and without this study included in our sample, and found no significant differences. In fact, in comparing the two resultant correlation matrices, only one of the fifty-five pairs of correlations differed by more than .05.
[183]
---------------
[184]
The combined use of meta-analysis and SEM is a relatively new analytic strategy. While this technique is promising, its validity hinges on a number of statistical assumptions. Efforts were made to address some commonly voiced concerns regarding this technique. However, more psychometric and simulation-based research regarding the limits and potential drawbacks of this approach is clearly needed.
In addition, we did not include any information on moderating variables, such as Growth Need Strength (GNS), in our analysis. This decision was made for several reasons, including: (a) the fact that few of the studies selected for our meta-analysis included information on GNS, (b) Tiegs, Tetrick & Fried (1992) offer compelling evidence that GNS is not, in fact, a significant moderator of the relationships in the model, (c) that the analysis of the GNS moderator in the manner originally proposed by Hackman and Oldham (moderation at two stages) is troublesome and would either require the addition of 14 additional paths to Model 1 or the splitting of continuous variables into categorical ones (Jaccard & Wan, 1996), and (d) the effects of moderators are tangential to the specific purpose of the present paper.
Finally, the correlations used as input to the structural equations analysis were not corrected for unreliability at either the meta-analytic stage or the SEM stage, although techniques for such corrections are commonly employed. There were two reasons for this decision. First, research on the JCM and the JDS have long acknowledged that common method variance and multicollinearity serve to inflate the correlations among the JCM constructs (Roberts & Glick, 1981; Taber & Taylor, 1990). While unreliability serves to attenuate correlations, correcting for this attenuating effect while ignoring the factors which serve to artificially inflate variable correlations would result in biased correlations which overstate the strength of the relationships among the JCM variables. Second, when the analyses were conducted using corrected correlations as input, several statistical problems were encountered. In particular, the inflated correlations led to suppressor effects among the independent variables in Model 2 (the abridged model). This led to several statistically troubling results, including a standardized path coefficient greater than 1.0 (1.41 between autonomy and satisfaction) and negative causal paths between variables whose zero-order correlations are positive.
[184]
---------------
[185]
Two potential causes of these supressor effects are that the average reliabilities calculated from the primary studies were consistently lower than acceptable standards for scale reliability (along the diagonal in Table 2), and that multicollinearity may exist among the variables in the model. Our findings are consistent with Roberts and Glick’s (1981) and Taber and Taylor’s (1990) conclusions that the JDS is a useful, albeit limited, instrument, but that additional and alternate measures and methodologies are required in order to advance the field of job redesign. Thus, due to statistical anomalies and our desire to remain conservative in our analyses, no corrections for attenuation were made.
In sum, the central finding of the present analysis is that, while the abridged two-stage model demonstrates adequate fit, JCM researchers need to pay more attention to the CPS. The results of our meta-analysis support recent contentions that "researchers and practitioners who are interested in the impact of jobs on employees might consider measuring psychological states more often than is commonly done" (Johns, et al., 1992, p. 672). Thus, this paper contributes quantitative evidence to support those who have criticized how research has commonly been conducted on the JCM (see Fried & Ferris, 1987; Fox & Feldman, 1988; Hogan & Martel, 1987; Renn & Vandenberg, 1995).
Failure to incorporate CPS into the JCM could lead to unexpected results and misdirected organizational interventions. This classic theory is quite complex and rich, and has implications for many of the workplace change initiatives (e.g., JIT, TQM, MBO) in use in organizations today. Even though the two-stage model represents a more parsimonious model, important information may be lost if the CPS are not included.
REFERENCES
Ajzen, I. (1991). 'The theory of planned behavior." Organizational Behavior and Human Decision Processes, 43:179-211.
Arnold, H.J. & House, R.J. (1980). "Methodological and substantive extensions to the job characteristics model of motivation." Organizational Behavior and Human Decision Processes, 25:161-183.
[185]
---------------
[186]
Barnabe, C. & Burns, M. (1994). "Teachers’ job characteristics and motivation." Educational Research, 36:171-185.
Becherer, R.C., Morgan, F.W. & Richard, L.M. (1982). "The job characteristics of industrial salespeople: Relationship to motivation and satisfaction." Journal of Marketing, 46:125-135.
Brief, A.P. & Aldag, R.J. (1975). "Employee reactions to job characteristics: A constructive replication." Journal of Applied Psychology, 60:182-186.
Carson, P.P., Carson, K.D. & Roe, C.W. (1993). "Social power bases: A meta-analytic examination of interrelationships and outcomes." Journal of Applied Social Psychology, 23:1,150-1,169.
Champoux, J.E. (1991)."A multivariate test of the job characteristics theory of work motivation." Journal of Organizational Behavior, 12:431-446.
Deci, E.L. & Ryan, R.M. (1991). "Intrinsic motivation and self-determination in human behavior." In Steers, R.M. & Porter, L.W. (Eds.) Motivation and Work Behavior, 5th Edition. New York: McGraw-Hill, Inc., pp. 44-58.
Donovan, J.J., & Radosevich, D.J. (1998). "The moderating role of goal commitment on the goal difficulty-performance relationship: A meta-analytic review and critical reanalysis." Journal of Applied Psychology, 83:308-315.
Evans, M.G., Kiggundu, M.N. & House, R.J. (1979). "A partial test and extension of the job characteristics model of motivation." Organizational Behavior and Human Performance, 24:354-381.
Fox, S. & Feldman, G. (1988). "Attention state and critical psychological states as mediators between job dimensions and job outcomes." Human Relations, 41:229-245.
[186]
---------------
[187]
Fried, Y. & Ferris, G.R. (1987). "The validity of the job characteristics model: A review and meta-analysis." Personnel Psychology, 40:287-322.
Griffeth, R.W. (1985). "Moderation of the effects of job enrichment by participation: A longitudinal field experiment." Organizational Behavior and Human Decision Processes, 35:73-93.
Hackman J.R. & Oldham, G.R. (1976). "Motivation through the design of work: Test of a theory." Organizational Behavior and Human Performance, 16:250-279.
Hackman J.R. & Oldham, G.R. (1980). Work Redesign. Reading, MA: Addison-Wesley Publishing Company Inc.
Hackman J.R. & Oldham, G.R. (1975). "Development of the job diagnostic survey." Journal of Applied Psychology, 60:159-170.
Hogan E.L. & Martell, D.A. (1987). "A confirmatory structural equations analysis of the job characteristics model." Organizational Behavior and Human Decision Processes, 39:242-263.
Horn, P.W., Caranikas-Walker, F., Prussia, G.E. & Griffeth, R.E. (1992). "A meta-analytical structural equations analysis of a model of employee turnover." Journal of Applied Psychology, 77:890-909.
Jaccard, J. & Wan, C.K. LISREL Approaches to Interaction Effects in Multiple Regression. Thousand Oaks, CA: Sage Publications.
Johns, G., Xie, J.L. & Fang, Y. (1992). "Mediating and moderating effects in job design." Journnal of Management, 18:657-676.
Kiggundu, M.N. (1980). "An empirical test of the theory of job design using multiple job ratings." Human Relations, 33:339-351.
Lin, T.H & Dayton, C.M. (1997). "Model selection information criteria for non-nested latent class models." Journal of Educational and Behavioral Statistics, 22:249-264.
[187]
---------------
[188]
Loher, B.T., Noe, R.A., Moeller, N.L. & Fitzgerald, M.P. (1985). "A meta-analysis of the relation of job characteristics and job satisfaction." Journal of Applied Psychology, 70:280-289.
Oldham, G.R., Hackman, J.R. & Stepina, L.P. (1979). "Norms for the Job Diagnostic Survey." JSAS Catalog of Selected Documents in Psychology (Ms. No. 1819), 9, 14.
Oldham, G.R. & Rotchford, N.L. (1983). "Relationships between office characteristics and employee reactions." Administrative Science Quarterly, 28:542-556.
Oswald, F.L. & Johnson, J.W. (1998). "On the robustness, bias, and stability of statistics from meta-analysis or correlation coefficients: Some initial monte carlo findings." Journal of Applied Psychology, 83:164-178.
Premack, S.L. & Hunter, J.E. (1988). "Individual unionization decisions." Psychological Bulletin, 103:223-234.
Renn, R.W. & Vandenberg, R.J. (1995). "The critical psychological states: An underrepresented component in job characteristics model research." Journal of Management, 21:279-303.
Roberts, K.H. & Glick, W. (19810). "The job characteristics approach to task design: A critical review." Journal of Applied Psychology, 66:193-217.
Salancik, G.R. & Pfeffer, J. (1978). "A social information processing approach to job attitudes and task design." Administrative Science Quarterly, 23:224-253.
Taber, T.D. & Taylor, E. (1990). "A review and evaluation of the psychometric properties of the job diagnostic survey." Personnel Psychology, 43:467-500.
Tiegs, R.B., Tetrick, L.E. & Fried, Y. (1992). "Growth need strength and context satisfaction and moderators of the relations of the Job Characteristics Model." Journal of Management, 18:575-593.
[188]
---------------
[189]
Viswesvaran, C. & Ones, D.S. (1995). "Theory testing: Combining psychometric meta-analysis and structural equations modeling." Personnel Psychology, 48:865-885.
Wall, T.D., Clegg, C.W. & Jackson, P.R. (1978). "An evaulation of the job characteristics model." Journal of Occupational Psychology, 51:183-196.
AUTHOR BIOGRAPHIES
Dr. Scott J. Behson is an Assistant Professor of Management at Fairleigh Dickinson University, where he teaches, conducts research, and provides consulting services in organizational change, organizational behavior and human resource management. Scott is also a member of the Center for Human Resource Management (CHRMS) (www.chrms.org) at FDU. Email: Behson@mailbox.fdu.edu, Website: www.scottbehson.homestead.com
Dr. Erik Eddy is a Project Director with The Group for Organizational Effectiveness. His interests include continuous and organizational learning, informal methods of knowledge acquisition, and organizational privacy. E-mail: Erik.Eddy@groupOE.com
Dr. Steven Lorenzet is a recent graduate of the University at Albany, SUNY and will be beginning an appointment in the fall at Rider University as an Assistant Professor of Human Resource Management. His interests include training and development, employee and organizational learning, teams, and multiple levels of analysis. E-mail: sloren1@worldnet.att.net
[189]
---------------