the causal order of job satisfaction and organizational commitment in models of employee turnover

THE CAUSAL ORDER OF JOB SATISFACTIONAND ORGANIZATIONAL COMMITMENT IN

MODELS OF EMPLOYEE TURNOVER

Douglas B. Currivan

University of Massachusetts, Boston, MA, USA

This study examines four possible models of the causal relationship

between job satisfaction and organizational commitment in models of

employee turnover: (1) satisfaction precedes commitment, (2) commitment

precedes satisfaction, (3) satisfaction and commitment have a reciprocal

relationship, and (4) satisfaction and commitment have no significant

relationship. Structural equation models with two-wave panel data reveal

no significant effects between satisfaction and commitment, and the overall

fit for each of the four alternative models is virtually identical. Although

analysis of the alternative models suggests the relationship between

satisfaction and commitment may be spurious due to common determi-

nants, alternative explanations for these results and conclusions for

turnover research are discussed.

The purpose of this study is to analyze the causal relationship between job

satisfaction and organizational commitment in models of employee turnover.

Like much research on employee attitudes and behaviors in the workplace,

researchers have proposed and tested various models of the process leading to

employee turnover without actually establishing the causal order among two

key determinants of turnover, satisfaction and commitment (Bluedorn, 1982;

Dougherty, Bluedorn, & Keon, 1985; Hom & Griffeth, 1995; Kim, Price,

Mueller, & Watson, 1996; Lee, Ashford, Walsh, & Mowday, 1992; Mobley,

1982; Mowday, Porter, & Steers, 1982; Mueller, Boyer, Price, & Iverson, 1994;

Price & Mueller, 1986a). Furthermore, demonstrating the causal order be-

tween determinants of turnover like satisfaction and commitment also re-

quires specifying the indirect causal influences of other determinants of

turnover. In addition to the employee orientations of satisfaction and commit-

ment, models of employee turnover commonly specify various individual

characteristics, workplace structures, and environmental conditions as deter-

Direct all correspondence to: Douglas B. Currivan, University of Massachusetts at Boston, Center for Survey

Research, 100 Morrissey Boulevard, Boston MA 02125-3393, USA. E-mail: [email protected]

Human Resource Management Review, Copyright # 2000Volume 9, Number 4, 1999, pages 495 524 by Elsevier Science Inc.

All rights of reproduction in any form reserved. ISSN : 1053 4822

minants of turnover (Iverson & Roy, 1994; Kim, et al., 1996; Price & Mueller,

1986a). A dearth of explicit assumptions on the causal relationships among

employee orientations like satisfaction and commitment, and among other

presumed determinants of turnover, has limited our understanding of causal

relationships in turnover research.

Furthermore, common limitations in research design of prior studies of the

turnover process have prevented researchers from conducting appropriate

analysis, and drawing precise conclusions, on the causal relationship between

satisfaction and commitment. Despite the well-known limitations of cross-

sectional data for studying causal relationships (Finkel, 1995; Menard, 1991),

relatively few studies of the determinants of turnover like satisfaction,

commitment, and intent to stay, have collected and analyzed longitudinal

data. Researchers have typically used data collected from a sample of

employees at one time point. Without incorporating a time element into the

analysis of proposed models, existing studies have been unable to empirically

demonstrate the causal ordering among satisfaction and commitment.

Furthermore, the few longitudinal studies that have been conducted do not

analyze a complete model of the determinants of turnover, risking serious

model specification errors. As a result, researchers must base existing

conclusions on the relationship between satisfaction and commitment in

turnover models primarily on implicit assumptions about the influences of

other determinants.

This study analyzes the causal relationship between job satisfaction and

organizational commitment often proposed in models of the turnover process.

Models frequently specify satisfaction and commitment as intervening

variables between structural and individual determinants and turnover, yet

the causal ordering between these two employee orientations has not been

firmly established. This study contributes to our understanding of the turn-

over process by specifying more precise assumptions about the relationship

between satisfaction, commitment, and other determinants in the model.

While the emphasis in this study is on the causal relationship between

satisfaction and commitment, this research also provides evidence on the

causal impact of structural and individual determinants on these two employ-

ee orientations and, indirectly, on turnover. Furthermore, the analysis

employs a two-wave panel design and structural equations modeling to

analyze causal relationships in a model of turnover. Although these research

methods cannot address all of the limitations of prior study designs, the

methodology provides a significant enhancement to existing research on

employee turnover.

THE CONCEPTS OF SATISFACTION, COMMITMENT, AND TURNOVER

The key conceptual distinction between satisfaction, commitment, and turn-

over is that the first two concepts are employee attitudes or orientations, but

turnover refers to an employee behavior. Job satisfaction is perhaps the most

HUMAN RESOURCE MANAGEMENT REVIEW VOLUME 9, NUMBER 4, 1999496

widely studied work orientation over the last four decades of organizational

research. Researchers have defined and measured satisfaction both as a global

construct and as a concept with multiple dimensions or facets (Price, 1997).

This study adopts the global approach over the facet approach, conceptualizing

job satisfaction as the degree of positive emotions an employee has toward a

work role (Kalleberg, 1977; Locke, 1976; Smith, Kendall, & Hulin, 1969).

Whereas satisfaction denotes positive emotions toward a particular job,

organizational commitment is the degree to which an employee feels loyalty to

a particular organization. (Mueller, Wallace, & Price, 1992; Price, 1997).

Similar to Meyer and Allens dimension of affective commitment, this research

conceptualizes organizational commitment as an affective form of commitment

based on feelings of loyalty toward the organization (Allen & Meyer, 1990;

Meyer & Allen, 1984; Meyer, Allen, & Gellatly, 1990). Furthermore, this study

focuses on commitment to an organization as a whole entity, rather than

various constituencies or subgroups within the organization (Becker, 1992;

Becker & Billings, 1992; Reichers, 1985).

Whereas satisfaction and commitment focus on employee orientations

toward their work and organization, turnover refers to actual movement across

the membership boundary of an organization (Price, 1977, 1997). The specific

form of turnover of interest in this study, and most organizational research, is

voluntary separations or quits (Bluedorn, 1982; Price, 1997). Since data on

employees who quit voluntarily is typically difficult to collect, researchers often

focus on the most direct determinant of turnover, intent to stay (Bluedorn,

1982; Iverson, 1992; Price, 1997). Intent to stay is the degree of likelihood of an

employee maintaining membership in an organization (Iverson, 1992; Mueller

et al., 1994; Price & Mueller, 1981, 1986a). Intent to stay (or leave) refers to an

employees behavioral intentions, and has been demonstrated to exert a strong

negative influence on actual turnover (Bluedorn, 1982; Iverson, 1992; Mueller

et al., 1992; Price & Mueller, 1981, 1986a).

Turnover (or intent to stay) is one of the most widely studied outcomes of

both satisfaction and commitment, based on an axiomatic connection research-

ers make between employee attitudes and behaviors (Bluedorn, 1982; Mowday

et al., 1982; Shore, Newton, & Thornton, 1990). Researchers often propose job

satisfaction and organizational commitment as intervening variables between

other determinants (e.g., structural and individual variables) and outcomes

like stay intentions and employee turnover (Iverson, 1992; Mueller et al., 1992;

Price & Mueller, 1986a). A substantial body of empirical evidence links greater

commitment to greater intent to stay and, consequently, lower turnover

(DeCotiis & Summers, 1987; Hom & Griffeth, 1995; Kalleberg & Berg, 1987;

Lee et al., 1992; Lincoln & Kalleberg, 1996; Mathieu & Zajac, 1990; Mowday et

al., 1982; Mueller et al., 1992; Price & Mueller, 1986a; Randall, 1990; Somers,

1995). Fewer studies support a direct link between satisfaction and turnover

(Mueller et al., 1994), yet several support an indirect influence through

commitment (Lincoln & Kalleberg, 1985, 1990; Mowday et al., 1982; Mueller

et al., 1994; Price & Mueller, 1986a; Wallace, 1995). Because the specific order

of the causal relationship between satisfaction and commitment has not been

THE CAUSAL ORDER OF JOB SATISFACTION AND ORGANIZATIONAL COMMITMENT 497

clearly established, each of these employee orientations could have significant

direct or indirect influences on turnover.

THE CAUSAL MODEL

The Causal Order of Satisfaction and Commitment

Fig. 1 presents the model of the turnover process used to illustrate the

relationship between job satisfaction and organizational commitment in this

study. The dominant view in the literature assumes satisfaction causes

commitment (Lincoln & Kalleberg, 1985, 1990; Mowday et al., 1982; Mueller

et al., 1994; Price & Mueller, 1986a; Wallace, 1995). Researchers taking this

position implicitly assume employee orientations toward a specific job neces-

sarily precede orientations toward the entire organization. Typically, research-

ers have coupled this view with a corollary assumption that, compared to

organizational commitment, job satisfaction varies more directly and instan-

taneously with changing working conditions (Mowday et al., 1982; Mueller

Price, & Wynn, 1996). A wealth of empirical evidence supports the causal

precedence of satisfaction over commitment (Bluedorn, 1982; Iverson, 1992;

Lincoln & Kalleberg, 1990; Mowday et al., 1982; Mueller et al., 1994; Price &

Mueller, 1986a; Wallace, 1995; Williams & Hazer, 1986). Yet empirical

research has not always confirmed this causal ordering, and researchers have

advanced alternative arguments.

At least two studies have found commitment to be causally prior to

satisfaction (Bateman & Strasser, 1984; Vandenberg & Lance, 1992). These

researchers have argued that employees adjust their satisfaction levels to be

consistent with their current commitment levels. This position is consistent

Figure 1. The Causal Model. Dashed Pathways () Represent Presumed Influences ofSatisfaction and Commitment on Intent to Stay and Turnover that are Not Analyzed

in This Study


with a general social psychological perspective that assumes individuals

develop attitudes consistent with situations to which they are already com-

mitted (OReilly & Caldwell, 1980, 1981; Staw, 1980). Yet the evidence for this

second proposition is rather limited, compared to the first assumption.

Furthermore, given arguments and evidence to the contrary, this position

does not account for the role specific job attitudes play in the formulation of

more general organizational attitudes (Lincoln & Kalleberg, 1990).

Several other studies have concluded a reciprocal relationship exists

between satisfaction and commitment (Farkas & Tetrick, 1989; Lance,

1991; Mathieu, 1991; Mottaz, 1988). With the Farkas and Tetrick (1989)

study as the one exception, three of the studies observed an asymmetric

relationship where satisfaction had a stronger effect on commitment than the

reverse. Likewise, Lincoln and Kalleberg (1990) actually observed a signifi-

cant reciprocal relationship between satisfaction and commitment in the US

and Japan, although they used the significantly stronger effect of the former

on the latter to conclude that satisfaction precedes commitment. Farkas and

Tetricks (1989) three-wave longitudinal analysis of the relationship between

satisfaction and commitment produced a reciprocal relationship, but their

results did not favor one causal direction over the other. While none of these

researchers advanced a cogent argument to explain the reciprocal relation-

ship, support for a reciprocal relationship between satisfaction and commit-

ment suggests the competing arguments outlined in the two preceding

paragraphs may be reconciled.

Based on a finding of no significant causal relationship between satisfac-

tion and commitment, other studies have suggested that the relationship

between these two work attitudes may be spurious due to their multiple

common causes (Curry, Wakefield, Price, & Muller, 1986; Dougherty et al.,

1985). This argument has some appeal, given the increasing tendency of

researchers to assume similar determinants affect the two affective orienta-

tions (Kim et al., 1996; Lincoln & Kalleberg, 1985, 1990; Mottaz, 1988;

Mueller et al., 1994; Wallace, 1995). Yet only two studies have produced this

non-significant relationship, and some compelling arguments causally link

satisfaction and commitment.

Unlike existing a priori and ad hoc arguments on the relationship between

job satisfaction and organizational commitment, Lawlers (1992) theory of

affective attachment provides a theoretical linkage between more immediate

positive emotions like satisfaction, and more enduring affective attachments

like commitment. His theory first makes a conceptual distinction between

more transitory emotions, like feelings of satisfaction, and more enduring

affect, such as commitment. The theory then assumes positive emotional states

precede affective attachments to collectivities, acting as a meditating factor

between social structures and affective attachments (Lawler, 1992). Mueller

and Lawler (1996) extend these assumptions to the relationship between

working conditions, job satisfaction, and organizational commitment. They

argue that working conditions regularly produce positive or negative emotions

such as job satisfaction, and employees attempt to understand the contextual


sources of such emotions. The organization (or a subunit of the organization)

becomes a target for these feelings and the organization (or unit) viewed as

responsible for positive emotions is most likely to elicit an affective attachment

from the employee (Mueller & Lawler, 1996).

Following Lawler and Muellers arguments, the model proposed in this

study assumes that job satisfaction plays a dual role as both an important

determinant of organizational commitment, and as an intervening variable

between structural and individual determinants and commitment. Despite

considerable variation in methodology and results, empirical research favors a

positive influence of satisfaction on commitment, as well as a potential

mediating role of satisfaction (Bluedorn, 1982; Iverson, 1992; Lincoln &

Kalleberg, 1985, 1990; Mowday et al., 1982; Mueller et al., 1994; Price &

Mueller, 1986a; Wallace, 1995a; Williams & Hazer, 1986). The model assumes

higher job satisfaction produces higher organizational commitment. Although

the model makes this assumption, the analysis examines each of the four

alternative relationships between satisfaction and commitment to assess the

causal ordering.

Structural and Individual Determinants of Satisfaction and Commitment

Although the causal model used in this study draws upon the broad

literature on satisfaction, commitment, and turnover, the work of Price and

Mueller and their colleagues provides the theoretical basis of the model (Han,

Ko, Price, & Mueller, 1995; Iverson, 1992; Iverson & Roy, 1994; Kim et al.,

1996; Price & Mueller, 1986a). Price and Muellers model seeks to provide a

more complete explanation of the turnover process than alternative theories by

integrating sociological, economic, and psychological perspectives on employee

turnover in organizations (Han et al., 1995; Iverson, 1992; Kim et al., 1996;

Mueller & Price, 1990; Price & Mueller, 1986a). In addition, the review and

meta-analysis of research on organizational commitment conducted by

Mathieu and Zajac (1990) provides a strong empirical basis for the model

developed by Price and Mueller and their colleagues. These researchers

performed meta-analyses for 48 determinants, correlates, and consequences

of commitment based on 124 published studies. Although this research is now

a decade old, the large number of studies examined in the meta-analysis

suggest their specific conclusions would not likely change if more recent

research were included.

The results of Mathieu and Zajacs (1990) meta-analysis provide guidance

for simplifying the casual model to be analyzed in this study. The proposed

model deletes determinants from the Price and Mueller model that the meta-

analysis revealed do not consistently have a significant empirical relationship

with commitment, the final dependent variable analyzed in the model. The

simplified model retains the structural variables of autonomy, routinization,

social support (peer support and supervisor support), job stress (role ambi-

guity, role conflict, and workload), and pay. The model specfically proposes

autonomy, peer support, supervisor support, and pay positively influence


satisfaction and commitment, while routinization, role ambiguity, role conflict,

and workload negatively influence the two employee orientations (DeCotiis &

Summers, 1987; Dunham, Grube, & Castaneda, 1994; Glisson & Durick, 1988;

Kalleberg & Reve, 1992; Kim et al., 1996; Ko, Price, & Mueller, 1997; Lincoln &

Kalleberg, 1985, 1990; Mottaz, 1988; Mueller & Price, 1990; Mueller et al.,

1994; Price & Mueller, 1986a; Wallace, 1995a).

In addition to social structures within the organization, the model incor-

porates one individual characteristic, job involvement (Kanungo, 1982). Em-

pirical evidence indicates job involvement is distinct from both satisfaction

and commitment (Brooke, Russell, & Price, 1988; Mathieu & Farr, 1991).

Employees with higher job involvement are more likely to receive organiza-

tional rewards. We can assume greater organizational rewards, in turn, lead

to greater job satisfaction and organizational commitment. Following this

argument, job involvement is expected to have positive effects on satisfaction

and commitment.

Furthermore, the model proposes satisfaction not only has a positive impact

on commitment, but also potentially mediates the effects of the structural and

individual variables on commitment. Although not analyzed in this article, the

logical extension of the model is that satisfaction and commitment also have

positive influences on intent to stay, which in turn exerts a strong negative

TABLE 1Definitions of Theoretical Constructs in the Model

Construct Definition

Organizational Commitment Degree to which an employee feels loyalty tothe organization

Job Satisfaction Degree to which an employee has positive emotionstoward the work role

Autonomy Degree to which an employee exercises discretion overthe performance of job tasks

Routinization Degree to which job tasks are repetitiveSocial Support Degree of consideration individuals receive from others

in their social networkPeer Support Degree of consideration expressed among employees

with similar positions in the organizationSupervisor Support Degree of consideration expressed from an employee

in a superior position to an employee in a subordinateposition in the organization

Job Stress Degree to which employees experience difficulty inperforming their job

Role Ambiguity Degree to which job expectations are unclearRole Conflict Degree to which job expectations are incompatibleWorkload Amount of performance required by a jobPay Remuneration for work performed in an organizationJob Involvement Extent to which an employee is willing to perform the

work role


influence on turnover. In Fig. 1, the unanalyzed consequences of satisfaction

and commitment are represented by dashed arrows in the model. Table 1

presents definitions for the theoretical constructs specified in the model.

The proposed model includes no demographic variables, but the analysis

controls for five demographic characteristics: tenure, education, age, marital

status, and gender. Although each of these variables has exhibited a significant

correlation with commitment in previous research (Mathieu & Zajac, 1990),

the analysis includes the five demographic variables only to check for mis-

specification of the model (Iverson, 1992; Kim et al., 1996). The proposed

causal model includes numerous explanatory variables, but the model may not

identify all relevant causes of satisfaction and commitment. The analysis

controls for the five demographic variables to assess whether the model omits

additional determinants of satisfaction and commitment.

General Assumptions of the Model

A further advancement in recent work by Price, Mueller, and their

colleagues is to explicitly specify some of the general assumptions of their

model of the turnover process (Kim et al., 1996). Following this precedent, a

few important general assumptions elaborate the theoretical basis of the

causal model. First, expectancy theory underlies the specific assumptions of

the model (Mowday et al., 1982; Porter & Steers, 1973; Porter, Steers,

Mowday, & Boulian, 1974; Steers, 1977; Vroom, 1964). This perspective

assumes employees enter work organizations with expectations and values

about the organization of their workplace. Expectations are beliefs about what

conditions will characterize the workplace, while values refer to employee

conceptions of desired outcomes in the workplace (Kalleberg, 1977; Kim et al.,

1996; Mottaz, 1988). The model assumes employees have specific expectations

regarding each of the eight structural determinants specified in the model.

Furthermore, the model assumes employees positively value autonomy, peer

and supervisor support, and pay, but negatively value routinization and the

three sources of job stress (role ambiguity, role conflict, and workload). To the

extent that employee expectations and values are met, positive emotions

toward the work role (satisfaction) and feelings of loyalty to the organization

(commitment) result.

Second, the proposed model explicitly adopts a social exchange perspective,

assuming that satisfaction and commitment develop through an exchange of

employers rewards for employees work. Rewards are resources that employ-

ees value in the workplace. Traditionally, explanations of commitment have

explicitly or implicitly assumed an exchange relationship between employees

and organizations, whereby work orientations are influenced by the degree to

which the employer provides (or is perceived to provide) desired rewards

(Farrell & Rusbult, 1981; Kalleberg, 1977; Mottaz, 1988; Price & Mueller,

1986a; Rusbult & Farrell, 1983).

A third important assumption relates to the exact nature of the influences

of the proposed determinants of satisfaction and commitment. The traditional


expectancy perspective assumes employee values and expectations interact

with workplace structures to influence employee attitudes (Kalleberg, 1977;

Locke, 1976; Mobley, 1982; Porter & Steers, 1973; Vroom, 1964). Further-

more, research has proposed two types of buffering relationships moderate

the relationship between job stress and job satisfaction (House, 1981; Kahn,

Wolfe, Quinn, Snoek, & Rosenthal, 1964). Yet, to date, empirical research

offers little support for these four types of interactions (Beehr & Drexler,

1986; Iverson, 1992; Kim et al., 1996; Price & Mueller, 1986a). Since the

weight of evidence supports additive effects of work values, job expectations,

autonomy, social support, and job stress on employee orientations, the model

assumes all relationships are additive and the analysis does not test any of

these interactions.

Scope Conditions of the Model

Scope conditions are universal statements that define the class of circum-

stances in which assumptions and hypotheses are applicable (Cohen, 1989).

The causal model has three scope conditions: (1) the model applies only to

individuals employed by work organizations, (2) the model applies to employ-

ees who have a considerable work role in the organization (such as full-time,

permanent employees), and (3) the model applies only to contexts where

employees can move freely among work organizations.

METHODS

Site and Sample

The site for this study was the Chicago Public Schools (CPS), the third

largest school system in the Untied States with over 550 elementary and

secondary schools, 410,000 enrolled students, and 23,500 regular teachers. The

sample frame consisted of classroom teachers who entered this school system

between Fall 1988 and Fall 1990 (Iverson, 1992). The sample excludes admin-

istrators and ancillary staff, so that all sample members are professional

school staff employees who instruct students. The sample is comprised of

full-time, permanent employees who can, and do, move into and out of teaching

positions in the CPS and other school systems. Two advantages accrue from

using this sample. First, limiting the sample to early-career teachers controls

for the confounding influence of tenure. Job rewards are tightly linked to

seniority in public school teaching, so many working conditions vary by teacher

tenure in the school system. Second, since the teachers work at both elemen-

tary and high schools throughout the school system, some variation in working

conditions is likely to exist. Although these teachers are likely to share

similarities in their basic working conditions, the different management styles

of local school administrators and the different structures of elementary vs.

high schools are likely to produce variation in working conditions.


Data Collection

The data collection followed a prospective panel design of two survey waves

(Menard, 1991). The first data collection period occurred in the 19891990

school year. The researchers mailed the first questionnaire on January 15,

1991 to 1,476 teachers selected from CPS personnel records (Iverson, 1992).

The research team encouraged participation by following the techniques

recommended by Dillman (1978), including reminder notices and multiple

mailings. By the designated close-out date of May 1, 1991, 838 teachers

representing 405 schools returned the questionnaire, for a first-wave response

rate of 57 percent (838/1,476).

On March 15, 1993, the research team sent respondents to the first

questionnaire who continued as classroom teachers in the CPS a follow-up

questionnaire at their schools. For the second survey administration, 482

teachers from 294 schools returned the questionnaire by the cut-off date of

July 1, 1993. After deletion of eight respondents who had become ineligible or

unreachable, the response rate for the second questionnaire was 58 percent

(482/830). As a result, the corrected panel attrition rate was 42 percent

({830482}/830).Data reported for this study comprise only the panel members who com-

pleted both questionnaires. As Menard (1991) points out, the results of

analysis of panel data with an attrition rate greater than 50 percent may

not be generalized beyond those respondents retained in the study. Since the

attrition rate in the present survey did not exceed that standard, we may

tentatively generalize the data beyond the panel to the original sample, but not

necessarily to the original population.

Panel Representativeness

Examination of key demographic characteristics provides a comparison of

the 482 panel members who completed both questionnaires with the 356 who

only completed the first questionnaire. Only two significant differences exist

between the characteristics of those who remained in the panel for both waves

versus those who did not: (1) panel members are older than those who left the

study and (2) panel members have more years of experience in teaching than

leavers have. No significant differences exist between panel members and

leavers with respect to race/ethnicity, gender composition, educational level,

school type, years teaching in the CPS, and teaching status. This analysis

suggests the panel of 482 teachers comprise a representative sub-sample of the

original sample of 838. Examination of key demographic characteristics

tentatively indicates that findings obtained from analysis of the panel data

may generalize to the sample frame of 1,476 teachers originally selected

(Menard, 1991).

Yet further comparisons indicate panel members and leavers differ signifi-

cantly on several measures used to assess the major constructs examined in

the study. For example, CPS teachers who completed both questionnaires are


significantly more committed to their schools and more satisfied with their jobs

than those who only completed the first questionnaire. In general, these

comparisons indicate teachers who left the panel after the first wave tend to

have more negative views of their working conditions than do those who

remained in the panel. These data indicate that caution must be exercised in

attempting to generalize findings obtained with the panel data to the original

sample frame of 1,476 CPS teachers originally selected. These differences

between panel members and leavers support the decision to analyze the data

from only those respondents who completed both questionnaires.

Measures

Multiple survey items assessed all of the theoretical constructs examined

in the study, with the exception of pay (which had a single standard item)

and job involvement (which was not measured in both survey waves). This

measurement strategy implicitly assumes employees subjective perceptions

of their working conditions meaningfully reflect objective conditions in their

workplace. The workplace is conceptualized as the individual school in this

study, and the measure of organizational commitment reflects this definition.

Although teachers work both at individual schools and for an entire school

system, Lawler (1992) argues commitment to a smaller, nested organiza-

tional unit (e.g., a school) is a necessary condition for commitment to the

larger organizational unit (i.e., school system). Furthermore, given that the

smaller unit is more salient to employees than the larger unit, positive

emotions like job satisfaction is likely to strengthen teacher commitment to

their schools more than to the school system (Mueller & Lawler, 1996). For

this reason, the measure of organizational commitment reflects teacher

commitment to their school.

All measures included in both survey waves have demonstrated acceptable

reliability and validity in previous studies of employee orientations in work

organizations (Iverson, 1992; Iverson & Roy, 1994; Kim et al., 1996; Ko et al.,

1997; Mottaz, 1988; Mueller et al., 1994; Mueller et al., 1992; Price & Mueller,

1986a,b). All questionnaire items employed a five-point scale with 1 assigned

to strongly agree and 5 assigned to strongly disagree. For these five-point

scales, the researchers reversed-coded positively-worded items so that the

data records higher scores for respondents who agreed with each item.

Appendix A presents the questionnaire items for the multiple-indicator vari-

ables and the corresponding scoring for the response categories.

Table 2 presents descriptive statistics for the theoretical variables measured

in the two survey waves, along with reliability coefficients for the multiple

indicator scales. Composite scoring for these two tables assumes each item for

a multiple-indicator variable is a parallel measure, so values for the variables

reflect a simple summation of the items. Subsequently, the analysis under-

takes confirmatory factor analysis (CFA) procedures to determine the appro-

priate loading (or weighting) of individual items for the multiple-indicator

measures of theoretical constructs.


TA

BL

E2

Descriptiv

eS

tatis

tics

for

Theore

ticalV

ariable

sin

the

Model

Constr

uct

Item

s

Tim

e1

Tim

e2

MS

DR

elia

bili

tyM

SD

Relia

bili

ty

Auto

nom

y3

10.2

02.3

60.5

310.3

72.4

30.6

2R

outin

izatio

n3

6.1

11.8

70.6

06.2

82.1

60.7

1P

eer

Support

311.2

32.4

30.7

911.1

32.4

10.8

1S

uperv

isor

Support

27.1

82.1

20.8

46.6

92.2

70.8

2R

ole

Am

big

uity

35.3

71.8

80.6

15.1

01.6

70.5

8R

ole

Confli

ct

24.9

81.6

40.6

65.0

41.6

30.6

6W

ork

Overload

310.3

12.6

40.6

610.3

02.5

80.6

7P

ay

1U

S$26,4

76

US

$4,8

73

U

S$33,6

36

US

$9,2

55

Job

Satis

factio

n4

16.6

12.7

00.7

716.2

12.9

10.8

1O

rganiz

atio

nalC

om

mitm

ent

415.1

53.3

00.8

014.9

53.1

50.8

0


Similar to pay, standard single questionnaire items from the first survey

wave tap each of the five demographic characteristics included in the causal

model. Tenure (M = 4.60, SD = 3.79) in the organization assesses how long the

respondent has taught in the CPS, following an ascending five-category scale

ranging from less than six months (0) to more than six years (10). The

indicator for education (M = 17.33, SD = 1.08) measures the highest level of

schooling completed on a five-category scale, ranging from a four-year college

degree to a PhD degree. Age (M = 39.58, SD = 9.51) is measured by an item

with five exhaustive ranges. Recoded from five original categories, marital

status (59% married) is coded as 0 for all unmarried respondents and 1 for all

currently married respondents. Gender (81% female) is recorded as 0 for males

and 1 for females.

Tables 3 and 4 present the standardized LISREL correlations among the

variable for waves 1 and 2, respectively. Three points regarding the correla-

tions are important to note. First, at both wave 1 (r = 0.72) and wave 2 (r =

0.69), the correlation between satisfaction and commitment is positive and

significant. Second, among all variables at waves 1 and 2, the highest correla-

tion is observed between autonomy and routinization. This correlation exceeds

0.80 for both waves and indicates a potential multicollinearity problem for

estimating a causal model with these two variables. Second, not shown in the

table are the correlations for the measures of routinization, role ambiguity, and

job satisfaction between the first and second survey waves. Despite measure-

ment instability in these variables across survey waves, the time 1 and time 2

measures for these three constructs are highly positively correlated like the

other variables.

Analysis

The analysis conducted in this study applies the statistical techniques of

linear structural relations (LISREL). Although LISREL and multiple regres-

sion share similarities in terms of underlying assumptions and interpreting

their results, the numerous advantages of LISREL favor its use in analysis

of the causal relationships among the variables specified in the model.

LISREL consists of both a measurement model and a structural equations

model (Bollen, 1989; Hayduk, 1987; Joreskog & Sorbom, 1994). The mea-

surement model specifies a confirmatory factor analysis (CFA) of proposed

relationships between the manifest (observed) indicators and latent (theore-

tical) constructs, while the structural equation model (SEM) specifies hypothe-

sized relationship among latent constructs. A set of variancecovariance

matrices, manipulated following the basic rules of matrix algebra, represents

the parameters in both components (Bollen, 1989; Hayduk, 1987; Joreskog &

Sorbom, 1994). The CFA component actually allows researchers to specify a

measurement model to assess how well observed indicators measure theo-

retical variables. Consequently, LISREL is particularly useful in this study

for estimating multiple-indicator variables, correlated residuals, and recur-

sive causality.


TA

BL

E3

LIS

RE

LS

tandard

ized

Zero

-ord

er

Corr

ela

tions

for

Wave

1V

ariable

s

Variable

s1

23

45

67

89

10

11

12

13

14

(1)

Auto

nom

y

(2)

Routin

izatio

n0

.81

(3)

Peer

Support

0.4

60

.35

(4)

Superv

isor

Support

0.6

80

.41

0.4

1

(5)

Role

Am

big

uity

0.5

30.4

30

.360

.49

(6)

Role

Confli

ct

0.5

40.2

70

.460

.54

0.4

8

(7)

Work

load

0.3

90.1

00

.180

.29

0.3

00.4

5(8

)P

ay

0.0

70

.06

0.0

80.0

40

.110

.010

.03

(9)

Tenure

0.0

20.0

20

.010

.01

0.0

20.0

10.0

10.3

4(1

0)

Educatio

n0

.02

0.0

10

.030

.04

0.0

10.0

20.0

10.5

10.0

0(1

1)

Age

0.0

20

.01

0.0

40.0

50

.010

.250

.01

0.0

50.2

80

.00

(12)

Marita

lS

tatu

s0

.07

0.0

80

.040

.06

0.0

70.0

50.0

40.0

10.0

10.0

10

.00

(13)

Gender

0.0

00

.000

.010

.010

.00

0.0

00.0

00

.000

.000

.010

.00

0.0

0(1

4)

Job

Satis

factio

n0.6

40

.78

0.3

40.4

70

.640

.430

.400

.100

.060

.07

0.0

10

.11

0.0

0

(15)

Org

aniz

atio

nal

Com

mitm

ent

0.5

40

.54

0.4

90.6

80

.430

.440

.31

0.0

10

.010

.05

0.0

80

.090

.01

0.7

2

Sta

ndard

ized

corr

ela

tions

ofj0.1

1jo

rgre

ate

rare

sig

nifi

cant

at

p=

0.0

5.


TA

BL

E4

LIS

RE

LS

tandard

ized

Zero

-ord

er

Corr

ela

tions

for

Wave

2V

ariable

s

Variable

s1

23

45

67

89

(1)

Auto

nom

y(2

)R

outin

izatio

n0

.83

(3)

Peer

Support

0.4

70

.41

(4)

Superv

isor

Support

0.7

50

.47

0.3

2(5

)R

ole

Am

big

uity

0.6

70.5

90

.29

0.4

6(6

)R

ole

Confli

ct

0.5

40.3

70

.56

0.5

10.5

1(7

)W

ork

load

0.3

80.0

80

.10

0.3

20.2

50.4

2(8

)P

ay

0.0

80.0

30.0

50.0

70

.05

0.0

20

.04

(9)

Job

Satis

factio

n0.7

40

.80

0.3

90.4

90

.61

0.4

30

.40

0.0

9(1

0)

Org

aniz

atio

nalC

om

mitm

ent

0.6

20

.57

0.3

80.7

20

.39

0.4

00

.26

0.0

50.6

9

Sta

ndard

ized

corr

ela

tions

ofj0.1

1jo

rgre

ate

rare

sig

nifi

cant

at

p=

0.0

5.


An important step before proceeding with the primary analysis is specifying

how the study design provides appropriate conditions for the assessment of

causal relationships. Data must satisfy three well-known conditions to estab-

lish causal relationships among two or more variables: (1) the variables must

covary, as indicated in non-experimental studies by a non-zero correlation, (2)

another variable or set of variables must not produce the covariation(s)

between the variables (i.e., non-spurious), and (3) the causal variable must

precede the effect variable in time, as indicated by the change in the cause

occurring no later than the associated change in the effect (Finkel, 1995;

Menard, 1991). The two-wave panel design implemented in this study can

satisfy these conditions of causality more thoroughly than previous research on

satisfaction, commitment, and turnover, which typically involves analysis of

cross-sectional data. The panel data allows the time sequencing of explanatory

variables and organizational commitment to be established more clearly,

rather than assuming simultaneous effects or simply ignoring time order

(Farkas & Tetrick, 1989; Finkel, 1995). Longitudinal designs remain relatively

rare, and the few existing studies are limited in their focus (Cramer, 1996;

Vandenberg & Lance, 1992; Vandenberg & Scarpello, 1994).

The primary analysis uses the two-wave panel data to test the causal

relationship between satisfaction and commitment, as well as the causal

influences of other determinants in the model. Several features of LISREL

provide for a more robust analysis of the causal relationships among the

workplace structures and employee orientations than alternative multivari-

ate techniques, such as regression analysis (Anderson & Williams, 1992;

Hom & Griffeth, 1995). The SEM techniques in LISREL incorporate true

stability coefficients to control for the time-lagged effects of a construct at

time 1 on that same construct at time 2 (Cramer, 1996; Hom & Griffeth,

1995; Mueller et al., 1996). Since correlated measurement errors between

repeated applications of the same indicators over time (autocorrelations) can

bias observed causal effects in longitudinal analyses, the LISREL analysis

explicitly models these correlations (Cramer, 1996; Hom & Griffeth, 1995;

Kessler & Greenberg, 1981; Mueller et al., 1996). Since all relevant causes

of commitment may not be identified by the model, the LISREL procedures

further reduce bias in causal effects by estimating correlations between the

disturbance terms of structural equations at each time point. This procedure

accounts for unmeasured common causes for the two employee orientations,

satisfaction and commitment (Finkel, 1995). Despite all these advantages of

using LISREL to model causal relationships with the two-wave panel data,

the statistical models must still make some assumptions about the causal

process and empirical relationships among the variables in order to produce

interpretable results (Kessler & Greenberg, 1981).

Table 5 presents the hypotheses to be analyzed as specified in the causal

model, including only those variables for which wave 1 and wave 2

measures are available. The final hypothesis that satisfaction causally

precedes commitment (Hypothesis 9) is of central interest in this study.

Given alternative arguments and mixed evidence on the causal relationship


between satisfaction and commitment, the analysis explicitly tests which of

four possible relationships between the two employee orientations is most

consistent with the two-wave panel data.

RESULTS

The Causal Relationship between Satisfaction and Commitment

The analysis elaborates the causal relationship between job satisfaction and

organizational commitment by performing competitive tests among four alter-

native models: (1) the proposed model in which satisfaction precedes commit-

ment, (2) a model where commitment precedes satisfaction, (3) a model with

reciprocal effects between satisfaction and commitment, and (4) a model where

satisfaction and commitment have no significant relationship ( Vandenberg &

Lance, 1992). These tests alter the specification of the relationship between

satisfaction and commitment by adding and deleting paths between satisfac-

tion and commitment at time 1, and satisfaction and commitment at time 2.

The proposed model, with a time-lagged effect from wave 1 satisfaction to wave

2 commitment, serves as the first model. Assuming commitment actually

precedes satisfaction, the second model deletes the effect from wave 1 satisfac-

tion to wave 2 commitment, and instead specifies a path from wave 1 commit-

ment to wave 2 satisfaction. The third model proposes commitment and

satisfaction have reciprocal causal influences, so it includes two time-lagged

effects from wave 1 satisfaction to wave 2 commitment and from wave 1

commitment to wave 2 satisfaction. Finally, assuming no causal relationship

TABLE 5Summary of Hypotheses Operationalized and Tested

(1a) Greater autonomy produces greater job satisfaction.(1b) Greater autonomy produces greater organizational commitment.(2a) Greater routinization produces lower job satisfaction.(2b) Greater routinization produces lower organizational commitment.(3a) Greater peer support produces greater job satisfaction.(3b) Greater peer support produces greater organizational commitment.(4a) Greater supervisor support produces greater job satisfaction.(4b) Greater supervisor support produces greater organizational commitment.(5a) Greater role ambiguity produces lower job satisfaction.(5b) Greater role ambiguity produces lower organizational commitment.(6a) Greater role conflict produces lower job satisfaction.(6b) Greater role conflict produces lower organizational commitment.(7a) Greater workload produces lower job satisfaction.(7b) Greater workload produces lower organizational commitment.(8a) Higher pay produces higher job satisfaction.(8b) Higher pay produces higher organizational commitment.(9) Greater job satisfaction produces greater organizational commitment.


exists between commitment and satisfaction, the fourth model specifies no

time-lagged paths between satisfaction and commitment.

Two indicators assess which of the alternative models of the relationship

between satisfaction and commitment best fits the data. First, significant

LISREL coefficients between the two variables provide evidence of causal

effects. Second, since the first, second, and fourth models are actually nested

within the third model (the reciprocal causal effects model), we can evaluate

the models both by comparing their likelihood chi-square (x2) ratios and theiroverall fit indices (AGFI, NNFI, and NFI) provided in the LISREL output. A

significant reduction in the likelihood x2 ratios for the three nested modelsprovides evidence of improved fit over the encompassing model (Bentler &

Bonett, 1980; James, Mulaik, & Brett, 1982).

Table 6 provides the LISREL estimates of the relationships between

satisfaction and commitment for each of the four models. In the proposed first

model the time-lagged effect from wave 1 satisfaction to wave 2 commitment is

not significant, so Hypothesis 9 is not supported. The path from wave 1

commitment to wave 2 satisfaction in model two is likewise not significant.

The third model proposes commitment and satisfaction have reciprocal causal

influences, but neither of the time-lagged effects from wave 1 satisfaction to

wave 2 commitment or from wave 1 commitment to wave 2 satisfaction is

significant. Model 4 specifies no time-lagged paths between satisfaction and

commitment, and this hypothesis is supported by the lack of significant LISREL

coefficients. The overall fit indices are virtually identical across models,

indicating none of the models fits the data significantly better than the others.

TABLE 6LISREL Estimates of the Relationship between Satisfaction and Commitment

Model 1. SAT!COM 2. COM!SAT 3. SAT$COM 4. Zero EffectsCausal Variable

Job Satisfaction 0.091 0.087 Organizational

Commitment 0.031 0.007

R2

Job Satisfaction 0.885 0.887 0.886 0.887Organizational

Commitment0.748 0.745 0.748 0.744

x2 5,577.98 5,578.86 5,577.97 5,579.13(df) (1,646) (1,646) (1,645) (1,647)

AGFI 0.710 0.710 0.710 0.710NFI 0.700 0.695 0.695 0.695NNFI 0.735 0.734 0.734 0.735

All coefficients reported for job satisfaction and organizational commitment are com-pletely standardized.


Overall, these findings offer qualified support for the fourth model, where no

causal effects between satisfaction and commitment are hypothesized.

Although this model does not provide a better overall fit to the data than

the other models, none of the effects between satisfaction and commitment is

significant in any of the other models. This result suggests the relationship

between the two employee orientations is spurious, the result of common

causes. Presumably, the workplace structures included in the causal model

could be some of the common causes of satisfaction and commitment. The

Determinants of Satisfaction and Commitment examines which determinants

in the model have significant effects on satisfaction and commitment.

The Determinants of Satisfaction and Commitment

A major innovation of this study is to examine the relationship between job

satisfaction and organizational commitment using a full causal model that

incorporates determinants of both employee orientations. The analysis of the

causal relationship between satisfaction and commitment controls for pre-

sumed causes of these two variables, and thereby provides a more accurate

estimate of the relationship. The finding of no significant effects between

satisfaction and commitment suggests common determinants among the two

employee orientations may partially account for the strong positive correlation

observed between the two at both survey waves.

Table 7 gives the LISREL estimates for the determinants of satisfaction

and commitment specified in the proposed causal model, where satisfaction

is causally prior to commitment. Autonomy, peer support, and supervisor

support have significant positive effects on job satisfaction, while routiniza-

tion and workload have significant negative effects on satisfaction. These

results support Hypotheses 1a, 2a, 3a, 4a, and 7a, but offer no support for

the remaining three hypotheses for satisfaction. For organizational commit-

ment, peer support, supervisor support, and role conflict have significant

positive influences, while routinization and workload have significant nega-

tive influences. These findings support Hypotheses 2b, 3b, 4b, and 7b, but

contradict Hypothesis 6b. Role conflict was expected to exert a negative

effect on commitment, so the significant negative effect is surprising. The

analysis provides no support for the hypotheses regarding the influences of

autonomy, role ambiguity, and pay on commitment. None of the five

demographic variables has a significant effect on either satisfaction or

commitment. Furthermore, these results do not vary for the models with

the three alternative specifications of the relationship between satisfaction

and commitment.

Overall, routinization, peer support, supervisor support, and workload have

similar effects on satisfaction and commitment in this model. The significant

determinants account for considerable variation in both employee orientations,

although they explain considerably more variance in satisfaction (0.885) than

in commitment (0.748). Yet the measures of overall model fit, including the

likelihood chi-square (x2) ratio and the Adjusted Goodness of Fit Index (AGFI),


the Normed Fit Index (NFI), and the Non-Normed Fit Index (NNFI), indicate

the model provides a relatively poor fit to the data.

One potential reason for the poor overall fit of the model could be high

multicollinearity between autonomy and routinization. As noted in the text

below Table 7, the coefficients for autonomy and routinization presented in the

table are reported from two alternative structural equations models where only

one of these variables is included. When both variables are included in a model,

the high multicollinearity between them produces significant negative effects

for both autonomy and routinization on satisfaction and commitment, and the

standardized coefficient for routinization is greater than 1. The result for

autonomy strongly contradicts existing research (DeCotiis & Summers, 1987;

Dunham et al., 1994; Han et al., 1995; Kalleberg & Reve, 1992; Ko et al., 1997;

Lincoln & Kalleberg, 1990; Mathieu & Zajac, 1990; Mottaz, 1988), and the

result for routinization exceeds the upper bound for LISREL coefficients of 1.00.

Appendices B and C present the complete estimates for models including

only autonomy and routinization, respectively. These results should be com-

TABLE 7LISREL Estimates for the Proposed Model (Satisfaction!Commitment)

Dependent Variable

Explanatory Variable Job Satisfaction School Commitment

Autonomy 0.442*** 0.030Routinization 0.567*** 0.236***Peer Support 0.118* 0.159**Supervisor Support 0.240** 0.822***Role Ambiguity 0.087 0.041Role Conflict 0.072 0.079*Workload 0.326*** 0.141*Pay 0.053 0.056Tenure 0.029 0.006Education 0.038 0.080Age 0.083 0.136Marital Status 0.090 0.022Gender 0.018 0.023Job Satisfaction 0.091School Commitment R2 0.885 0.748x2 5,577.98(df) (1,646)AGFI 0.710NFI 0.700NNFI 0.735

All coefficients reported are completely standardized, while significance tests are based onunstandardized coefficients (*p < 0.05; **p < 0.01; ***p < 0.001). The coefficients for autonomyare based on an alternative model with routinization omitted, while the coefficients forroutinization are based on an alternative model with autonomy omitted. Complete results forthese two alternative models are found in Appendices B and C.


pared to the results for the proposed model, since some changes in the

coefficients are observed. In the autonomy model, the coefficients for super-

visor support, role ambiguity, and tenure differ somewhat from the original

model. In the routinization model, differences for supervisor support, work-

load, and tenure are obtained. Notably, the multicollinearity between these

two variables affects some of the coefficients for the determinants of satisfac-

tion and commitment, but has no impact on the estimates of the relationship

between the two employee orientations. Furthermore, the two alternative

models produce only small improvements in the overall fit indices, suggesting

that the multicollinearity problem is only part of the explanation for the

generally poor fit. Since the primary goal of the study is to analyze the

relationship between satisfaction and commitment, and not to perfect a model

of the turnover process, the poor empirical fit is not a critical issue.

DISCUSSION

This study examines the causal relationship between job satisfaction and

organizational commitment in models of employee turnover. The analysis of

four alternative causal relationships between satisfaction and commitment

provides fresh evidence on this unresolved question in the turnover literature.

The finding of no significant effects between satisfaction and commitment over

time raises important theoretical and methodological issues. Additionally, the

results provide strong evidence on the causal influences of proposed determi-

nants of satisfaction and commitment. Unlike existing research, the analysis

undertaken in this study assesses the impact of workplace structures and

individual characteristics on satisfaction and commitment using two-wave

panel data and structural equation models. Both components of the analysis

represent important innovations in research on the relationship between

satisfaction and commitment in models of the turnover process, and the

findings suggest several important points for discussion.

Using SEM procedures and panel data, this study contributes convincing

evidence to the ongoing debate over the causal ordering between satisfaction

and commitment in turnover models (Farkas & Tetrick, 1989; Lance, 1991;

Lincoln & Kalleberg, 1990; Mathieu, 1991; Vandenberg & Lance, 1992). Two

potential explanations seem most plausible to explain the lack of a significant

causal relationship found between satisfaction and commitment. First, the

observed correlation between the two employee orientations may be spurious,

the result of common determinants included in the causal model. Two previous

studies suggest this kind of a spurious relationship (Curry et al., 1986;

Dougherty et al., 1985). The argument has some appeal, given significant

findings of common causes such as routinization, peer support, supervisor

support, and workload.

At the same time, two important points argue against this interpretation

of the results. The correlations between satisfaction and commitment at

both time 1 and time 2 remain significant even with the influences of these


common causes included in the structural equation models. If the relation-

ship between satisfaction and commitment were truly spurious, these

correlations should decline to insignificance. Furthermore, existing analyses

of the relationship between satisfaction and commitment (including one

study using this same data set) provide considerable empirical evidence

for significant positive effects between the two (Bluedorn, 1982; Iverson,

1992; Lincoln & Kalleberg, 1990; Mowday et al., 1982; Mueller et al., 1994;

Price & Mueller, 1986a; Wallace, 1995; Williams & Hazer, 1986). Since most

of these studies involve analysis of cross-sectional data, a counter-argument

to this second point is that the current study indicates that the relationship

simply does not hold up in a more robust longitudinal analysis.

A second potential explanation for the finding of no relationship between

satisfaction and commitment involves a methodological issue, the lag time.

The panel data analyzed in this study spans a time interval of two years. If two

years is not an appropriate timing for the causal effects between satisfaction

and commitment, the analysis may obscure a significant relationship. While

some researchers have argued for longer time intervals in longitudinal studies

of satisfaction and commitment (Farkas & Tetrick, 1989), two years is longer

than previous studies of this causal relationship. Furthermore, the proposed

causal model assumes satisfaction and commitment are endogenous at both

time 1 and time 2, so the analysis cannot fully account for change in these

variables over time as in the analysis of the impact of the proposed determi-

nants on these endogenous variables. Since the relationship between satisfac-

tion and commitment involves cross-wave effects, specification of an

appropriate lag time is more critical to the analysis of this causal relationship

than to the analysis of the causal effects of the determinants on satisfaction

and commitment (Finkel, 1995; Kessler & Greenberg, 1981).

The analysis of the determinants of satisfaction and commitment proposed

in the model also raises important theoretical and methodological issues. Only

four of the proposed determinants significantly influence both satisfaction and

commitment: routinization, peer support, supervisor support, and workload.

Autonomy also has a strong positive effect on satisfaction only. Role ambiguity,

role conflict, and pay have small or insignificant effects on the two employee

orientations. Because measures are not available in the survey data, the

analysis is unable to test the proposition that job involvement positively

influences satisfaction and commitment. The lack of significant findings for

some of the hypothesized determinants indicates the need for future work to

clarify the assumptions supporting these hypotheses. Furthermore, the severe

multicollinearity problem between autonomy and routinization, which re-

quired separating these two variables to estimate the structural equation

models, highlights a need to review the relationship between these two

variables and how the variables are measured (Evans & Fischer, 1992).

Although the two-wave panel analysis provides a more stringent test of the

causal relationship between satisfaction and commitment than existing re-

search, a few methodological limitations of the study suggest directions for

future research on turnover. Like virtually all studies on the turnover process,


the present research relies upon a restricted sample of employees from very

similar organizations (public schools) and a single occupation (teaching) This

study reiterates the need for samples including employees from different jobs,

occupations, industries, sectors, and societies.

The data collection procedures raise two further issues. The length of the

questionnaire likely discouraged respondent participation, contributing to

response rates of less than two-thirds of the eligible respondents at both

survey waves. Scattering the multiple items (presented in Appendix A)

throughout the questionnaire likely lowered the reliability of several mea-

sures, creating threats to measurement validity. Including additional indica-

tors may have improved reliability and validity for some measures, but this

step would mean increasing the length of an already long questionnaire.

Finding a balance between greater reliability (and, hence, validity) and lower

non-response bias is likely to remain an important challenge in surveys of

employee attitudes and behaviors.

The two-wave panel analysis is perhaps the greatest strength of the study,

but this research design also reflects a very important methodological limita-

tion. As previously cited, the two-wave panel data limited the specification of

the relationship between satisfaction and commitment in the model. Empirical

models using three waves of data can incorporate the influences of explanatory

variables at time 2, controlling for the influence of each variable at time 1, on

outcome variables at time 3. Designing and implementing panel studies with

at least three waves of data is an important goal for research on the turnover

process, and could more clearly untangle the alternative relationships between

satisfaction and commitment (Finkel, 1995; Kessler & Greenberg, 1981;

Menard, 1991).

The implication from this study that satisfaction and commitment have a

spurious relationship due to common determinants suggests the need for

revising models of turnover, which often include satisfaction and commitment

as intervening variables. Furthermore, this research suggests a few of the

proposed determinants of satisfaction and commitment often included in

models of turnover may have limited influences on the two employee orienta-

tions. While this study extends existing research on the turnover process,

especially the work of Price and Mueller and their colleagues, further studies

are needed to elaborate the assumptions and conditions linking satisfaction

and commitment.

ACKNOWLEDGMENTS

This research is based on the authors doctoral work completed in the

Department of Sociology at The University of Iowa. The Lloyd A. Fry

Foundation of Chicago provided financial support for collection of the data

reported in this article.


APPENDIX AQuestionnaire Items and Response Categories for

Multiple-indicator Constructs

Autonomy

1. I never control the scheduling of my work.2. I have little to no influence over things that affect me on the job.3. I have a lot of input in deciding what tasks or parts of tasks I will do. (r)

Routinization

1. My job has variety (e.g., always something new to teach).2. The duties in my job are repetitious (e.g., do same lessons over and over). (r)3. I have the opportunity to do a number of different things in my job.

Peer Support

1. My co-workers (i.e., other teachers) can be relied upon when things get tough. (r)2. My co-workers are willing to listen to my job-related problems. (r)3. My co-workers are helpful to me in getting my job done. (r)

Supervisor

1. My administrator can be relied upon when thinks get tough on my job. (r)2. My administrator is willing to listen to my job-related problems. (r)

Role Ambiguity

1. I do not know what my responsibilities are in performing my job. (r)2. I know exactly what is expected of me in my job.3. I know how to get my job done.

Role Conflict

1. I get conflicting job requests from my co-workers (e.g., other teachers). (r)2. Job requests from my administrator and co-workers are often conflicting. (r)

Workload

1. I have to work very fast on the job (e.g., cover a lot of material). (r)2. I do not have enough time to get everything done on my job. (r)3. The workload on my job is too heavy. (r)

Job Satisfaction

1. I find enjoyment in my job. (r)2. Most days I am enthusiastic about my job. (r)3. I am often bored with my job.4. I feel dissatisfied with my job.

Organizational Commitment

1. The school in which I work is the best of all possible places to work. (r)2. I do not care about the fate of the school in which I work.3. I speak highly of the school in which I work to my friends. (r)4. I am proud to tell others I am part of the school in which I work. (r)

Response Categories: Strongly Agree 1; Agree 2; Neither Agree nor Disagree 3; Disagree 4;Strongly Disagree 5.


APPENDIX BLisrel Estimates for the Proposed Model with Autonomy Only

(Satisfaction!Commitment)

All coefficients reported are completely standardized, while significance tests are based onunstandardized coefficients (*p < 0.05; **p < 0.01; ***p < 0.001).

APPENDIX CLisrel Estimates for the Proposed Model with Routinization Only

(Satisfaction!Commitment)

All coefficients reported are completely standardized, while significance tests are based onunstandardized coefficients (*p < 0.05; **p < 0.01; ***p < 0.001).

Dependent VariableExplanatory Variable Job Satisfaction School CommitmentAutonomy 0.442*** 0.030Peer Support 0.126* 0.172**Supervisor Support 0.004 0.604***Role Ambiguity 0.116* 0.006Role Conflict 0.089 0.118Workload 0.126* 0.006Pay 0.081* 0.026Tenure 0.358*** 0.536***Education 0.016 0.066Age 0.073 0.129Marital Status 0.067 0.022Gender 0.008 0.025Job Satisfaction 0.099School Commitment R2 0.749 0.669x 2 4,488.13(df ) (1,325)AGFI 0.728NFI 0.721NNFI 0.757

Dependent VariableExplanatory Variable Job Satisfaction School CommitmentRoutinization 0.567*** 0.236***Peer Support 0.084* 0.113*Supervisor Support 0.057 0.546***Role Ambiguity 0.045 0.094Role Conflict 0.044 0.068Workload 0.228*** 0.016Pay 0.055 0.038Tenure 0.361*** 0.553***Education 0.008 0.044Age 0.071 0.123Marital Status 0.097 0.014Gender 0.003 0.004Job Satisfaction 0.040School Commitment R2 0.841 0.685x 2 4,453.99(df ) (1,325)AGFI 0.727NFI 0.726NNFI 0.763


REFERENCES

Agho, A. O., Mueller, C. W., & Price, J. L. (1993). Determinants of employee job satisfaction:

An empirical test of a causal model. Human Relations, 46, 10071027.

Allen, N. J., & Meyer, J. P. (1990). The measurement and antecedents of affective, continu-

ance, and normative commitment to the organization. Journal of Occupational Psy-

chology, 63, 118.

Anderson, S. E., & Williams, L. J. (1992). Assumptions about unmeasured variables with

studies of reciprocal relationships: The case of employee attitudes. Journal of Applied

Psychology, 77, 638650.

Bacharach, S. B., Bamberger, P., & Conley, S. (1990). Work processes, role conflict, and role

overload: The case of nurses and engineers in the public sector. Work and Occupa-

tions, 17, 199228.

Bateman, T. S., & Strasser, S. (1984). A longitudinal analysis of the antecedents of organiza-

tional commitment. Academy of Management Journal, 27, 95112.

Becker, T. E. (1992). Foci and bases of commitment: Are they distinctions worth making?

Academy of Management Journal, 35, 232244.

Becker, T. E., & Billings, R. S. (1992). Profiles of commitment: An empirical test. Journal of

Organizational Behavior, 14, 177190.

Becker, T. E., Billings, R. S., Eveleth, D. M., & Gilbert, N. L. (1996). Foci and bases of

employee commitment: Implications for job performance. Academy of Management

Journal, 39, 464482.

Beehr, T. A., & Drexler, J. A., Jr. (1986). Social support, autonomy, and hierarchical level as

moderators of the role characteristicsoutcome relationship. Journal of Occupational

Behavior, 7, 207214.

Bentler, P. M., & Bonett, D. G. (1980). Significance tests and goodness of fit in the analysis of

covariance structures. Psychological Bulletin, 88, 588606.

Blau, P. (1960). A theory of social integration. American Journal of Sociology, 65,

545556.

Bluedorn, A. C. (1982). The theories of turnover: Causes, effects, and meaning. In S. B.

Bacharach (Ed.), Research in the sociology of organizations (Vol. 1, pp. 75128).

Greenwich, CT: JAI Press.

Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley.

Brooke, P. P., Russell, D. W., & Price, J. L. (1988). Discriminant validation of measures of job

satisfaction, job involvement, and organizational commitment. Journal of Applied


Cohen, B. P. (1989). Developing sociological knowledge: Theory and method. Chicago:

Nelson Hall.

Cramer, D. (1996). Job satisfaction and organizational continuance commitment: A two-wave

panel study. Journal of Organizational Behavior, 17, 389400.

Cronbach, L. J. (1951). Coefficient alpha and the internal structure of tests. Psychometrika,

16, 297334.

Curry, J. P., Wakefield, D. S., Price, J. L., & Mueller, C. W. (1986). On the causal ordering of

job satisfaction and organizational commitment. Academy of Management Journal,

29, 847858.

DeCotiis, T. A., & Summers, T. P. (1987). A path analysis of a model of the antecedents and

consequences of organizational commitment. Human Relations, 40, 445470.

Dillman, D. A. (1978). Mail and telephone surveys: The total design method. New York: Wiley.

Dougherty, T. W., Bluedorn, A. C., & Keon, T. L. (1985). Precursors of employee turnover: A

multiple-sample causal analysis. Journal of Occupational Behaviour, 6, 259271.


Dunham, R. B., Grube, J. A., & Castaneda, M. B. (1994). Organizational commitment: The

utility of an integrative definition. Journal of Applied Psychology, 79, 370380.

Evans, B. K., & Fischer, D. G. (1992). A hierarchical model of participatory decision-making,

job autonomy, and perceived control. Human Relations, 45, 11691189.

Farkas, A., & Tetrick, L. (1989). A three-wave longitudinal analysis of the causal ordering of

satisfaction and commitment in turnover decisions. Journal of Applied Psychology, 74,

855868.

Farrell, D., & Rusbult, C. E. (1981). Exchange variables as predictors of job satisfaction, job

commitment, and turnover: The impact of rewards, costs, alternatives, and invest-

ments. Organizational Behavior and Human Performance, 28, 7895.

Finkel S. E. (1995). Sage University paper series on quantitative applications in the social

sciences. Causal analysis with panel data. (pp. 07105). Thousand Oaks, CA: Sage.

Glisson, C., & Durick, M. (1988). Predictors of job satisfaction and organizational commit-

ment in human service organizations. Administrative Science Quarterly, 33, 6181.

Han, N. C., Ko, J. W., Price, J. L., & Mueller, C. W. (1995). Organizational commitment in

South Korea. Research and Practice in Human Resource Management, 3, 3968.

Hayduk, L. A. (1987). Structural equation modeling with LISREL: Essentials and advances.

Baltimore: The Johns Hopkins University Press.

Hom, P. W., & Griffeth, R. W. (1995). Employee turnover. Cincinnati, OH: South-Western.

House, J. S. (1981). Work stress and social support. Reading, MA: Addison-Wesley.

Iverson, R. D. (1992). Employee intent to stay: An empirical test of a revision of the

Price and Mueller model. Iowa City, IA: The University of Iowa (unpublished

doctoral dissertation).

Iverson, R. D., & Roy, P. (1994). A causal model of behavioral commitment: Evidence from a

study of Australian blue-collar employees. Journal of Management, 20, 1541.

James, L. R., Mulaik, S. A., & Brett, J. M. (1982). Causal analysis: Assumptions, models, and

data. Beverly Hills, CA: Sage.

Jo, D. G. (1995). Employee attachment among teachers in South KoreaA comparative

analysis with U.S. teachers. Iowa City, IA: The University of Iowa (unpublished

doctoral dissertation).

Joreskog, K. G., & Sorbom, D. (1994). LISREL 8.14. Chicago: Scientific Software

International.

Kahn, R. L., Wolfe, D. M., Quinn, R. P., Snoek, J. D., & Rosenthal, R. A. (1964). Organiza-

tional stress: Studies in role conflict and ambiguity. New York: Wiley.

Kalleberg, A. L. (1977). Work values and job rewards: A theory of job satisfaction. American

Sociological Review, 42, 124143.

Kalleberg, A. L., & Berg, I. (1987). Work and industry: Structures, markets, and processes.

New York: Plenum.

Kalleberg, A. L., & Reve, T. (1992). Contracts and commitment: Economic and sociological

perspectives on employment relations. Human Relations, 45, 11031132.

Kanungo, R. N. (1982). Work alienation. New York: Praeger.

Kessler, R. C., & Greenberg, D. F. (1981). Linear panel analysis: Models of quantitative

change. New York: Academic Press.

Kim, S. W., Price, J. L., Mueller, C. W., & Watson, T. W. (1996). The determinants of

career intent among physicians at a U.S. Air Force hospital. Human Rela-

tions, 49, 947976.

Ko, J. W., Price, J. L., & Mueller, C. M. (1997). Assessment of Meyer and Allens three-

component model of organizational commitment in South Korea. Journal of Applied


Kuhnel, S. (1996). Linear panel analysis of ordinal data using LISREL: Reality or science


fiction? In U. Engel, & J. Reinecke (Eds.), Analysis of change: Advanced techniques in

panel data analysis (Chap. 5, pp. 87112). Berlin: Walter de Gruyter.

Lance, C. E. (1991). Evaluation of a structural model relating job satisfaction, organizational

commitment, and precursors to voluntary turnover. Multivariate Behavioral Re-

search, 26, 137162.

Lawler, E. J. (1992). Affective attachments to nested groups: A choice-process theory. Amer-

ican Sociological Review, 57, 327339.

Lee, T. W., Ashford, S. J., Walsh, J. P., & Mowday, R. T. (1992). Commitment propensity,

organizational commitment, and voluntary turnover: A longitudinal study of organi-

zational entry processes. Journal of Management, 18, 1532.

Lincoln, J. R., & Kalleberg, A. L. (1985). Work organization and workforce commitment: A

study of plants and employees in the U.S. and Japan. American Sociological Review,

50, 738760.

Lincoln, J. R., & Kalleberg, A. L. (1990). Culture, control and commitment: A study of work

organization and work orientations in the United States and Japan. Cambridge:

Cambridge University Press.

Lincoln, J. R., & Kalleberg, A. L. (1996). Commitment, quits and work organization:

A study of U.S. and Japanese plants. Industrial and Labor Relations Review,

50, 738760.

Locke, E. A. (1976). The nature and causes of job satisfaction. In M. D. Dunnette (Ed.),

Handbook of industrial and organizational psychology (pp. 12971349). Chicago:

Rand-McNally.

Mathieu, J. E. (1991). A cross-level nonrecursive model of the antecedents of organizational

commitment and satisfaction. Journal of Applied Psychology, 76, 607618.

Mathieu, J. E., & Farr, J. L. (1991). Further evidence for the discriminant validity of mea-

sures of organizational commitment, job involvement, and job satisfaction. Journal of

Applied Psychology, 76, 127133.

Mathieu, J. E., & Hamel, K. (1989). A causal model of the antecedents of organizational

commitment among professionals and non-professionals. Journal of Vocational Beha-

vior, 34, 299317.

Mathieu, J. E., & Zajac, D. M. (1990). A review and meta-analysis of the antecedents,

correlates, and consequences of organizational commitment. Psychological Bulletin,

108, 171194.

Menard, S. (1991). Sage University paper series on quantitative applications in the social

sciences, Longitudinal research (pp. 776). Newbury Park, CA: Sage.

Meyer, J. P., & Allen, N. J. (1984). Testing the side-bet theory of organizational com-

mitment: Some methodological considerations. Journal of Applied Psychology,

69, 372378.

Meyer, J. P., Allen, N. J., & Gellatly, I. R. (1990). Affective and continuance commitment to

the organization: Evaluation of measures and analysis of concurrent time-lagged re-

lations. Journal of Applied Psychology, 75, 710720.

Mobley, W. H. (1982). Employee turnover: Causes, consequences, and control. Reading, MA:

Addison-Wesley.

Mottaz, C. J. (1988). Determinants of organizational commitment. Human Relations, 41,

467482.

Mowday, R. T., Porter, L. W., & Steers, R. M. (1982). Employeeorganization linkages: The

psychology of commitment, absenteeism, and turnover. New York: Academic Press.

Mowday, R. T., Steers, R. M., & Porter, L. W. (1979). The measurement of organizational

commitment. Journal of Vocational Behavior, 14, 224247.

Mueller, C. W., & Lawler, E. J. (1996). Commitment to different foci: The case of nested


organizational units. Iowa City, IA: Department of Sociology, The University of Iowa

(unpublished manuscript).

Mueller, C. W., & Price, J. L. (1990). Economic, psychological, and sociological determinants

of voluntary turnover. The Journal of Behavioral Economics, 19, 321335.

Mueller, C. W., Boyer, E. M., Price, J. L., & Iverson, R. D. (1994). Employee attachment and

noncoercive conditions of work: The case of dental hygenists. Work and Occupations,

21, 179212.

Mueller, C. W., Price, J. L., & Wynn, T. (1996). Nested organizational units: Differences in the

temporal stability of employee satisfaction and commitment. Paper presented at the

Academy of Management annual meetings, Organizational Behavior division.

Mueller, C. W., Wallace, J. E., & Price, J. L. (1992). Employee commitment: Resolving some

issues. Work and Occupations, 19, 211236.

OReilly, C. O., & Caldwell, D. (1980). Job choice: The impact of intrinsic and extrinsic

factors on subsequent satisfaction and commitment. Journal of Applied Psychology,

65, 559565.

OReilly, C. O., & Caldwell, D. (1981). The commitment and job tenure of new employees:

Some evidence of postdecisional justification. Administrative Science Quarterly, 26,

597616.

Porter, L. W., & Steers, R. M. (1973). Organizational, work, and personal factors in employee

turnover and absenteeism. Psychological Bulletin, 80, 5176.

Porter, L. W., Steers, R. M., Mowday, R. T., & Boulian, P. V. (1974). Organizational commit-

ment, job satisfaction, and turnover among psychiatric technicians. Journal of Applied


Price, J. L. (1977). The study of turnover. Ames, IA: Iowa State University Press.

Price, J. L. (1997). Handbook of organizational measurement. Bradford, UK: MCB

University Press.

Price, J. L., & Mueller, C. W. (1981). A causal model of turnover for nurses. Academy of

Management Journal, 24, 543563.

Price, J. L., & Mueller, C. W. (1986b). Absenteeism and turnover of hospital employees.

Greenwich, CT: JAI Press.

Price, J. L., & Mueller, C. W. (1986). Handbook of organizational measurement. Cambridge,

MA: Ballinger.

Randall, D. M. (1990). The consequences of organizational commitment: Methodological in-

vestigation. Journal of Organizational Behavior, 11, 361378.

Reichers, A. E. (1985). A review and reconceptualization of organizational commitment.

Academy of Management Review, 10, 465476.

Rusbult, C. E., & Farrell, D. (1983). A longitunidal test of the investment model: The impact

on job satisfaction, job commitment, and turnover of variations in rewards, costs,

alternatives, and investments. Journal of Applied Psychology, 68, 429438.

Shore, L. M., Newton, L. A., & Thornton, G. C. (1990). Job and organizational attitudes in

relation to employee behavioral intentions. Journal of Organizational Behavior, 11,

5767.

Smith, P. C., Kendall, L. M., & Hulin, C. L. (1969). The measurement of satisfaction in work

and retirement. Chicago: Rand McNally.

Somers, M. J. (1995). Organizational commitment, turnover, and absenteeism: An exam-

ination of direct and interaction effects. Journal of Organizational Behavior, 16,

4958.

Staw, B. M. (1980). Rationality and justification in organizational life. In L. L. Cummings, &

B. M. Staw (Eds.), Research in organizational behavior (Vol. 2, pp. 4580). Greenwich,

CT: JAI Press.


Steers, R. M. (1977). Antecedents and outcomes of organizational commitment. Administra-

tive Science Quarterly, 22, 4656.

Vandenberg, R. J., & Lance, C. E. (1992). Examining the causal order of job satisfaction and

organizational commitment. Journal of Management, 18, 153167.

Vandenberg, R. J., & Scarpello, V. (1994). A longitudinal assessment of the determinant

relationship between employee commitments to the occupation and the organization.

Journal of Organizational Beha

the causal order of job satisfaction and organizational commitment in models of employee turnover

Documents