Job Satisfaction, Dissatisfaction
of Texas Newspaper Reporters





Submitted by Li-jing Arthur Chang
and George Sylvie

Li-jing Arthur Chang
Assistant Professor
School of Communication Studies
Nanyang Technological University
Nanyang Avenue
Singapore 639798
Tel: (65)790-5773
Email: [log in to unmask]

George Sylvie
Associate Professor
Department of Journalism
CMA 6.144
University of Texas at Austin
Austin, Texas 78712-1093
Tel: (512)471-1783
Email: [log in to unmask]



Submitted to
Newspaper Division
Association for Education in Journalism and Mass Communication
Kim Walsh-Childers
3044 Weimer Hall
University of Florida
Gainesville, FL 32611-8400
April 1, 1999









Abstract:

        This study surveyed 365 Texas daily newspaper reporters in spring 1998 to
explore the conceptual distinction between job satisfaction and dissatisfaction
as well as factors behind job satisfaction and dissatisfaction.
        Results indicated the reporters' job satisfaction and dissatisfaction were
different concepts. Path analysis showed the predictors of their job
satisfaction include sense of achievement, personal growth, newsroom policy,
impact on community, and autonomy. The analysis also showed the predictors of
their dissatisfaction include pay and work conditions.












INTRODUCTION

        Journalists' job satisfaction can be viewed as an important indicator of job
commitment (Weaver & Wilhoit, 1994, 1996), which has been found to affect
employee turnover (Igbaria & Greenhaus, 1992). Job satisfaction in the newsroom
has gained more importance as newspapers face shrinking readership (Dizard,
1994) and need to stem revenue loss resulting from high turnover (White, 1995).
        Because of high turnover rates of newspaper reporters (Moss, 1978; Tharp,
1991), understanding job dissatisfaction is crucial because dissatisfaction,
rather than satisfaction, is among the major reasons journalists of all types
(including newspaper reporters) want to leave their field (Weaver & Wilhoit,
1996). In addition, it is important to study job satisfaction because reasons
behind satisfaction are closely tied to job performance (Herzberg, Mausner &
Snyderman, 1959; Herzberg, 1968, 1987; Shore & Martin, 1989).
        However, most journalism researchers have tackled the issue by treating both
concepts as opposite ends of the same continuum (Bergen & Weaver, 1988; Chusmir,
1984; Stamm & Underwood, 1993; Weaver & Wilhoit, 1994). Such an approach
reflects the prevailing view that job satisfaction and dissatisfaction are
determined by the same group of factors (Iiacqua, Schumacher & Li, 1995). But
Herzberg and associates argue job satisfaction and dissatisfaction are different
concepts affected by a different set of human needs (factors) (Herzberg et al.,
1959; Herzberg, 1968, 1987). Dozens of studies testing his theory found support
for it (Bockman, 1971; Moorhead & Griffin, 1989). However, none were directly
related to newspaper journalists.
        Because past research on newspaper journalists seems to suggest that job
satisfaction and dissatisfaction are influenced by different factors (Bergen &
Weaver, 1988; Chusmir, 1984; Joseph, 1981, 1982a, 1982b, 1982c; Tharp, 1991;
Weaver & Wilhoit, 1994, 1996), Herzberg's theory should be tested on newspaper
journalists, especially on the reporters, who have a high turnover rate (Moss,
1978). Support by such a test may provide a better model to explain journalist
satisfaction and, we contend, have an important bearing on newsroom management.
So, in order to test the theory in the context of newspaper reporters, we
identify and statistically analyze factors affecting the reporters' job
satisfaction and dissatisfaction to see if they affect different concepts.
        But first, we will review the pertinent literature and explain which
theoretical model for job satisfaction is most appropriate.


LITERATURE REVIEW

Theoretical Framework

        Explaining the appropriateness of Herzberg's theory also means discussing
different factors affecting newspaper reporters' job satisfaction and
dissatisfaction, as well as rival theories exploring job satisfaction.
        For example, according to Likert (1961, 1967), organizations with
democratic--as opposed to autocratic--management styles produce employees with
higher levels of job satisfaction. Although Likert implied that work autonomy
affects job satisfaction, he did not account for other factors affecting
reporter satisfaction, such as salaries and newsroom policy.
        On the other hand, Maslow (1954) identifies five types of innate needs that
motivate humans. The three lower-level needs--physiological (such as food and
water), security (such as shelter), and belongingness (such as acceptance,
love)--are deficiency needs that must be satisfied before a person can fulfill
the two higher-level growth needs: esteem (such as recognition) and
self-actualization (creativity).
        Compared with Likert, Maslow is more useful in explaining reporters' job
satisfaction because he recognized different types of needs that motivate people
(Lacy, Sohn & Wicks, 1993). For example, salaries can be viewed as a
physiological need that motivates reporters, while quality of their newspapers
can be seen as a self-actualization need. But Maslow's theory also is
sequentially rigid in that people have to satisfy basic needs before they can
fulfill higher needs. Maslow also does not discuss what happens if higher-level
needs stop being fulfilled (Lacy et al., 1993).
        Alderfer's (1972) theory addresses these issues by arguing that human beings
have three basic needs: existence, relatedness, and growth (ERG). As to reporter
satisfaction, Alderfer's theory is better in that it recognizes a person can be
motivated by more than one need at a time, and it also explains that if
higher-level needs are not satisfied, the person may return to lower-level needs
as motivators (Lacy et al., 1993). But the theory fails to distinguish between
satisfaction and dissatisfaction--an important distinction because a synthesis
of past research on newspaper journalists indicates that satisfaction and
dissatisfaction are determined by different groups of factors (Bergen & Weaver,
1988; Chusmir, 1984; Joseph, 1981, 1982a, 1982b, 1982c; Stamm & Underwood, 1993;
Tharp, 1991; Weaver & Wilhoit, 1994, 1996). Even though not directly testing on
the satisfaction-dissatisfaction distinction, these studies suggest that
newspaper reporters' satisfaction is affected by factors such as job autonomy,
the chance to help readers, and newspaper policy, while dissatisfaction is
affected by pay, work conditions and promotion opportunities.
        Unlike the Alderfer, Herzberg rejected the assumptions that job satisfaction
and dissatisfaction are at different ends of one continuum (Herzberg et
al.,1959; Herzberg, 1968, 1987). He argued satisfaction and dissatisfaction are
different concepts, each affected by a different set of human needs (factors),
called work values. Herzberg described those needs determining job
dissatisfaction as hygiene factors--extrinsic factors found in the job
environment and related to biological needs for survival and innate drives to
avoid pain. On the other hand, Herzberg described those needs affecting job
satisfaction as motivators--intrinsic factors found in the job content and are
human needs for psychological growth, such as sense of achievement and autonomy.
Therefore, Herzberg considers multiple factors (e.g., autonomy, sense of
achievement, newsroom policy, work conditions and pay) related to job
satisfaction and dissatisfaction. Herzberg's motivator-hygiene theory has been
scrutinized more closely than most other theories in organizational behavior
(Pinder, 1984). Herzberg's theory may have stimulated more research on job
satisfaction than any other satisfaction theory (Shaver, 1978).
        Over the past four decades, dozens of studies conducted on workers in different
occupations have found support for Herzberg's theory (Bockman, 1971; Caston &
Braito, 1985; Cohen, 1974; Iiacqua et al., 1995; Knoop, 1994; Phillipchuk &
Whittaker, 1996; Rosenfeld & Zdep, 1971; Schawartz, Jenusaitis & Stark, 1963;
Shaver, 1978), while some studies extended it (Rosenfeld & Zdep, 1971; Iiacqua
et al., 1995). We reviewed the empirical support for Herzberg's theory to build
the theoretical framework for the current study. Major findings from the studies
confirming Herzberg's theory include the following: (1) Motivators (intrinsic
values) primarily contribute to job satisfaction and hygiene factors (extrinsic
values) primarily contribute to job dissatisfaction; (2) specific motivators or
hygiene factors may change from profession to profession; and (3) a minority of
work values may neither be motivators nor hygiene factors and should be
categorized as "neutral" factors.

Different Factors Behind Satisfaction and Dissatisfaction

        Most studies that confirmed Herzberg's theory did so by finding that job
satisfaction and dissatisfaction are attributable to different groups of
factors--i.e., job satisfaction is affected primarily by motivators (or
intrinsic factors), while job dissatisfaction is influenced primarily by hygiene
factors (or extrinsic factors). For example, researchers (Schawartz, Jenusaitis
& Stark, 1963) asked 111 subjects from 21 public utility companies to recall two
experiences, one pleasant and the other unpleasant, from their work experiences.
The researchers identified achievement, recognition, the work itself,
responsibility, and advancement as motivators (which contributed to job
satisfaction). On the other hand, 11 items related to the job context were
identified as hygiene factors (which contributed to job dissatisfaction).
        As another example, Shaver (1978) polled 235 journalism graduates from the
classes of 1960 and 1970 to explore educators' questions about journalism
graduates' job satisfaction. The subjects were asked to select from a list of 24
factors those that affected their job satisfaction or dissatisfaction. The study
showed job satisfaction and dissatisfaction were primarily caused by different
groups of factors. The four motivators cited most frequently as contributing to
subjects' satisfaction were possibility of growth, the work itself,
responsibility, and achievement--all intrinsic factors. In contrast, the two
factors contributing most to subjects' dissatisfaction were poorly handled
company policy and low pay, both of which are extrinsic factors.
        In an attempt to test the validity of Herzberg's theory in higher-education
teachers, Iiacqua, Schumacher and Li (1995) surveyed 83 of the 137 faculty
members at an independent private business college. The study provides support
for Herzberg's study more than three decades after the theory was proposed,
showing that the theory can sustain the test of time. It also supports
Herzberg's theory may need modification due to the existence of "a neutral
factor," which affects satisfaction and dissatisfaction about equally. The
second point is especially intriguing: Since there is at least one factor that
affects both satisfaction and dissatisfaction about equally, satisfaction and
dissatisfaction cannot be entirely unrelated.
        Although Herzberg's original 1959 study (Herzberg et al.) used fairly generic
motivators (achievement, recognition, work itself, responsibility, and
advancement) and hygiene factors (company policy and administration,
supervision, salary, interpersonal relations, and work conditions), the exact
combination of work values may change from profession to profession. For
example, community college instructors' motivators and hygiene factors (Cohen,
1974) are different from those of mining company workers (Rosenfeld & Zdep,
1971). Therefore, we need to examine possible work values for the journalism
profession.
        A synthesis of past research (Bergen & Weaver, 1988; Chusmir, 1984; Joseph,
1981, 1982a, 1982b, 1982c; Stamm & Underwood, 1993; Tharp, 1991; Weaver &
Wilhoit, 1994, 1996) shows that reporters' job satisfaction and dissatisfaction
are likely to be affected by different groups of factors. The synthesis suggests
that the following factors are more likely to affect reporters' job satisfaction
than their dissatisfaction: autonomy [1], sense of achievement [2], personal
growth [3], interesting nature of work [4], and impact on the community [5].
These factors (or, in Herzberg's terms, work values) are directly related to the
job content. Specifically, autonomy relates to the amount of freedom a reporter
could get from his or her job, sense of achievement is a feeling about
completing a job, personal growth refers to what one can learn from the job,
interesting nature of work is a feeling one would get from doing the job, and
impact on community is an outcome of one's job.
        The synthesis also suggests that the following factors are more likely to
affect newspaper reporters' job dissatisfaction than their satisfaction: pay
[6], work conditions [7] and promotion opportunities [8] (Tharp, 1991; Weaver &
Wilhoit, 1994, 1996). These factors (or work values) are directly related to the
job environment. Specifically, pay relates to the financial compensation for the
job, work conditions refer to the situations under which the job is carried out
(such as work schedule, job stress and workplace facilities), and promotion
opportunities relate to job status. In addition, although evidence is
inconclusive as to whether job security should be a factor behind satisfaction
or dissatisfaction (Weaver & Wilhoit, 1994, 1996), job security is related more
to the work environment (than the job content) and should influence reporters'
job dissatisfaction (Herzberg et al. 1959; Herzberg, 1968, 1987).[9]
        Finally, different from other work values, which either affect satisfaction or
dissatisfaction, newsroom policy (on whether news quality or ad income should
take the priority) was found to affect both the reporters' job satisfaction and
dissatisfaction (Bergen & Weaver, 1988; Stamm & Underwood, 1993; Weaver &
Wilhoit, 1994).[10]
        In sum, although past studies on newspaper reporters' job satisfaction did not
specifically test Herzberg's theory (i.e., they did not treat satisfaction and
dissatisfaction as separate concepts), they implied the possibility that
reporters' satisfaction and dissatisfaction may be affected by different groups
of factors. Some studies also suggested that newsroom policy is likely to affect
both the satisfaction and dissatisfaction of newspaper reporters. If these
implications are true, our knowledge about the satisfaction-dissatisfaction
distinction is incomplete. Such a distinction is important because it is job
dissatisfaction--rather than satisfaction--that is a major contributor to why
journalists leave the field (Weaver & Wilhoit, 1996).

Hypotheses

        If job satisfaction and dissatisfaction are different concepts, they have to be
affected by different groups of factors (except for a minority of work values
that are neutral factors, whose existence suggests that the two concepts may be
somewhat correlated[11]). Based on the rationale, we proposed the following
hypothesis:
        Hypothesis One: Newspaper reporters' job satisfaction and dissatisfaction are
affected by different groups of factors.
        According to Herzberg (Herzberg et al., 1959; Herzberg, 1968, 1987) and based
on the synthesis of past research discussed earlier, motivators or intrinsic
factors behind newspaper reporters' job satisfaction, should include autonomy,
sense of achievement, personal growth, interesting nature of work, and impact on
the community. These factors are intrinsic factors because they are more related
to job content. For example, a reporter's autonomy refers to the amount of
professional freedom he or she has in daily work (daily story assignments),
sense of achievement reflects how reporters feel they have achieved through
daily work, and personal growth is part of the experience of daily work.
Therefore, we proposed a second hypothesis:
        Hypothesis Two: Newspaper reporters' job satisfaction is predicted by
motivators, or factors related to their job content, such as autonomy, sense of
achievement, personal growth, interesting nature of work, and impact on
community.
        Autonomy, sense of achievement, personal growth, interesting nature of work,
and impact on community are directly related to the job content and are more
likely to allow the reporters to experience psychological growth. Therefore,
these factors likely enable the reporters to have higher job satisfaction. On
the other hand, pay, work conditions, and promotion opportunities are not
directly related to the content of the jobs (in the case of newspaper reporters,
the daily assignments) but more directly related to either monetary reward
(pay), work conditions, or the chance to get ahead in organization (promotion
opportunities). All of these--pay, work conditions, and promotion
opportunities--are directly related to the environment under which the job is
performed. Therefore, pay, work conditions, and promotion opportunities should
be hygiene factors, or factors affecting the reporters' job dissatisfaction.
Also, although no empirical evidence showing whether job security should be a
factor affecting job satisfaction (motivator) or dissatisfaction (hygiene
factor), job security is likely to be a hygiene factor since it is more related
to the job environment. Therefore, we proposed a third hypothesis:
        Hypothesis Three: Newspaper reporters' job dissatisfaction is predicted by
hygiene factors, or factors related to their job environment, such as pay, job
security, promotion opportunities, and work conditions.
        In addition, newsroom policy (on whether news quality or ad income should take
the priority) could be focused on either news quality or the need to earn more
advertising dollars depending on a newspaper's policy emphasis (Stamm &
Underwood, 1993). For example, if an editor decided to publish a public
relations article for a potential advertiser on a regular news page, it would
hurt news quality. On the other hand, if an editor decided not to publish such
an article, it would help news quality but may hurt potential ad income. Related
both to the job content (news quality) and job environment (ad income), newsroom
policy (on whether news quality or ad income should take the priority) could be
both a hygiene factor and motivator (i.e., a neutral factor that contributes to
job satisfaction and dissatisfaction about equally).[12] Therefore, we proposed
a fourth hypothesis:
        Hypothesis Four: Newsroom policy (on whether news quality or advertising income
should take the priority) is a predictor of both newspaper reporters' job
satisfaction and dissatisfaction.


METHODS

Population

        The population of the study is daily newspaper reporters in Texas, where the
circulation distribution pattern of newspapers is similar to that of the entire
United States (Editorial & Publisher International Yearbook, 1997).[13]
Therefore, studying Texas newspapers should not pose a significant problem for
external validity.

Approach

        We chose to conduct a survey since its application is aimed at understanding
the general patterns of a larger population from which the sample was initially
selected (Babbie, 1990). In terms of various survey methods, this study used a
mail survey because of the anonymity (so respondents may be more likely to
answer sensitive questions truthfully) and the low cost it provided (Wimmer &
Domnick, 1994).

Pretest

        A pretest was conducted on a central Texas daily newspaper in March and April
1997. The pretest questionnaire consisted of 96 items, including 70 items
revised from the widely used job descriptive index (JDI) that measures overall
job satisfaction (Ironson, Smith, Brannick, Gibson & Paul, 1989; Yeager, 1981);
22 items related to possible work values of newspaper reporters developed from a
synthesis of past research (Bergen & Weaver, 1988; Chusmir, 1984; Joseph, 1981,
1982a, 1982b, 1982c; Stamm & Underwood, 1993; Tharp, 1991; Weaver & Wilhoit,
1994, 1996), and four demographic items. The JDI was used because it measures
job satisfaction (Smith, Kendall & Hulin, 1969) with consistency and reasonably
high reliabilities across different samples (Jung, Dalessio & Johnson, 1986),
and is recommended for continual use because its validity is considered good
among available methods (Buckley, Carraher & Cote, 1992). Although the small
pretest sample size (n = 12) may render what was found to be non-generalizable
to the population of reporters, the findings provided some preliminary evidence
that newspaper reporters' job satisfaction and dissatisfaction could be
determined by different factors.[14]
        In addition, further support of Herzberg's theory was found by statistically
regressing the intrinsic dimensions of job satisfaction (satisfaction with work,
satisfaction with supervision, and satisfaction with coworkers) on the
motivators as a composite scale (scale reliability coefficient alpha = .90).[15]
The findings showed that motivators as a whole were significant predictors of
intrinsic dimensions (satisfaction with work, satisfaction with supervision and
satisfaction with coworkers)[16] of overall job satisfaction as measured by the
JDI, but were not significant predictors of the extrinsic dimensions of overall
job satisfaction (satisfaction with pay and satisfaction with promotions).[17]
Since the intrinsic dimensions of the overall job satisfaction are closely
related to job satisfaction [18], Herzberg's theory received support for the
part that posits a link between motivators and job satisfaction.
        Regression analysis on pretest data also showed that hygiene factors as a
composite scale (scale reliability coefficient alpha= .66) [19] were significant
predictors for one extrinsic dimension (satisfaction with promotions) (R square
= .33, p = .05) but not a significant predictor of another extrinsic dimension
(satisfaction with pay) and all intrinsic dimensions (satisfaction with work,
satisfaction with supervision and satisfaction with coworkers). Since the
extrinsic dimensions of overall job satisfaction are closely related to job
dissatisfaction [20], the findings that hygiene factors as a whole were a
significant predictor of one of the two extrinsic job satisfaction dimension
provides limited support for Herzberg's theory (Herzberg et al., 1959, Herzberg,
1968, 1987).
        The pretest also showed the reliability of the motivators and hygiene factors
as a scale fell within acceptable ranges.[21] Similarly, alpha reliability tests
showed the reliability for the five JDI dimensions were satisfactory.[22]
Overall then, the reliabilities of the measurement scales in the pretest are
satisfactory.
        Finally, the pretest's 50 percent response rate from three mailings was used to
estimate the minimum number of original subjects needed to get the desired
amount of valid responses for the present study. In addition, the fact that very
few missing data were found in the pretest survey showed that the questionnaire
items were clear and easy to answer.

Operational Definitions

        The main goal of the survey was to explore whether job satisfaction and
dissatisfaction are different concepts. Past research on newspaper reporters
showed those factors affecting newspaper reporters' overall job satisfaction
include the following 10 work values: autonomy, sense of achievement, impact on
community, personal growth, interesting nature of work, newsroom policy (on
whether news quality or ad income should take the priority), pay, job security,
promotion opportunities, and work conditions. The synthesis of past research on
newspaper reporters' overall job satisfaction showed the group of work values
affecting reporters' job satisfaction would include those directly related to
their job content: autonomy, sense of achievement, the work itself, impact on
the community, personal growth, and interesting nature of work. On the other
hand, the group of work values affecting the reporters' job dissatisfaction
would include work values directly related to their job environment: pay, job
security, promotion opportunities, and work conditions. In addition, the work
value of newsroom policy (on whether news quality or ad income should take the
priority) could affect satisfaction and dissatisfaction about equally.
        Basically, this study attempts to measure whether the 10 work values
traditionally found to affect newspaper reporters' job satisfaction can be
categorized primarily into two groups, each affecting either job satisfaction or
dissatisfaction. The 10 work values served as the source of questionnaire items.
To gauge the work values, 12 questionnaire items were used.[23] For example, the
item "autonomy at work" was used to gauge the work value "autonomy," and "salary
and benefits" was used to gauge the work value "pay." For each item, the
respondents were asked to respond on a 7-point Likert scale that described the
concept. For instance, for "autonomy at work," the respondents were asked to
respond on a 7-point scale ranging from "not at all" to "a great deal." For the
item "facilities and equipment in newsroom," respondents were asked to respond
on a 7-point scale ranging from "not adequate at all" to "very adequate."
        The questionnaire measured the respondents' overall job satisfaction by using
70 items revised from the widely applied 72-item job descriptive index (JDI)
(Smith et al., 1969), which has five dimensions: work, supervision, pay,
promotions and coworkers. Among the five dimensions, satisfaction with work,
satisfaction with supervision, and satisfaction with coworkers are more related
to job content, so they are likely to be intrinsic and reflect job satisfaction.
In contrast, satisfaction with pay and satisfaction with promotions are more
related to the job environment, so they are more likely to be extrinsic and
reflect job dissatisfaction. In order to more validly measure the intrinsic JDI
dimensions (reflecting job satisfaction) and the extrinsic JDI dimensions
(reflecting job dissatisfaction), the statistical procedure of structural
equation modeling was used (Buckley et al., 1992).

Hypotheses Testing

        First of all, we hypothesized that newspaper reporters' job satisfaction and
dissatisfaction are affected by different groups of factors.
        For the two concepts (satisfaction and dissatisfaction) to be affected by
different groups of factors, they should be different concepts. To test this, a
confirmatory factor analysis was used.
        After the JDI items were coded,[24] a confirmatory factor analysis was used to
determine whether job satisfaction (extracted from intrinsic dimensions) and
dissatisfaction (extracted from extrinsic dimensions) were indeed separate
concepts (factors). The procedure was employed to confirm the literature review
that indicated the intrinsic dimensions (reflecting job satisfaction) should
load on one factor (latent variable) and the extrinsic dimensions (reflecting
job dissatisfaction) should load on another.
        Secondly, we hypothesized that newspaper reporters' job satisfaction is
predicted by motivators, or factors related to job content, such as autonomy,
sense of achievement, personal growth, interesting nature of work, and impact on
community.
        Next, we hypothesized that newspaper reporters' job dissatisfaction is
predicted by hygiene factors, or factors related to job environment, such as
pay, job security, promotion opportunities, and work conditions.
        Finally, we predicted that newsroom policy is a predictor of both newspaper
reporters' job satisfaction and dissatisfaction.
        Because newsroom policy is hypothesized to affect both job satisfaction and
dissatisfaction, we tested it separately in two separate theoretical path (or
causal) models.[25] In the first theoretical path model, newsroom policy (on
whether news quality or ad income should take the priority) and motivators
(autonomy, sense of achievement, personal growth, interesting nature of work,
and impact on community) together affect job satisfaction (extracted from the
intrinsic dimensions). In the second theoretical path model, job dissatisfaction
(extracted by the extrinsic dimensions) is posited to be affected by newsroom
policy, and hygiene factors, including pay, job security, promotion
opportunities, and work conditions (reflected by questionnaire items: hours and
schedules of work, stress from working in the newsroom, and newsroom equipment
and facilities).
        To measure the two models, the work values were coded based on the
questionnaire used. We asked the survey subjects to describe, on a 7-point
Likert scale, the 12 questionnaire items [26] measuring the 10 work
values--autonomy, sense of achievement, impact on community, personal growth,
interesting nature of work, newsroom policy (on whether news quality or ad
income should take the priority), pay, job security, promotion opportunities,
and work conditions.[27]

Sampling

        Following a general rule of thumb that for each parameter estimation there
should be five subjects (Bentler & Chou, 1987), only about 70 subjects (five
times 14 for the first path model) were needed to accommodate the most
complicated path model.[28] Because we obtained a 50 percent response rate in
the pretest, the original subject pool had to be at least 140 newspaper
reporters. Although 140 reporters would seem sufficient, all available subjects
(whose names and addresses were obtained for the study) were used in the
original sample pool, because the bigger the sample size, the more stable the
structural equation model will be (Hatcher, 1998, p. 200). The original subject
pool included all 702 reporters whose names and addresses were obtained via two
methods. First, lists of reporters were compiled from all 17 daily newspapers
available in the public libraries.[29] Some 614 reporters' names were compiled
from bylines appearing in the newspapers.[30] Second, because there are 87 daily
newspapers in Texas (Editor & Publisher International Year Book, 1997), the list
of 69 other newspapers was obtained from the Editor & Publisher International
Year Book. A letter was sent to the editors of the 69 newspapers to invite them
to participate in the study. Eleven of the 69 newspapers agreed to participate
and provided lists of their reporters, for a total of 88.[31] As a result, the
original subject pool (which included the 614 reporter names obtained from
newspapers and 88 others provided by newspapers' editors) contained 702
reporters.[32]
        Except for five evening newspapers, the majority of the 28 daily newspapers
studied were morning newspapers. The circulation distribution patterns of the
newspapers were similar to those of U.S. morning newspapers.[33] For this
reason, the sample for present study seemed to represent the 686 morning
newspapers in the United States.[34] Even though the sample did not represent
all newspapers in the United States and did not represent the other 846 evening
newspapers in the country,[35] its representation of the morning papers still is
very important because morning newspapers usually are more urban and more
influential (having greater circulation sizes).[36]

Data Collection

        The survey questionnaire, cover letter, and self-addressed, stamped envelope,
was mailed to each subject in February and March 1998. With one original mailing
and one follow-up,[37] 365 valid responses were collected. Because four among
the original 702 subjects either had incorrect addresses or had quit their jobs,
the actual sample pool is 698 subjects. Based on this, the response rate exceeds
52 percent,[38] better than the 47 percent average response rate for mail
surveys (Wimmer & Domnick, 1994). After the questionnaires were collected, the
data were entered in a computer statistical program for analysis.


RESULTS

        The sample of 365 newspaper reporters was 60.5% male, 81.6% White, 50.1%
married, and 97.5% with at least a college education (82.2% were college
graduates and 15.3% had graduate degrees). The median age was 34, the average
tenure was 7 years, and the median salary ranged from $30,000 to $34,999.[39]
Compared with national samples of American newspaper journalists (Jennings,
1995; Weaver & Wilhoit, 1994), the demographics for the Texas daily newspaper
reporters are similar,[40] suggesting that the sample's external validity is
reasonably high.
        The sample was randomly assigned into two equal-sized sub-samples[41] to test
the four theoretical statements. Of the two sub-samples, one served as an
analysis sample and the other as a holdout sample, so that the three theoretical
models posited in the study could be tested twice to get better evidence about
the models' fit. Such cross-validation is recommended for models derived through
structural equation modeling (Hatcher, 1998). The cross-validation worked as
follows: For each theoretical model, the analysis sample was used to confirm or
modify the model.[42] Then, the model derived through the analysis sample was
tested in the holdout sample to see if the model still fit and if the parameter
estimates were comparable. If the model fit well in both samples and the
parameter estimates were stable for both, then there is a better chance that
such a model would fit other samples as well (Hatcher, 1998).

Hypothesis One

        We predicted that newspaper reporters' job satisfaction and dissatisfaction are
affected by different groups of factors. For the two concepts (satisfaction and
dissatisfaction) to be affected by different groups of factors, they have to be
different concepts. To test whether the two concepts are different, we used a
confirmatory factor analysis.
        But first, the reliability of each of the five JDI dimensions (work,
supervision, pay, promotions and coworkers) needed examining to ensure each was
a reliable measurement scale (since each dimension consisted of eight to 18
items). By testing the reliability of the five JDI dimensions on the analysis
sample,[43] the results showed that the reliability coefficient alpha of each
JDI dimension either approximated or exceeded the acceptable minimum level of
.80 for well-established measurement scales (Nunnally & Bernstein, 1994, p.
265), indicating that each of the five dimensions was a reliable scale.[44]
        After the five JDI dimensions were proven to be reliable measurement scales, we
used a confirmatory factor analysis to confirm that job satisfaction (extracted
from intrinsic dimensions) and dissatisfaction (extracted from extrinsic
dimensions) were indeed separate concepts. The procedure also was used to
confirm that the intrinsic dimensions load only on one factor (job satisfaction)
and the extrinsic dimensions load only on another (job dissatisfaction), as
suggested by the literature review.[45]

1. Testing the theoretical factor solution model on the analysis sample

        The analysis sample (n = 182) was first tested using the SAS CALIS program
(Hatcher, 1998)[46] to examine the fit of the confirmatory factor analysis
model. The procedure confirmed the factor model by showing it has good fit.[47]

2.  Validating the factor model in the holdout sample

        After the fit of the theoretical model was confirmed by testing it on the
analysis sample, the model was tested again on the holdout sample (n = 183) to
see if it would fit a different sample. The results of testing the model on the
holdout sample also supported the model.[48] Since the confirmatory factor
analysis model (representing the first theoretical statement) was confirmed in
both the analysis and holdout samples, the first hypothesis received ample
support. Testing the first hypothesis also revealed the correlation between job
satisfaction and dissatisfaction,[49] indicating the possible existence of
neutral factors that affect both concepts.

Hypotheses Two, Three and Four

        The second hypothesis posits that newspaper reporters' job satisfaction is
predicted by motivators, or factors related to job content, such as autonomy,
sense of achievement, personal growth, interesting nature of work, and impact on
community. The third hypothesis posits that newspaper reporters' job
dissatisfaction is predicted by hygiene factors, or factors related to their job
environment, such as pay, job security, promotion opportunities, and work
conditions. The fourth hypothesis posits that newsroom policy (on whether news
quality or ad income should take the priority) is a predictor of both newspaper
reporters' job satisfaction and dissatisfaction.
        The three hypotheses were tested in two theoretical path models. In the first
model, motivators and newsroom policy serve as independent variables, and the
latent variable of job satisfaction serves as the dependent variable. In the
second path model, hygiene factors and newsroom policy serve as independent
variables and the latent variable job dissatisfaction serves as the dependent
variable.
        Both theoretical path models were tested using a structural equation modeling
procedure executed through the SAS CALIS program (Hatcher, 1998). Each model was
first tested on the analysis sample and, after being modified, tested again on
the holdout sample for cross-validation.

1. Testing and modifying the first theoretical path model

        The model was tested first on the analysis sample (n = 182) to determine if
autonomy, sense of achievement, personal growth, interesting nature of work,
impact on community, and newsroom policy influences job satisfaction. Goodness
of fit indices for testing the model on the analysis sample exhibited mixed
results.[50]
        The mixed results in the fit indices indicated that the theoretical  model
needs modification. A look at the modification index of the GAMMA matrix
revealed that the correlation between "interesting nature of work" and
"satisfaction with work," and that between "interesting nature of work" and
"satisfaction with supervision" were the two largest unexplained correlations.
Because in path model modification it is better not to add a new path but to
eliminate an existing path (Hatcher, 1998), the best approach seemed to be
dropping the variable "interesting nature of work," a move that also appeared
justified because the path coefficient for that variable was statistically
insignificant (t = 1.2237, p = n.s.[51]).
        After "interesting nature of work" was removed from the original theoretical
model, the revised model was tested with the analysis sample. The resulting fit
indices showed evidence of a good fit for the model.[52]
        Exhibiting a good fit in the analysis sample, the revised path model was tested
again in the holdout sample (n = 183). The resulting fit indices readings all
showed signs of a good fit.[53] Since all the fit indices looked good, the
revised model was validated in the holdout sample.

2. Examining standardized path coefficients for the first path model

        Since testing the revised first path model on both samples produced two
standardized path coefficients for each independent variable, the two sets of
path coefficients had to be examined to see if they are similar in statistical
significance and whether they have positive or negative relationships with the
dependent variable.[54] In addition, standardized path coefficients' sizes would
indicate the effect of the independent variables on the dependent variable
(Hatcher, 1998).
        The revised path model (Figure 1) included five independent variables
(autonomy, sense of achievement, impact on community, personal growth, and
newsroom policy on whether news quality or ad income should take the priority)


Figure 1:       Revised Path Model 1 with Coefficients
                of Independent Variables







Note: "*" stands for a statistical significance. The path coefficients not in
the parentheses are those obtained from testing the revised model on the
analysis sample. The path coefficients in the parentheses are those obtained
from testing the model on the holdout sample.





that affect the dependent variable job satisfaction. Among the five, sense of
achievement, personal growth and newsroom policy had statistically significant
path coefficients (p < .05), whether the model was tested on the analysis sample
or on the holdout sample. Among the three independent variables with significant
path coefficients for both samples, sense of achievement had the largest path
coefficients and appeared to exert the most significant effect on job
satisfaction.
        Autonomy and impact on community, also included in the model, had significant
coefficients (p < .05) only for a single sample (instead of both samples).
        In sum, in this revised first path model, sense of achievement, personal growth
and newsroom policy were confirmed as significant predictors of job satisfaction
(dependent variable) in both samples. As hypothesized predictors of the
dependent variable, autonomy and impact on community received limited support
because each was confirmed as a significant predictor in a single sample (not
both). Except for "interesting nature of work," all four other independent
variables hypothesized in the second hypothesis (sense of achievement, personal
growth, autonomy and impact on community) received a certain degree of support
from the samples tested as predictors of the dependent variable job
satisfaction. In short, the second hypothesis was basically supported. In
addition, the finding that newsroom policy was a significant predictor of job
satisfaction only constituted partial support for the fourth hypothesis.

3. Testing and modifying the second theoretical path model

        The second theoretical path model was tested first on the analysis sample (n =
182) to determine whether pay, promotion opportunities, job security, hours and
schedule of work, job stress, newsroom facilities, and newsroom policy
influences job dissatisfaction.[55] Goodness-of-fit indices readings exhibited a
lack of a good fit [56], indicating the second theoretical path model also
needed modification.
        A look at the GAMMA matrix (obtained from testing the theoretical model on the
analysis sample) showed the correlation between the variables "promotion
opportunities" and "satisfaction with pay," and that between the variables
"promotion opportunities" and "satisfaction with promotions" were the two
largest unexplained correlations. Because path model modification usually aims
at eliminating an existing path (Hatcher, 1998), we dropped "promotion
opportunities" from the model, a move seemed justified because "promotion
opportunities" did not have a significant path coefficient (t = -.9123, p =
n.s.).
        The first revised model was tested on the analysis sample to see its goodness
of fit. The fit indices showed that the model was improved but problems still
existed.[57] The lack of an overall fit for the first revised model showed that
the model needed another modification.
        A look at the GAMMA matrix (obtained from testing the first revised model)
showed that correlation between the variables "newsroom policy" and
"satisfaction with pay," and that between the variables "newsroom policy" and
"satisfaction with promotions" were the two largest unexplained correlations.
Because path model modification aims to eliminate existing path, (Hatcher,
1998), we dropped "newsroom policy" from the model, a move seemed justified
because "newsroom policy" was almost statistically insignificant (t = -.1.7878
(one-tailed), the absolute value of which was barely larger than the t-value at
p = .05 level).
        The fit indices reading from testing second revised model on the analysis
sample showed a good fit.[58] The second revised model was then tested again on
the holdout sample (n = 183). The resulting fit indices readings also looked
good,[59] but the model suffered from a negative error variance when the latent
variable of job dissatisfaction was extracted from the observed variable of
satisfaction with pay. Since negative error variance cannot exist, a restraint
was set on the variance so that it could only be positive and the model was
tested again on the holdout sample. The resulting goodness of fit looked
good.[60] Therefore, the second revised model was confirmed for the holdout
sample.

4. Examining standardized path coefficients for second path model

        Because the second revised path model was confirmed in both the analysis sample
and holdout samples[61], it is important also to examine standardized path
coefficients for the independent variables in the revised model to see if they
are similar in their statistical significance and positive/negative signs (which
indicate positive/negative links with the dependent variable). In addition,
standardized path coefficients should be examined because their sizes indicate
the effect of the independent variables on the dependent variable (Hatcher,
1998).
        The revised path model (see Figure 2) has five independent variables  (pay, job
security, hours and schedule of work, job stress, and newsroom facilities) that
affect the dependent variable job satisfaction. Among the five, pay and job
stress had statistically significant path coefficients (p < .05) for at least
one of the two samples tested. Of the two, pay was more


Figure 2:       Revised Path Model 2 with Standardized Path Coefficients








Note: "*" stands for a statistical significance. The path coefficients not in
the parentheses are those obtained from testing the revised model on the
analysis sample. The path coefficients in the parentheses are those obtained
from testing the model on the holdout sample.





important, having path coefficients with larger absolute values. Moreover, pay,
with negative coefficients for both samples, had a negative relationship with
the dependent variable job dissatisfaction, while job stress, with positive
coefficients, had a positive relationship with job dissatisfaction. Overall,
judging by their statistical significance, job security, hours and schedule of
work, and newsroom facilities were not good predictors of job dissatisfaction.
        In sum, testing the second revised path model in the analysis sample and the
holdout sample showed pay and job stress were important predictors of job
dissatisfaction, while job security, hours and schedule of work, and newsroom
equipment were not. In addition, since the work value of work conditions was
reflected by job stress, hours and schedule of work, and newsroom facilities,
the fact that job stress was a good predictor of job dissatisfaction provided
some support that work conditions was a good predictor of job dissatisfaction.
Overall, among the four hygiene factors posited in the third hypothesis to
affect job dissatisfaction--pay, job security, promotion opportunities, and work
conditions--only pay and work conditions were good predictors of job
dissatisfaction. Therefore, the third hypothesis received mixed support. In
addition, newsroom policy--though a good predictor of job satisfaction--was not
found to be a significant predictor of job dissatisfaction. Therefore, the
fourth hypothesis also received mixed support.


DISCUSSION

        The results have several direct implications. First, knowing that job
satisfaction and dissatisfaction are distinct concepts, as Herzberg held,
suggests future researchers of job satisfaction should measure job
dissatisfaction on a different scale from job satisfaction, instead of following
the traditional path of treating both concepts as different ends of the same
continuum. The distinction between job satisfaction and dissatisfaction also is
important to newspaper managers. Treating the two concepts separately would help
managers to better identify reasons for job satisfaction to motivate reporters
to high performance, and to better identify reasons for job dissatisfaction to
help lower turnover rate. Also, the finding about the correlation between job
satisfaction and dissatisfaction supports the theoretical notion that some
factors may affect job satisfaction and dissatisfaction about equally. Such
"neutral factors" may be important because they are useful to managers. Suppose,
for example, newsroom policy (on whether news quality or ad income should take
the priority) equally contributes to reporters' job satisfaction and
dissatisfaction. A newspaper manager ignoring this likelihood may make policy
that lead to low satisfaction or high dissatisfaction.
        Second, confirming that sense of achievement, personal growth, newsroom policy,
autonomy, and impact on community are important positive predictors of job
satisfaction yields many intriguing possibilities. Interestingly, sense of
achievement had the highest path coefficient of all five predictors, meaning it
appears to have the greatest influence on job satisfaction--more influential
than autonomy and personal growth. Previous studies showed autonomy as an
important factor behind how newspaper journalists rated their jobs (Weaver &
Wilhoit, 1994), but also showed that newspaper reporters need achievement
(Chusmir, 1984) and like personal achievement most about their jobs (Tharp,
1991). As writing news stories and having them published is a short-term
process, reporters probably like to see quick results from their work. Since
personal growth often takes time, reporters may find it less desirable than the
daily sense of achievement. The finding that sense of achievement is the most
important predictor of reporter job satisfaction also is intriguing because past
national surveys on journalists repeatedly indicated autonomy to be a major
factor behind journalist job satisfaction, while overlooking sense of
achievement (Pollard, 1995; Weaver & Wilhoit, 1994, 1996). A possible
explanation may come from the belief that, as professionals, journalists should
have and value autonomy (Pollard, 1995). On the other hand, newspaper reporters
care a lot about their writing (Olson, 1989), need achievement (Chusmir, 1984),
and believe personal achievement (from writing good stories) is their strongest
like about their jobs (Tharp, 1991). Apparently, sense of achievement as a
factor behind reporters' satisfaction is worth more research attention than it
receives.
        Third, the finding that pay and work conditions (as reflected by the variable
job stress) are key predictors of job dissatisfaction is interesting because of
the different positive/negative signs of the two variables' path coefficients.
Because pay has a negative coefficient and job stress has a positive one, this
means (1) the higher the pay, the lower the job dissatisfaction and (2) the
higher the job stress, the higher the job dissatisfaction. Both findings makes
inherent sense and are consistent with earlier research (Fedler, Buhr & Taylor,
1988). In addition, the finding that pay is by far the most important predictor
also is consistent with past findings (Tharp, 1991; Weaver & Wilhoit, 1994,
1996).
        Fourth, while the study supports Herzberg's theory (that job satisfaction is
affected by motivators and job dissatisfaction is influenced by hygiene
factors), factors found to affect job satisfaction--sense of achievement,
newsroom policy, autonomy, and impact on community--also directly relate to
media content. For example, reporters have the reputation of being eager to
crusade. For some reporters, having a high sense of achievement means they need
to write such stories. Relatedly, reporters with high autonomy might also use it
to write crusading stories. Another motivator, newsroom policy, when with a
strong emphasis on news quality, also may affect media content by improving its
quality. Likewise, impact on community also could affect content because a
community-conscious reporter would write more stories catering to readers'
needs.
        Fifth--and probably most important--because job satisfaction and
dissatisfaction are different concepts, a reporter could be satisfied and
dissatisfied at the same time. A reporter dissatisfied because of, say, pay and
work conditions but satisfied because of a sense of achievement and personal
growth might still produce good-quality news stories. Taking care of reporters'
job dissatisfaction is important, but it is insufficient to motivate them to be
productive. As job satisfaction has been found to motivate better job
performance (Cranny et al., 1992), newsroom managers need to pay attention to
factors behind reporter satisfaction. This study showed that the leading
predictor of reporter satisfaction is sense of achievement, followed by (in
order of their importance) personal growth, newsroom policy, impact on
community, and autonomy. Based on the findings, newsroom managers need to pay
most attention to reporters' sense of achievement, which can motivate reporters
most effectively.
        While having several implications, the findings are also limited. One
limitation involved the measurement of the neutral factor newsroom policy.
Although it was more appropriate to use two items to measure the factor, only a
single item was used. As a result, the influence of newsroom policy on reporter
job dissatisfaction was not identified. Another limitation concerned the sample
size. Although the sizes of the two sub-samples were adequate a larger sample
could lead to the estimation of an even more stable model. Finally, the fit of
the tested models in the data does not rule out the possibility that alternative
models also may fit the data well. Because of this, the development of the
original theoretical model (for testing in the data) is crucial and should be
based on both theory and empirical evidence.
        Despite such limitations, this study may have provided stronger support for
Herzberg's theory than past studies that confirm the theory on the basis of
subjects' self-reports (Cohen, 1974; Schawartz, Jenusaitis & Stark, 1963;
Shaver, 1978), exploratory factor analysis (Knoop, 1994), or regression analysis
(Iiacqua et al., 1995). Besides providing better evidence to confirm Herzberg's
theory, another strength of the present study is that it may have marked the
first attempt to test and confirm the theory in the context of newspaper
reporters. The confirmation suggests that researchers and newsroom managers
should pay more attention to the different factors behind satisfaction and
dissatisfaction. The present research also may have greater generalizability
than other research on journalists' job satisfaction that relied on the
subjects' self-reports to determine the link between the factors and job
satisfaction (Bergen & Weaver, 1988; Weaver & Wilhoit, 1994, 1996). Since such
self-reported data constitute descriptive statistics, the models derived from
the present study using inferential statistics (i.e. structural equation
modeling) seem to stand a better chance of generalizability to the population of
newspaper reporters. What's even more interesting is that the models derived
from past research on journalists' job satisfaction are largely confirmed in
this study.
        In summary, then, "satisfaction" of reporters is more complex than
communication scholars have previously indicated. More work needs to be done to
explore the satisfaction-performance link as well as the
dissatisfaction-turnover connection. And managers need to use this new-found
complexity to explore new ways to motivate reporters through appropriate
intrinsic and extrinsic factors to achieve strategic--and happeir--ends.




Notes:

[1] Weaver and Wilhoit (1996) found in a survey that autonomy was the leading
reason cited as affecting satisfaction of newsworkers in newspapers, wire
services, news magazines, and television and radio stations. About 25 percent of
the 1,156 respondents cited autonomy as the top reason affecting their
satisfaction, rather than dissatisfaction. In a poll of U.S. and Canadian
journalists, Pollard (1995) also found staffers' autonomy related to their
overall job satisfaction. Likewise, In Weaver and Wilhoit's (1994) poll of 636
newspaper journalists, autonomy was a major reason affecting how the journalists
rated their jobs, cited by 52.9 percent of the respondents. Also, from surveys
on U.S. and Canadian newspaper journalists, Joseph (1981, 1982a, 1982b, 1982c)
found reporters wanted more decision-making power (or autonomy).

[2] In his study using journalists' occupational orientation codes created by
the U.S. Department of Labor, Chusmir (1984) concluded that by raising
reporters' sense of achievement, their overall job satisfaction would increase.
Also, in Tharp's (1991) national poll of 600 small-daily newspaper reporters,
personal achievement was cited as the reporters' strongest like (rather than
dislike) about their jobs, cited by 23 percent of the subjects. Such a result
suggests that the reporters' sense of achievement affects their job happiness
rather than their unhappiness.

[3] In Tharp's (1991) national poll on newspaper reporters, personal growth was
found as the reporters' second strongest like (rather than dislike) about their
jobs, cited by 22 percent of the subjects. Similarly, in Weaver and Wilhoit's
(1994) national poll, the chance to develop a specialty was found as a major
factor in how journalists rate their jobs, cited by 43.7 percent of the
respondents.

[4] Tharp's (1991) national poll of newspaper reporters also found the work
itself as one of the reporters' strongest likes (rather than dislikes) about
their jobs, cited by 19 percent of the subjects. Also, a national poll of
journalists showed interesting nature of work was the second most important
factor affecting the job satisfaction of newsworkers in newspapers, wire
services, news magazines, and televisions and radio stations (Weaver & Wilhoit,
1996).

[5] Chusmir (1984) concluded that newspaper reporters are the types of persons
who seek information that could influence the lives of others. Also, a national
survey of journalists showed impact on community was a major factor affecting
newsworkers' job satisfaction (rather than their dissatisfaction) (Weaver &
Wilhoit, 1996). Likewise, from a re-analysis of a national poll of 470 newspaper
journalists, Bergen and Weaver (1988) found that the newspaper's performance in
informing the public is a factor affecting the journalists' overall job
satisfaction.

[6] From their national survey, Weaver and Wilhoit (1996) found salary is one of
the two most important factors affecting job dissatisfaction (rather than
satisfaction) of newsworkers in newspapers, wire services, news magazines,
televisions and radios, cited by 50 percent of the respondents (Weaver &
Wilhoit, 1996). Also, Weaver and Wilhoit (1994) found that pay is a major factor
affecting how newspaper journalists rate their jobs, cited by 21.4 percent of
the subjects. Likewise, Tharp (1991) found in a poll of reporters that one of
the reporters' top two dislikes (rather than likes) about their jobs is salary,
cited by 28 percent of the respondents.

[7] In his national survey of small newspaper reporters, Tharp (1991) found
"work load" and "work conditions" to be one of the two strongest dislikes
(rather than likes) of the reporters about their jobs, cited by 28 percent of
the respondents. Also in the same survey, internal pressure was found to be a
major reason the reporters dislike their jobs, cited by 21 percent of the
respondents. On the other hand, Weaver & Wilhoit's (1996) national survey found
that work conditions is a weak factor affecting the job satisfaction journalists
of all types, cited by only 5 percent of the respondents.

[8] Promotion opportunities were found in a national survey to be a major
factor affecting the job dissatisfaction (rather than satisfaction) of
journalists in newspapers, wire services, news magazines, televisions and
radios, cited by about one-fifth of the 1,156 respondents (Weaver & Wilhoit,
1996). Another national survey also found promotion opportunities to be a major
factor affecting how newspaper journalists rate their jobs, cited by 35.5
percent of the respondents (Weaver & Wilhoit, 1994).

[9] Judging from past research, it is unclear whether job security affects
newspaper reporters' satisfaction or dissatisfaction. From a 1992 poll on
newspaper journalists, job security was found to be a key factor affecting how
newspaper journalists rate their jobs, cited by 62 percent of the respondents
(Weaver & Wilhoit, 1994). However, another recent national poll showed that job
security is a weak factor affecting job satisfaction of newsworkers in
newspaper, wire services, news magazines, and television and radio stations,
cited by only 5 percent of the respondents (Weaver & Wilhoit, 1996). Although it
is unclear whether job security affects the reporters' satisfaction or
dissatisfaction, it is clearly an extrinsic work value--as it is related more to
the work environment (rather than the job content). According to Herzberg's
theory, such a work value should influence the reporters' job dissatisfaction
(Herzberg et al. 1959; Herzberg, 1968, 1987).

[10] Weaver and Wilhoit (1996) found management policy is one of the two most
important factors affecting job dissatisfaction (rather than satisfaction) of
newsworkers in newspapers, wire services, news magazines, and television and
radio stations, cited by 50 percent of the respondents. Bergen and Weaver (1988)
found that the newsroom policy is a major factor affecting newspaper
journalists' overall job satisfaction. Weaver and Wilhoit (1994) also found that
editorial policy is the most important factor affecting how newspaper
journalists rate their jobs, cited by 66.8 percent of the respondents. In
addition, from a poll of 429 newspaper journalists, Stamm and Underwood (1993)
found that journalists' overall job satisfaction rises when the policy results
in news quality improvement. Their overall satisfaction declines (i.e., their
dissatisfaction will increase) if, after a policy change, professional autonomy
takes a back seat to business concerns. In other words, newsroom policy may
contribute to the journalists' job satisfaction if it improves quality but may
contribute to job dissatisfaction if it emphasizes business concerns. These
studies suggest that newsroom policy is likely to affect newspaper reporters'
job satisfaction and dissatisfaction about equally. When newsroom policy focuses
on news quality, it relates to job content (daily work) and should affect job
satisfaction. On the other hand, when newsroom policy favors advertising
revenues, it relates more to the environment under which the job is carried out,
and should affect job dissatisfaction.

[11] Because the possible existence of neutral work values that affect the two
concepts (job satisfaction and dissatisfaction) about equally, the two are,
however, somewhat correlated.

[12] Newsroom policy (on whether news quality or ad income should take the
priority) is related both to the job content (news quality) and to job
environment (ad income) because (1) newsroom policy may focus on news quality,
which is related to the reporters' daily assignments (job content), and (2)
newsroom policy also may emphasize advertising revenue, which is more related to
newspaper profitability (job environment) than to the daily assignments (job
content). Therefore, it can be both a motivator and hygiene factor.

[13] According to Editor & Publisher International Year Book (1997), the
circulation makeup of Texas newspapers is similar to that of the United States.
For example, about 0.5 percent of U.S. daily newspapers belong to the
circulation category of more than 500,000, similar to the 2.3 percent ratio in
Texas. For the 250,001-500,000 circulation category, the similarity is
striking--2.1 percent of U.S. dailies fall under the category, compared with 2.3
percent of dailies in Texas.

[14] According to the self-reports by pretest respondents, factors affecting the
reporters' job satisfaction, or "motivators," were primarily work-related (e.g.,
sense of achievement, interesting nature of job, the chance to learn a great
deal, and the chance to help readers). In contrast, factors affecting the
reporters' job dissatisfaction, or "hygiene factors," were primarily related to
the job environment (e.g., newsroom facilities, job security, and pay). In
addition, newsroom policy (on whether news quality or ad income should take the
priority) was found to affect job satisfaction and dissatisfaction about
equally.

[15] According to Nunnally and Bernstein (1994), the reliability coefficient
alpha reading was better than expected since the reliability for a hypothesized
measure of a construct (in this case, the scale is composed by motivators) is
considered satisfactory if the coefficient is .70 or higher.

[16] Among the five JDI dimensions (satisfaction with work, satisfaction with
supervision, satisfaction with pay, satisfaction with promotions, and
satisfaction with coworkers), satisfaction with work, satisfaction with
supervision and satisfaction with coworkers contain items most directly related
to daily work. For example, almost every item under the dimension of
satisfaction with work (such as "fascinating," "endless," "routine," and
"challenging") is directly related to daily work. Likewise, most items under the
dimension of satisfaction with supervision are also directly related to daily
work (such as "tells me where I stand," "up-to-date," "knows my job well," and
"praises good work"). Similarly, the majority of the items under the dimension
of satisfaction with coworkers (such as "stimulating," "responsible," "fast,"
and "loyal") are also daily work-related. Because these dimensions relate
closely to daily work, they are considered intrinsic dimensions of overall job
satisfaction and, according to Herzberg's theory (Herzberg et al., 1959,
Herzberg' 1968, 1987), should be closely related to job satisfaction.

[17] Among the five JDI dimensions, satisfaction with pay and satisfaction with
promotions are not related to daily work, but rather the work environment. For
example, items under satisfaction with pay are directly related to the financial
rewards given in the work environment (such as "underpaid," "highly paid,"
"adequate for normal expenses," and "barely live on income"). Similarly, items
under satisfaction with promotions are related to the reward of higher job
status found in the job environment (such as "good opportunities for
advancement," "dead-end job," "good chance for promotion," and "opportunities
somewhat limited"). Because these dimensions relate closely to the work
environment, they are considered extrinsic dimensions of overall job
satisfaction and, according to Herzberg's theory (Herzerg et al. 1959; Herzberg,
1968, 1987), should be closely related to job dissatisfaction.

[18] Among the five JDI dimensions (satisfaction with work, satisfaction with
supervison, satisfaction with pay, satisfaction with promotions, and
satisfaction with coworkers), satisfaction with work, satisfaction with
supervision and satisfaction with coworkers contain items most directly related
to the daily work. For example, almost every item under the dimension of
satisfaction with work (such as "fascinating," "endless," "routine," and
"challenging") is directly related to the daily work. Likewise, most items under
the dimension of satisfaction with supervision are also directly related to the
daily work (such as "tells me where I stand," "up-to-date," "knows my job well,"
and "praises good work"). Similarly, the majority of the items under the
dimension of satisfaction with coworkers (such as "stimulating," "responsible,"
"fast," and "loyal") are also daily work-related. Because these dimensions
relate closely to the daily work, they are considered intrinsic dimensions of
the overall job satisfaction and, according to Herzberg's theory (Herzberg et
al., 1959; Herzberg, 1968, 1987), should be closely related to job satisfaction.

[19] According to Nunnally and Bernstein (1994), the reliability coefficient
alpha reading was roughly appropriate since the reliability for a hypothesized
measure of a construct (in this case, the scale was composed by hygiene factors)
is considered satisfactory if the coefficient is .70 or higher.

[20] Among the five JDI dimensions, satisfaction with pay and satisfaction with
promotions are not related with daily work, but rather the work environment. For
example, items under satisfaction with pay are directly related to the financial
rewards given in the work environment (such as "underpaid," "highly paid,"
"adequate for normal expenses," and "barely live on income"). Similarly, items
under satisfaction with promotions are related  to the reward of higher job
status found in the job environment (such as "good opportunities for
advancement," "dead-end job," "good chance for promotion," and "opportunities
somewhat limited"). Because these dimensions relate closely to the work
environment, they are considered extrinsic dimensions of the overall job
satisfaction and, according to Herzberg's theory (1959, 1968, 1987), should be
closely related to job dissatisfaction.

[21] The realiability alpha readings were .90 for motivators as a whole and .66
for hygiene factors as a whole. The readings were acceptable because they wer
either higher than the .70 acceptable level or close to that level. The
reliability for a hypothesized measure of a construct is considered satisfactory
if the coefficient is .70 or higher (Nunnally & Bernstein, 1994).

[22] The reliability coefficient alpha readings were .90 for satisfaction with
work, .95 for satisfaction with supervision, .76 for satisfaction with pay, .86
for satisfaction with promotions, and .92 for satisfaction with coworkers.
According to Nunnally and Bernstein (1994), it can be argued that for
well-established measurement scales, increasing reliability much beyond a
reliability coefficient alpha reading of .80 is often unnecessary.

[23] The 12 questionnaire items (or the 12 variables used in the statistical
analysis of the data collected from the present study) include autonomy at work,
salary and benefits, job security, feeling of accomplishment from work, the
extent to which my job is interesting, chance of advancement in the paper,
opportunities to influence readers, opportunities to learn a great deal, hours
and schedule of work, stress from working in newsroom, newsroom policy,
facilities and equipment in newsroom, and the extent to which my job is
challenging. The 12 items with the work values they represent in the parentheses
are as follows: autonomy at work (autonomy); salary and benefits (pay); job
security (job security); feeling of accomplishment in the paper (sense of
achievement); the extent to which my job is interesting (interesting nature of
work); chance of advancement (promotion opportunities); opportunities to
influence readers (impact on the community); opportunities to learn a great deal
(personal growth); newsroom policy (newsroom policy); hours and schedule of work
(work conditions); facilities and equipment in newsroom (work conditions);
stress from working in newsroom (work conditions).

[24] For each item, a score of 3 was assigned to a "yes" response to a positive
item (such as "fascinating" under the dimension of "satisfaction with work") and
to a "no" to a negative item (e.g., "less than I deserve" under the dimension of
"satisfaction with pay"). In other words, a score of 3 to any single item
indicated "satisfaction" (for items in the intrinsic JDI dimensions) or "lack of
dissatisfaction" (for items in the extrinsic JDI dimensions). On the other hand,
a score of 0 was assigned to a "no" response to a positive item (e.g., "ask my
advice" under "satisfaction with supervision") or to a "yes" to a negative item
(e.g., "boring" under "satisfaction with coworkers"). Therefore, a score of 0
indicated "lack of satisfaction" (for items in the intrinsic dimensions) or
"dissatisfaction" (for items in the extrinsic dimensions). In addition, a score
of 1 was assigned to an "unsure" response to any item. The scoring scheme
followed empirically tested guidelines proposed by Smith, Kendall and Hulin
(1969). For the present study, an unsure response to an intrinsic-dimension item
is more indicative of "lack of satisfaction" than "satisfaction." Similarly, an
unsure response to an extrinsic-dimension item is more indicative of
"dissatisfaction" than "lack of dissatisfaction."

[25] Path models may be referred to as causal models (Hatcher, 1998). The models
will be tested using path analysis under the statistical procedure of structural
equation modeling, which would improve the validity of the JDI items (Buckley et
al., 1992). Another advantage to using structural equation modeling is that the
procedure tests all paths (relationships between variables) simultaneously and
estimates the errors at the same time (Schumacker & Lomax, 1996) (source:
Schumacker, R. E., & Lomax, R. G. (1996). A Beginner's Guide to Structural
Equation Modeling. Mahwah, NJ: Lawrence Erlbaum Associates, Publishers).

[26] The 12 questionnaire items (or the 12 variables used in the analysis of the
data collected from the present study) are autonomy at work, salary and
benefits, job security, feeling of accomplishment from work, the extent to which
my job is interesting, chance of advancement in the paper, opportunities to
influence readers, opportunities to learn a great deal, hours and schedule of
work, stress from working in newsroom, newsroom policy, facilities and equipment
in newsroom, and the extent to which my job is challenging. The 12 items with
the work values they represent in the parentheses are as follows: autonomy at
work (autonomy); salary and benefits (pay); job security (job security); feeling
of accomplishment in the paper (sense of achievement); the extent to which my
job is interesting (interesting nature of work); chance of advancement
(promotion opportunities); opportunities to influence readers (impact on the
community); opportunities to learn a great deal (personal growth); newsroom
policy (newsroom policy); hours and schedule of work (work conditions);
facilities and equipment in newsroom (work conditions); and stress from working
in newsroom (work conditions).
After the data were collected, each of the 12 items was assigned a value.
Specifically, each response was assigned a value ranging from 0 to 6. The seven
continuous integers were chosen to correspond to the original 7-point scale.
Except for questionnaire item of newsroom policy, the value of 0 was used to
indicate zero salience of a specific questionnaire item (e.g., autonomy at
work). In contrast, the value of 6 was used to indicate full-blown salience of a
questionnaire item (e.g., job security). For the questionnaire item of newsroom
policy, the value of 0 was used to indicate great policy emphasis on business
concerns while the value of 6 was used to indicate great policy emphasis on news
quality.

[27] Single items (instead of multiple items) were used to measure most of the
10 work values. There are two reasons for such treatment:
(1) Single items have been used to gauge factors behind job satisfaction in past
survey research on newspaper journalists' job satisfaction (Bergen & Weaver,
1988; Weaver & Wilhoit, 1996);
(2) Using multiple items would make the number of questionnaire items too large
(the first two parts of the questionnaire already contains 82 items, see
appendices 1 and 2), thus reducing the response rate.
For more information on the questionnaire design, contact the authors.

[28] The sample used here is related to the number of parameters estimated in
each of the three models (one confirmatory factor analysis model and two path
analysis models) (Bentler & Chou, 1987). In the confirmatory factor analysis,
there are five observed variables (five JDI dimensions), two latent variables
(or factors), five error terms (for extracting factors from each observed
variable), and one correlation (see endnote No. 11) between the two latent
variables (or factors), a total of 13 parameters to estimate. For the first path
model, there are nine observed variables (five motivators, newsroom policy, and
three intrinsic JDI dimensions), one latent variable and four error terms (three
errors from extracting a factor from three observed variables--satisfaction with
work, satisfaction with supervision and satisfaction with coworkers--and one
error from the path equation using six observed variables as predictors of the
latent variable), a total of 14 parameters to be estimated. Following the same
rationale as that for the first path model, the second path model requires 13
parameters to be estimated.

[29] To protect the anonymity of the survey respondents as promised in the cover
letter, the names of the 17 newspapers found in the public libraries are not
disclosed here.

[30] All reporters' names were collected from a week's worth of newspapers for
each newspaper to ensure that almost every reporter's name for each newspaper
was collected. The time of name collection was in late December 1997 and January
1998, just before the survey started. The timing was intended to ensure that the
reporters' names collected from the newspapers were current.

[31] To protect the anonymity of survey respondents, the names of the 11
newspapers whose editors agreed to participate in the present study are not
revealed here.

[32] This study merges reporters obtained from the two sampling methods into one
sample because their demographics (as obtained from those responding to the
survey) were mostly comparable (except for median income). Some 365 subjects
completed every item of the survey. Among the 365 subjects, 320 were collected
through the first sampling method while 45 were collected through the 2nd
sampling method.

[33] According to Editor & Publisher International Year Book (1997), circulation
distribution patterns of the sampled 28 Texas daily newspapers for the present
study are similar to that of all 686 morning newspapers in the United States.
For example, 7.1 percent of the dailies in the sample belong to the circulation
category of 250,001-500,000, compared to the 4.7 percent ratio for the U.S.
morning papers. A striking similarity occurs at the 50,001-100,000 circulation
category, where the difference between the ratio for the sample and that for the
U.S. morning papers is only 0.4 of a percentage point. Except for the
circulation categories of 10,001-25,000 and 25,001-50,000 (where the difference
in the ratio is at most 7.6 percentage points), the difference in the ratios for
the sample and U.S. morning newspapers is less than 5 percentage points.

[34] According to the Editor & Publisher International Year Book (1997), there
are a total of 686 morning newspapers in the United States.

[35] According to the Editor & Publisher International Year Book (1997), there
are 1,520 daily newspapers (including both morning and evening dailies) and 846
evening newspapers in the United States.

[36] According to the Editor & Publisher International Year Book (1997), morning
papers tend to be more urban than afternoon papers. Morning papers concentrate
in areas with larger populations, while afternoon papers concentrate in areas
with smaller populations. Also, according to the year book, morning papers
usually have larger circulation sizes than afternoon papers.

[37] Each of the mailed questionnaires left out the subject's name, title, or
name of the newspaper where he or she worked. The only identification on the
questionnaire was a number, which was used only for identifying non-responding
subjects for a follow-up mail only.

[38] If responses with missing data are included, there were 448 responses,
making the response rate 64 percent.

[39] For more informationon on gender, ethnicity, marital status, educational
backgrounds, and income levels, please contact the authors.

[40] The demographics of the Texas reporters sampled in this study are very
comparable to national samples of American newspaper journalists in terms of
gender, ethnicity, education, age, and salary, as the following table shows:

----------------------------------------------------------------------
                Demographic             National        Texas
                categories              sample          sample
----------------------------------------------------------------------
Gender          Male                    66.1%           60.5%
                Female                  33.9%           39.5%

Ethnicity       White                   89.1%*          81.6%
                Non-white               10.1%*          18.4%

Education       College graduates**     84.3%           97.5%
                Non-college graduates   15.7%           2.5%

Others          Median age              37              34
                Median salary           35,000          30,000-34,999
----------------------------------------------------------------------
* Except for the figures for ethnicity, statistics in all other demographic
categories for the national sample come from Weaver and Wilhoit's 1992 national
survey (Weaver & Wilhoit, 1994). The figures for ethnicity for the national
sample comes from a 1994 survey conducted by the American Society of Newspaper
Editors (Jennings, 1995).
** College graduates refer to those with at least a bachelor's degree.

[41] One sub-sample contains 182 subjects while the other contains 183 subjects.

[42] Because Herzberg's theory had not been tested on newspaper reporters and
the theoretical models tested for the present study were not yet a
well-developed model, we modified the hypothesized theoretical models to derive
a better model (based on the analysis sample). The model modification, if there
is any, will be further justified through testing the revised model on the
holdout sample.

[43] The reliability analysis was only performed on the analysis sample because
it was the sample where the reliability of the measurement should be established
before the start of statistical analysis. The second sample served as holdout
sample merely to confirm or disconfirm the model derived from the analysis
sample but not used to estimate the reliability of the measurement scales.

[44] The readings for reliability coefficients alpha were .81 for the intrinsic
dimension of satisfaction with work, .88 for the intrinsic dimension of
satisfaction with supervision, .86 for the intrinsic dimensions of satisfaction
with coworkers, .79 for the extrinsic dimensions of satisfaction with pay, and
.90 for the extrinsic dimensions with promotions.

[45] The confirmatory factor analysis requires the researcher to pre-determine
which factor loadings are zero (McDonald, 1985). Because a variable will only
have zero loading on a factor if the variable is totally not loaded on that
factor (McDonald, 1985), it is clear that "satisfaction with work,"
"satisfaction with supervision," and "satisfaction with coworkers" were
hypothesized to load only on Factor 1 (job satisfaction) and to have zero
loadings on Factor 2 (job dissatisfaction) (source: McDonald, R. P. (1985).
Factor analysis and related methods. Hillsdale, N.J. : Lawrence Erlbaum
Associates). In contrast, "satisfaction with pay" and "satisfaction with
promotions" were hypothesized to load only on Factor 2 and to have zero loadings
on Factor 1.

[46] The CALIS program of the SAS package is designed to perform confirmatory
factor analysis and path analysis, both structural equation modeling techniques
(Hatcher, 1998).

[47] The fit index readings were good for both the analysis sample (n = 182) and
the holdout sample (n = 183), showing that the model fits both samples. The GFI,
AGFI and NFI readings were all above the .90 required fit level for good fit
(Schumacker & Lomax, 1996) (source: Schumacker, R. E., & Lomax, R. G. (1996). A
Beginner's Guide to Structural Equation Modeling. Mahwah, NJ: Lawrence Erlbaum
Associates, Publishers). In addition, the Chi-square statistics were
insignificant for both samples, another sign of good fit because these figures
show the model built from the data is not different from the theoretical model
(Schumacker & Lomax, 1996). Also, the average absolute residuals from testing
the two samples were both small, indicating the success of the factor solutions
(McDonald, 1985) (source: McDonald, R. P. (1985). Factor analysis and related
methods. Hillsdale, N.J. : Lawrence Erlbaum Associates). The following table
shows the fit criteria readings:

Table: Fit Criteria Readings of Confirmatory Factor Analysis
----------------------------------------------------------------
                                        Analysis        Holdout
Fit criteria                            sample          sample
----------------------------------------------------------------
Goodness of Fit Index (GFI)             .9878           .9967
GFI Adjusted for Degree of Freedom(AGFI).9544           .9875
Normed Fit Index (NFI)                  .9611           .9835
Chi-square probability                  .2120           .8158
Average Absolute Residual               .02005          .01172
----------------------------------------------------------------

[48] See endnote No. 47.

[49] The factor-loading structures obtained from both samples clearly indicated
that "satisfaction with work," "satisfaction with supervision," and
"satisfaction with coworkers" loaded only on Factor 1 (job satisfaction) while
"satisfaction with pay" and "satisfaction with promotions" loaded only on Factor
2 (job dissatisfaction). The following table shows the factor loading structures
obtained from analysis and holdout samples:

Table: Factor-Loading Structure (Analysis Sample vs. Holdout Sample)
-----------------------------------------------------------------
                        Factor loadings
-----------------------------------------------------------------
        Observed                Factor 1        Factor 2
        variables               (Job            (Job
                                satisfaction)   dissatisfaction)
-----------------------------------------------------------------
Satisfaction with work          .7001 (.6918)   0 (0)
Satisfaction with supervision   .7253 (.6038)   0 (0)
Satisfaction with coworkers     .3617 (.4501)   0 (0)
Satisfaction with pay           0 (0)           .2775 (.3011)
Satisfaction with promotions    0 (0)           .8893 (.5474)
-----------------------------------------------------------------
Correlation between factors: .7272 (analysis sample);
                             .8236 (holdout sample)

Note: n = 182 (analysis sample); n = 183 (holdout sample). Loadings not in the
parentheses were loadings obtained from testing the model in the analysis
sample, while loadings in the parentheses were obtained from testing the model
on the holdout sample.

[50] The fit indices showed mixed results for testing the first theoretical
model on the analysis sample. Although the GFI, AGFI and NFI readings are either
higher than or nearing the .90 required level of good fit (Schumacker & Lomax,
1996), a significant Chi-square showed the model's a poor fit (Schumacker &
Lomax, 1996) (source: Schumacker, R. E., & Lomax, R. G. (1996). A Beginner's
Guide to Structural Equation Modeling. Mahwah, NJ: Lawrence Erlbaum Associates,
Publishers). The fit indices revealed that the revised model is better than the
original theoretical model. Compared with the GFI, AGFI and NFI readings from
testing the theoretical model, the readings from testing the revised model on
both the analysis and holdout samples were higher and all exceeded the .90
minimum level of good fit. Also, while the Chi-square probability level from
testing the theoretical model was significant, the probability levels from
testing the modified model on both samples were insignificant. The insignificant
Chi-square indicated the model estimated from the data was not significantly
different from the revised model. The following table shows the results of
testing the theoretical model and the revised model:

Table:  Fit Criteria Readings (Theoretical Model 1 vs. Revised Model)
-------------------------------------------------------------------
Model       |   Sample     |    GFI     AGFI    NFI     Chi-square
            |              |                            (p)
-------------------------------------------------------------------
Theoretical |   Analysis   |
model       |   sample     |    .9656   .8708   .9428   .0028
--------------------------------------------------------------------
            |   Analysis   |
Revised     |   sample     |    .9781   .9213   .9527   .0876
model       |-------------------------------------------------------
            |   Holdout    |
            |   sample     |    .9811   .9318   .9506   .1245
-------------------------------------------------------------------

Note: n = 182 (analysis sample); n = 183 (holdout sample). GFI = Goodness of Fit
Index; AGFI = GFI Adjusted for Degrees of Freedom; NFI = Normed-Fit Index;
Chi-square (p) = Chi-square probability level.

[51] "n.s." stands for a statistically insignificant probability level.

[52] See endnote No. 50.

[53] See endnote No. 50.

[54] Since path coefficients represent the size of the effect a given
independent variable has on a dependent variable (Hatcher, 1998), the
negative/positive sign in front of the coefficient indicates the
positive/negative relationship the independent variable has with the dependent
variable.

[55] When the latent variable "job dissatisfaction" was extracted from the
manifest variables of "satisfaction with pay" and "satisfaction with
promotions," the relationships between the latent variable and manifest
variables were prescribed as negative in the second path model. For example, the
factor (latent variable) extraction part of the path model looked as follows:

V1 = - L1 F1 + E1
V2 = - L2 F1 + E2
where
V1 = satisfaction with pay;
V2 = satisfaction with promotions;
F1 = job dissatisfaction;
L1, L2 = factor loadings;
E1, E2 = error terms from factor extraction.
(Note: The negative signs before L1 and L2 indicate negative relationships
between F1 and V1 and between F1 and V2.)

The prescription of negative relationships was intended to recode the latent
variable so that it represented "dissatisfaction," rather than "satisfaction."
The original manifest (observed) variables ("satisfaction with pay" and
"satisfaction with promotions") were composite variables formed by items scored
on a 0-to-3 scale, with 0 indicating "dissatisfaction" and 3 indicating "lack of
dissatisfaction." If the latent variable was to represent job dissatisfaction,
the original manifest variables needed to be recoded so that 0 would represent
"lack of dissatisfaction" and 3 would stand for "dissatisfaction." Instead of
recoding each of the 17 items constituting the two original composite (manifest)
variables of "satisfaction with pay" and "satisfaction with promotions," the
author simply prescribed negative relationships between the manifest variables
and the latent variables (of "job dissatisfaction"). Such prescription of
negative relationship is equivalent to recoding the manifest variables before
extracting job dissatisfaction from them.

[56] The fit indices readings showed that the original second theoretical path
model lacked a good fit. All of the GFI, AGFI and NFI readings were lower than
the required .90 level of good fit (Schumacker & Lomax, 1996), while a
significant Chi-square also indicated a lack of fit (Schumacker & Lomax, 1996)
(source: Schumacker, R. E., & Lomax, R. G. (1996). A Beginner's Guide to
Structural Equation Modeling. Mahwah, NJ: Lawrence Erlbaum Associates,
Publishers). The first revised model showed an improvement of the fit, with the
GFI and NFI exceeding .90 level, but the AGFI reading was still low and the
Chi-square was still significant. Unlike the original model and the first
revised model, the second revised model showed evidence of a good fit. The NFI,
AGFI and NFI readings from testing the second revised model on both samples all
exceeded the .90 level, and the Chi-square probability level was insignificant.
The insignificant Chi-square indicated that the model estimated from the data is
not significantly different from the seond revised model. The following table
shows the results from testing the models:

Table:  Fit Criteria Readings (Theoretical Model 2 vs. Revised Models)
---------------------------------------------------------------------
Model         | Sample     |    GFI     AGFI    NFI     Chi-square
              |            |                            (p)
---------------------------------------------------------------------
Theoretical   | Analysis   |
model         | sample     |    .8905   .1788   .7315   .0001
---------------------------------------------------------------------
First revised | Analysis   |
model         | sample     |    .9766   .8313   .9512   .0026
---------------------------------------------------------------------
              | Analysis   |
Second        | sample     |    .9880   .9157   .9775   .0955
revised       |------------------------------------------------------
sample        | Holdout    |
              | sample     |    .9911   .9503   .9746   .3266
---------------------------------------------------------------------

Note: n = 182 (analysis sample); n = 183 (holdout sample, tested with an error
restraint set on the extraction of "job satisfaction with pay" to generate the
latent variable of job dissatisfaction). GFI = Goodness of Fit Index; AGFI = GFI
Adjusted for Degrees of Freedom; NFI = Normed-Fit Index; Chi-square (p) =
Chi-square probability level.

[57] See endnote No. 56.

[58] See endnote No. 56.

[59] The fit indices readings were .9929 for GFI, .9503 for AGFI and .9798 for
NFI, all exceeding the .90 level required for a good fit (Schumacker & Lomax,
1996). The Chi-square was insignificant, adding more evidence for the fit of the
model (Schumacker & Lomax, 1996) (source: Schumacker, R. E., & Lomax, R. G.
(1996). A Beginner's Guide to Structural Equation Modeling. Mahwah, NJ: Lawrence
Erlbaum Associates, Publishers).

[60] See endnote No. 56.

[61] The path coefficients for the holdout sample reported here were those
obtained with error variance restraint for the extraction of the observed
variable "satisfaction with pay" (to form latent variable of job
dissatisfaction). The restraint is used to ensure that the error variance is
positive, since a negative error restraint is not possible.



References:

Alderfer, C. P. (1972). Existence, relatedness, and growth. New York: Free
Press.

Babbie, E. (1990). Survey Research Methods (2nd ed.). Belmont, C.A.: Wadsworth
Publishing Company.

Bentler, P.M., & Chou, C.P. (1987). Practical issues in structural modeling.
Sociological Methods and Research, Vol. 16, No. 1, p. 78-117.

Bergen, L. A., & Weaver, D. (1988). Job satisfaction of daily newspaper
journalists and organization size. Newspaper Research Journal, Vol. 9, No. 2, p.
1-13.

Bockman, V. M. (1971). The Herzberg Controversy. Personnel Psychology, Vol. 24,
p. 155-89.

Buckley, E. R. , Carraher, S. M., & Cote, J. A. (1992). Measurement issues
concerning the use of inventories of job satisfaction. Educational and
Psychological Measurement, Vol. 52, No. 3, p. 529-543.

Caston, R. J., & Braito, R. (1985). The worker-to-job 'fit' hypothesis: Some
further evidence. Work and Occupations, Vol. 12, No. 3, p. 269-84.

Chusmir, L. H. (1984). Profile of motivational needs of individuals in the
newspaper industry. Newspaper Research Journal, Vol. 5, No. 1, p. 33-41.

Cohen, A. M. (1974). Community college faculty job satisfaction. Research in
Higher Education, Vol. 2, No. 4, p. 369-376.

Cranny, C. J., Smith, P. C., & Stone, E. F. (1992). Job Satisfaction: How People
Feel About Their Jobs and How It Affects Their Performance. New York: Lexington
Books.

Dizard, W. (1994). Old Media/New Media: Mass Communications in the Information
Age. White Plains, N.Y.: Longman.

Editor & Publisher International Year Book (1997). Anderson, I. E. (Eds.). 77th
edition, part 1. New York: Editor & Publisher.

Fedler, F., Buhr, T., & Taylor, D. (1988). Journalists who leave the news media
seem happier, find better jobs. Newspaper Research Journal, Vol. 9, No. 2, p.
15-23.

Hatcher, L. (1998). A Step-by-Step Approach to Using the SAS System for Factor
Analysis and Structural Equation Modeling. Cary, N.C.: SAS Institute Inc.

Herzberg, F., Mausner, B. & Snyderman, B. B. (1959). The Motivation to Work, New
York: Wiley.

Herzberg, F. (1968). One more time: How do you motivate employees. Harvard
Business Review, Vol. 46, No. 1, p. 53-62.

Herzberg, F. (1987). One more time: How do you motivate employees. Harvard
Business Review, Vol. 65, No. 5, p. 109-117.

Igbaria, M., & Greenhaus, J. H. (1992). Determinants of MIS employees' turnover
intentions: A structural equation model. Communications of the ACM, Vol. 35, No.
2, p. 35-49.

Iiacqua, J. A., Schumacher, P., & Li, H. C. (1995). Factors contributing to job
satisfaction in higher education. Education, 116(1), p. 51-61.

Ironson, G. H., Smith, P. C., Brannick, M. T., Gibson, W. M., & Paul, K. B.
(1989). Construction of a Job in General Scale: A Comparison of Global,
Composite, and Specific Measures. Journal of Applied Psychology, Vol. 74, Vol.
2, p. 193-200.

Jennings, V. (1995). Minorities are now 10.9 percent of the newsroom work force
in the U.S. ASNE Bulletin, No. 767, p. 22-23.

Joseph, T. (1981). Existing decision-making practices on American dailies.
Newspaper Research Journal, Vol. 2, No. 4, p. 56-60.

Joseph, T. (1982a). Daily newspaper reporters' and editors' preferences toward
reporter decision-making and existing practices. Human Relations, Vol. 35, No.
7, p. 539-545.

Joseph, T. (1982b). Reporters' and editors' preferences toward reporter decision
making. Journalism Quarterly, Vol. 59, No. 2, p. 219-222 & 248.

Joseph, T. (1982c). Decision-making preferences and practices. Newspaper
Research Journal, Vol. 3, No. 3, p. 37-44.

Jung, K. G. , Dalessio, A., & Johnson, S. M. (1986). Stability of the factor
structure of the Job Descriptive Index. Academy of Management Journal, Vol. 29,
p. 609-616.

Knoop, R. (1994). Work values and job satisfaction. Journal of Psychology, Vol.
128, No. 6, p. 683-690.

Lacy, S., Sohn, A. B., & Wicks, J. L. (1993). Media Management: A Casebook
Approach. Hillsdale, N.J.: Lawrence Erlbaum Associates, Publishers.

Likert, R. (1961). New Patterns of Management. New York: McGraw-Hill Book Co.,
Inc.

Likert, R. (1967). The Human Organization: Its Management and Value. New York:
McGraw-Hill Book Co., Inc.

Maslow, A. H. (1954). Motivation and personality. New York: Harper & Row.

Moorhead, G., & Griffin, R. W. (1989). Organizational Behavior. Boston: Houghton
Mifflin Co.

Moss, M. (1978). Reporter turnover on Texas daily newspapers. Journalism
Quarterly, Vol. 55, No. 2, p. 354-356.

Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric Theory. New York:
McGraw-Hill, Inc., p. 265.

Olson, L. D. (1989). Job satisfaction of journalists and PR personnel. Public
Relations Review, Vol. 15, No. 4, p. 37-45.

Phillipchuk, J., & Whittaker, J. (1996). An inquiry into the continuing
relevance of Herzberg's motivation theory. Engineering Management Journal, 8(1),
p. 15-20.

Pinder, C. C. (1984). Work Motivation: Theory, Issues and Application. Glenview,
IL: Scott, Foresman.

Pollard, G. (1995). Job satisfaction among newsworkers: the influence of
professionalism, perceptions of organizational structure, and social attributes.
Journalism Quarterly, Vol. 72, No. 3, p. 682-697.

Rosenfeld, M., & Zdep, S. M. (April, 1971). Intrinsic-extrinsic aspects of work
and their demographic correlates," Psychological Reports, Vol. 2, p. 359-62.

Schawartz, M. M., Jenusaitis, E., & Stark, H. (1963). Motivational factors among
supervisors in the utility industry. Personnel Psychology, Vol. 16, p. 45-53.

Shaver, H. C. (1978). Job satisfaction and dissatisfaction among journalism
graduates. Journalism Quarterly, Vol. 55, No. 1, p. 54-61 & 108.

Shore, L. M., & Martin, H. J. (1989). Job satisfaction and organizational
commitment in relation to work performance and turnover intentions. Human
Relations, Vol. 42, No. 7, p. 625-638.

Smith, P. C., Kendall, L M., & Hulin, C. L. (1969). The Measurement of
Satisfaction in Work and Retirement: A Strategy for the Study of Attitudes.
Chicago, IL.: Rand McNally & Company.

Stamm, K. & Underwood, D. (1993). The relationship of job satisfaction to
newsroom policy changes. Journalism Quarterly, Vol. 70, No. 3, p. 528-541.

Tharp, M. (1991). Turnover and mobility at small daily newspapers. Newspaper
Research Journal, Vol. 12, No. 1, p. 76-90.

Weaver, D. H., & Wilhoit, G. C. (1994). Daily newspaper journalists in the
1990s. Newspaper Research Journal, Vol. 15, No. 3, p. 3-21.

Weaver, D. H., & Wilhoit, G. C. (1996). The American Journalist in the 1990s:
U.S. News People at the End of an Era. Mahwah, N.J.: Lawrence Erlbaum
Associates, Publishers.

White, G. L. (1995). Employee turnover: The hidden drain on profits. Hr Focus,
Vol. 72, No. 1, p. 15-17.

Wimmer, R. D., & Domnick, J. R. (1994). Mass Media Research: An Introduction
(4th ed.). Belmont, CA: Wadsworth Publishing Co.

Yeager, S. J. (1981). Dimensionality of the Job Descriptive Index. Academy of
Management Journal, Vol. 24, p. 462.