Hackett 1985

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 42

ORGANIZATIONAL BEHAVIOR AND HUMAN DECISION PROCESSES 35, 340-381,(1985)

A Reevaluation of the Absenteeism-Job


Satisfaction Relationship
RICK D. HACKETTANDROBERT M. GUION
Bowling Green State University

Previous reviews of the relationship of employee absenteeism to job satis-


faction have largely neglected the size of the relationships reported and the
artifacts that can affect statistical tests of significance. This paper applies the
E L. Schmidt-J. E. Hunter (1977, Journal ofAppLied Psychology, 62, .529-
540) model of validity generalization in assessing the nature and strength of
the relationship of absence to attitudes. Issues concerning the reliability and
validity of absence measures are addressed, correlations between absence and
job satisfaction are compiled and summarized, and an agenda for future re-
search is set out. Considering the reliability estimates reported for the Fre-
quency, Attitudinal, and Time Lost indices, the Time Lost Index was found
to be the most reliable (ru = .66, SD = .28). Factor analyses of intercorre-
lations among absence measures provided tentative support for a voluntary-
involuntary absenteeism distinction. Combining all measures of satisfaction
and all measures of absences, the mean correlation between absence and
attitudes is - .09 (SD = .13). In addition to more comprehensive theory-
guided multivariate research, future studies should aim toward a reconcep-
tualization of absenteeism as a construct to take into consideration the per-
ceptions of the workers themselves. 0 1985 Academic press, ~nc.

Employee absenteeism is a costly personnel problem (Mowday, Porter,


& Steers, 1982, p. 76) that has attracted the attention of managers and
researchers alike. A major focus of the research on this topic concerns
the relationship between absenteeism and employee attitudes. These
studies assess the intuitively appealing proposition that workers who are
less satisfied with their jobs are more likely to be absent from work
(Brayfield & Crockett, 1955; Vroom, 1964). While earlier reviews of the
literature offer general support for this proposition (Herzberg, Mausner,
Peterson & Capwell, 1957; Muchinsky, 1977; Porter & Steers, 1973;
This paper covers the research literature up to July 1983.It was written in partial fulfillment
of formal program requirements for the Ph.D. degree and was supervised by the second
author. We gratefully acknowledge Peter Bycio, C. J. Cranny, Michael Doherty, June Hahn,
Martin Morf, and two anonymous reviewers for their comments on earlier drafts of this
manuscript. Gratitude is also extended to those who provided us with findings from their
unpublished reports, particularly Nigel Nicholson, and Patricia Pelletier for her assistance
and overall support during the study. Finally, the senior author wishes to thank Dr. Robert
Guion for his supervision throughout the project. It was a pleasure and privilege to work
with him. Requests for reprints should be sent to Robert M. Guion, Psychology Department,
Bowling Green State University, Bowling Green, OH 43403.
340
0749-5978/85$3.OO
Copyright 0 1985 by Academic Press. Inc.
All rights of reproduction in any form reserved.
ABSENTEEISM AND JOB SATISFACTION 341

Vroom, 1964), some researchers have cautioned that the relationship is


likely to be indirect and/or tenuous (cf. Nicholson, Brown, & Chadwick-
Jones, 1976; Steers & Rhodes, 1978; Vroom, 1964). The reviews to date
have given only modest regard to the magnitude of the relationships
found. Typically, a summary table is presented merely indicating for each
study whether a positive, negative, or zero relationship between absence
and attitudes was reported (i.e., Clegg, 1983; Muchinsky, 1977; Nicholson
et al., 1976; Porter & Steers, 1973; Porwoll, 1980). Most of the findings
in this area have been based on bivariate correlations, with the magnitude
of the coefficients and tests of significance depending largely on mea-
surement error and sample sizes. A simple summary tabulation of the
number of statistically significant versus nonsignificant findings could
therefore be misleading (cf. Meehl, 1978); a more pragmatic concern is
the actual strength of these relationships. This paper examines the re-
search on absence as it relates to job satisfaction, focusing particularly
on the size of the relationships reported. Through application of the
Schmidt and Hunter model of validity generalization (see Hunter,
Schmidt, & Jackson, 1982) the possibility of moderators of the relation-
ship is also investigated.
The paper is organized into three main sections. First, problems en-
countered in the conceptualization and measurement of absenteeism are
highlighted. Commonly used measures of absence are reviewed and their
reliability and validity addressed. Second, the results of a series of meta-
analyses (Hunter et al., 1982) performed on separate sets of correlations
between various measures of absence and job satisfaction are presented.
Third, methodological shortcomings of the literature are discussed and
suggestions for future research offered.

ABSENTEEISM: ITS CONCEPTUALIZATION AND MEASUREMENT


Definitional Problems
In an earlier review of absence research, Muchinsky (1977) remarked
that “the single most vexing problem associated with absenteeism as a
meaningful concept involves the measure of absenteeism” (p. 317).
Gaudet (1963) noted, for example, that at least 41 different indices of
absence had been used throughout the literature. The unknown psycho-
metric properties of many of these measures and the varying kinds of
information they provide led Lyons (1972) to describe absenteeism re-
search as representing “a hodgepodge of conceptually and operationally
differing definitions” (p. 279). All of these indices, however, were de-
veloped to “tap” two main types of absences: “voluntary” and “invol-
untary.” Since voluntary absences are by definition under the direct con-
trol of individual workers, they are generally thought more amenable to
342 HACKETT AND GUION

change. Accordingly, they have been the focus of most organizational


attempts to reduce absenteeism. Involuntary absences are defined as
being beyond the immediate control of workers and thought to be pre-
cipitated by such factors as transportation problems, family funerals,
sickness, and the like.
Using the terms presented by Chadwick-Jones, Brown, Nicholson, and
Sheppard (1971), the most common purported measure of voluntary ab-
sence is the Frequency Index, defined as simply the number of absences
over a specified time period, excluding holidays and rest days. This index
was introduced by Fox and Scott (1943), who reasoned that voluntary
absences were likely to be of short duration, and that these brief absences
would be best reflected in the number of times a worker was absent,
disregarding the duration of each absence. The most common purported
measure of involuntary absenteeism is the Time Lost Index, defined as
the total number of days absent, excluding rest days and holidays. Here
it is reasoned that only long-term illness is likely to have a heavy impact
on the total day absent of an employee. In contrasting the Time Lost and
Frequency indices, then, an employee with three 2-day absences would
have a Frequency count of 3 and a Time Lost count of 6. Note, however,
that the Time Lost Index could be largely made up of a high frequency
of short absences (i.e., the Time Lost Index would be the same for 2
weeks off work because of measles as for 10 Monday morning hang-
overs- although the Frequency Index in each case would differ). Slight
variations of the Time Lost and Frequency indices have appeared in the
literature. For example, in a study involving employees of a governmental
agency whose policy was to pay workers for accumulated unused sick
leave and vacation leave, Cheloha and Farr (1980) recorded separately
for each leave type (a) the total number of days a worker was absent for
more than 1 hour (Time Lost) and (b) the total number of periods of
absence, where a period of consecutive absences was defined as attrib-
utable to a single cause (Frequency). Hammer, Landau, and Stern (1981)
first classified the absences of workers from a small furniture manufac-
turing firm into voluntary and involuntary, and then recorded for each
type the total hours absent, total days absent, and frequency of absences.
Voluntary absences were defined as those taken for personal reasons or
for which no excuse was given-involuntary as those considered legiti-
mate according to personnel policy (i.e., sickness, funeral attendance).
Another purported measure of voluntary absenteeism commonly used
is the Attitudinal Index, defined as the number of l-day-only absences
accumulated over a stated period of time, excluding holidays and rest
days (Chadwick-Jones et al., 1971); the number of absences of three
consecutive days or less has also been used (cf. Mirvis 8z Lawler, 1977),
as has the total number of unpaid l- and 2-day absences (Terborg, Lee,
ABSENTEEISM AND JOB SATISFACTION 343

Smith, Davis, & Turbin, 1982). Other purported measures of voluntary


absenteeism, much less common than either the Frequency or Attitudinal
Index, include: (a) “other reasons”- the number of days lost within a
specified time period for any reason other than holidays, rest days, and
certified sickness (sometimes referred to as “unexcused absences”); (b)
the Blue Monday Index-the difference between Monday’s and Friday’s
absences per hundred workers averaged over a specified time; and (c)
the Worst Day Index-the difference between the number of absences
on the “worst” day and “best” day of the week, holidays and rest days
excluded (Chadwick-Jones et al., 1971). These latter two indices are
group measures based on the rationale .that since involuntary absences
are equally likely to occur on any given day of the week, large weekly
variations should reflect predominantly voluntary absenteeism (Argyle,
Gardner, & Cioffi, 1958).
Reliability
The literature is not clear on precisely what is measured by an index
of absenteeism. If absence is a behavior that “comes prepackaged in
discrete objectively measurable units” (Nicholson, 1977, p. 236), then
surely the total number of days an employee is absent over a given time
period can be accurately recorded; the error of measurement must be
zero unless there are routine, clerical errors. Any inconsistencies that
appear over time must be inconsistencies in behavior, not inconsistencies
in measurement; perhaps absence is simply an unstable behavior (Latham
& Pursell, 1977).
However, if inferences are drawn from these “objective” indices of
absence to the underlying reasons for the absence (i.e., voluntary or
involuntary), then we become concerned with errors in the measurement
of the underlying construct inferred. Latham and Purse11(1977) stated
that “absenteeism measures are frequently loaded with error vari-
ance . . . as soon as a company attempts to measure absences, reliability
may be attenuated because it is difficult for the recorder to be certain of
the true cause and, hence, correct classification” (p. 369). How to sep-
arate error variance from systematic variance remains unresolved (Ilgen,
1977; Latham & Pursell, 1975, 1977).
Even the objective fact of presence or absence, correctly recorded, can
be a source of error in the measurement of an underlying construct. For
example, it might be argued that absence may be used as a measure of
work avoidance. Work avoidance can be conceptualized as a continuum;
a worker may, at any given instant, have either a weak or strong tendency
to avoid going to work. A threshold can also be conceptualized on that
continuum. Above the threshold, the worker stays away; below it, the
worker goes to work. Some workers who are present may be barely below
344 HACKETT AND GUION

the threshold, and some who are absent may be barely above it. About
the threshold may be a region of error, in which the behavior may be
thought to be random. Even with no recording error, there is some
random error variance in an attendance record used to measure tenden-
cies to approach or avoid work.
The psychometric concept of reliability is therefore not irrelevant to
the evaluation of various absenteeism measures. The common definitions
of reliability include the consistency of measurement across (a) indepen-
dent administrations of a single instrument (test-retest), (b) independent
administrations of presumably equivalent forms of the instrument (alter-
nate forms reliability), and (c) single administration of one form (internal
consistency). A common approach to estimating reliability of measures
of absenteeism combines the logic of a retest with practical problems of
equivalent forms when absences during one time period are correlated
with absences during a subsequent time period. If each time period is a
full year, and if no dramatic changes (e.g., recessions or boom periods)
occurred in one of them, then the two years may be considered essentially
equivalent in that they are subject to the same seasonal and similar
sources of variation in attendance. In contrast, the fourth quarter of one
year (October through December) ordinarily cannot be considered equiv-
alent for attendance purposes to the first quarter (January through March)
of the next year. The lack of equivalent circumstances may, therefore,
be one source of error variance in inferences drawn from records of
absences.
Whereas test-retest reliabilities provide an index of a measure’s sta-
bility over two time periods, split-half test reliabilities provide an index
of a measure’s internal consistency within one administration or, in the
case of absenteeism, within one clearly specified time. Thus, we could
consider weekly absences recorded for each exployee over a 6-month
period as individual observations contributing to a “total 6-month absen-
teeism score.” The internal consistency of the absenteeism data could
then be determined by coefficient (Y or by dividing the observations in
halves (e.g., alternate weeks). (Pearson product-moment correlations be-
tween any two halves need to be corrected by the Spearman-Brown
formula.) As with the test-retest reliabilities, internal consistency esti-
mates can be expected to be lower than those typically reported in the
test construction literature; a person is likely to change less in the time
required for two consecutive multiple choice questions than over two
consecutive weeks.
Method. A search through published and unpublished research on em-
ployee absenteeism was done and reported reliabilities of absence mea-
sures cumulated within the three main absence types-Frequency, At-
titudinal, and Time Lost. To obtain a more stable estimate of the reliability
ABSENTEEISM AND JOB SATISFACTION 345

of each of these measures, mean reliabilities were computed. Fisher


transformations were not used in these computations, nor were negative
reliabilities included.
Summary overview of reliabilities. The reliabilities of the Frequency,
Attitudinal, and Time Lost indices are summarized in Tables 1, 2, and 3.
respectively.
Twenty-seven reliability coefficients were reported for the Frequency
Index. This distribution had a mean of .51 and SD of .18 (median = 51;
range = - .25 to .75). For the Attitudinal Index, 19 coefficients were
found, with the mean of this distribution being .41, SD = .21 (median =
.38; range = - .08 to .75). Twenty-nine reliability coefficients were found
for the Time Lost Index; mean = 66, SD = .28 (median = .75, range
= .17 to .99). The relatively high standard deviations of these distribu-
tions are not surprising, given the differences across studies in both the
type of reliability estimate used and the length of the time periods for
recording absences. While the variability of the distributions prevents

TABLE I
RELIABILITIESFORTHE FREQUENCY INDEX
r Time interval
Test-retest
Huse and Taylor (1962) .61 Two l-year periods
Chadwick-Jones ef al. (1971) .43 Two l-year periods
Bernardin (1976) .37 (7) Three i-year periods
Nicholson and Goodge (1976) .57 Two l-year periods
Waters and Roach (1979) .62 Two l-year periods
Breaugh (1981) .62 (7))” Four l-month periods
Hammer and Landau (1981)b .58 Two IS-month periods
Clegg (1983) .54 (V) Four l-month periods
Spearman-Brown
Turner (1960) .74 (Plant 1) 12 months
.60 (Plant 2)
Farr et al. (1971) .39
Nicholson (1975)’ -.25 (Plant I) .28 52 weeks
-.20 .39
.66 .65
.46 .7s
.47 .66
.23 .oo
.51 .43
-.25 .69 (Plant 16)
---
a Average interperiod r.
b In this study absences were classified into voluntary vs involuntary and frequency
counts made within each class. The reliability estimate reported is for voluntary absences.
c In this study 16 different companies of four contrasting technologies were sampled.
346 HACKETTANDGUION

TABLE2
RELIABILITIESFORTHE A~ITUDINAL INDEX
r Time interval

Test-retest
Huse and Taylor (1962) .52
Chadwick-Jones et al. (1971) .38
Mirvis and Lawler (1977) .49 (7))” Three l-month periods
Spearman-Brown
Nicholson (1975)b .45 (Plant 1) .37 52 weeks
.08 .lO
.66 .36
.35 .75
.55 .66
.I8 - .08
.36 .69
.06 .41 (Plant 16)
a Average interperiod r.
b In this study 16 different companies of four contrasting technologies were sampled.

firm conclusions, it does appear that all three indices are quite unreliable.
Finding that the highest mean reliability to be associated with the Time
Lost Index is contrary to the generally held notion that the Frequency
Index is the most reliable measure of absenteeism currently available (cf.
Hammer & Landau, 1981; Johns, 1978; Muchinsky, 1977).
Validity
Do absence measures reflect what they are purported to measure-
namely, the underlying voluntary or involuntary nature of the absence?
If, for example, it is to be argued that the Frequency Index reflects pre-
dominantly “voluntary” absences, then this index ought to correlate
higher with other measures of voluntary absenteeism than with measures
of involuntary absence. In an earlier review of the literature, Muchinsky
(1977) remarked that “there are no known studies that directly assessed
the validity of absenteeism measures” (p. 319). Since his review, only
one such investigation has been done. Chadwick-Jones, Nicholson, and
Brown (1982) devoted a full chapter of their recent book to assessing the
convergent and discriminant validity of various indices of absence-
largely using data collected previously by Nicholson et al. (1976).
Twenty-one separate intercorrelation matrices of absence measures,
based on data collected over 52 weeks from 21 different organizations,
were examined. In general, correlations between the Attitudinal and Fre-
quency indices were found to be consistently statistically significant and
higher than those between Frequency and Time Lost which in turn tended
to be higher than the correlations between Attitudinal and Time Lost (cf.
ABSENTEEISM AND JOB SATISFACTION 347

TABLE 3
RELIABILITIESFORTHE TIME LOST INDEX
r Time interval
Test-retest
Huse and Taylor (1962) .23 Two l-year periods
Chadwick-Jones et al. (1971) .19 Two l-year periods
Nicholson and Goodge (1976) 99 Two l-year periods
Ilgen and Hollenback (1977) .17 (7) Three 6-month periods
Breaugh (1981) .42 (T)” Four l-year periods
Split-half
Terborg et al. (1982)b 55 (Store 1). .62 11 weeks
.20 .33
.40 .52 (Store 6)
Spearman-Brown
Rosensteel (1953) .88
Nicholson (1975)” .75 (Plant 1) .97 52 weeks
.68 .94
.75 .80
.17 .95
.88 .97
.86 .92
.73 .85
.89 .95 (Plant 16)
Coefficient (Y
Adler and Golan (1981) .64 12 months
a Average interperiod r.
b Data were collected from six separate retail stores of a national retail-sales organization.
c In this study 16 different companies of four contrasting technologies were sampled.

Chadwick-Jones et al., 1982, p. 148). The higher correlations between


the Frequency and Attitudinal indices, compared to those between Fre-
quency and Time Lost, were interpreted as supporting the contention
that the Attitudinal and Frequency indices measure the same construct-
“voluntary absenteeism.” The slightly higher correlations of Frequency
with Time Lost, compared to those between the Time Lost and Attitu-
dinal indices, were explained by the greater reliability of the Frequency
Index compared to the Attitudinal Index. Statistically significant corre-
lations of both the Attitudinal and Frequency indices with the Time Lost
Index were accepted as due to overlapping data. Where the Worst Day
and Blue Monday indices (both purported measures of voluntary absence)
were included in the matrices, the former was observed to correlate highly
with the Attitudinal Index, but no consistent pattern emerged for the
latter. Treating organization as the unit of analysis, correlations between
348 HACKETT AND GUION

six absence measures and the yearly turnover rates of 16 companies were
also examined. The only significant correlation obtained was with the
Attitudinal Index, and this was interpreted as suggesting that short-term
absences and labor turnover may have a common cause.
Another way to determine the constructs underlying various indices of
absence would be to intercorrelate these measures and then to do ex-
ploratory factor analyses of the resulting matrices. Such a procedure is
mathematically questionable because of overlapping data among some of
the absence measures (e.g., the Attitudinal Index is wholly contained
within the Frequency Index), but factor analyses would help to generate
hypotheses about what is being “tapped” by these indices. If several
such analyses generate similar hypotheses, the greater the confidence we
can have in them. There are three studies that collected absence data
from a single sample over two consecutive years and presented separate
correlation matrices for each year (cf. Chadwick-Jones et al., 1970; Huse
& Taylor, 1962; Nicholson & Goodge, 1976). These studies therefore not
only permitted exploratory factor analyses to be done, but also allowed
for an assessment of the stability across years of the resulting factor
patterns.
A factor-analytic assessment of the validity of absence measures:
Method. Before the matrices were factored, the correlation coefficients
were corrected for the unreliability of both measures in each correlation,
using the reliability estimates reported. Analyses were also done on cor-
relations corrected for the unreliability of only the measure with the
lowest reliability. Although less clear, the factor patterns that emerged
were similar to those obtained with correlations made for the unreliability
of both measures. Corrections for unreliability were done because of the
wide variations in reliabilities for absence measures. The principal axes
procedure was used, with squared multiple correlations as communality
estimates. Since these estimates were 1.0 in all cases-not surprising in
view of the considerable overlap of data among absence measures-the
unrotated factors are identical to those that would be obtained through a
principal components analysis. The number of factors extracted from
each unrotated solution was decided upon through combined consider-
ation of the scree test (Cattell, 1966), parallel analyses (Montanelli &
Humphreys, 1976), and “interpretability” (Rummel, 1970, p. 362). Since
it would be unrealistic to expect any voluntary or involuntary absence
factors that might have emerged to be orthogonal to one another, all
unrotated solutions were rotated to oblique simple structure by the
Promax procedure (Hendrickson & White, 1964). Finally, similarity in
the pattern and magnitude of the factors across years was assessed by
using the coefficient of congruence (Harman, 1967, p. 269). This coeffr-
ABSENTEEISM AND JOB SATISFACTION 349

cient is interpreted like a correlation coefficient, with high factor stability


indicated by a coefficient of 90 or greater. All computations used the
Statistical Analysis System (cf. Helwig & Council, 1979).
Factor-analytic assessment of validity: Findings. The results of these
analyses are presented in Tables 4 through 6.
In the Huse and Taylor (1962) data set, the two unrotated factors from
the 1957 and 1958 solutions accounted, respectively, for 99 and 96% of
the total variance in the correlation matrices. Table 4 shows for both
solutions that the Attitudinal and Frequency indices loaded on a common
factor, and that the Time Lost Index loaded on a different one. The
coefficients of congruence for the first and second factors of each solution
were .99 and .92, respectively.
Chadwick-Jones et al. (1971) included a Lateness index to assess its
relationship with the various absence measures. The three unrotated fac-
tors retained from the analyses of the 1968 and 1969 data sets accounted
for 99 and 91%, respectively, of the variance in the correlation matrices.
Table 5 shows that the Attitudinal, Frequency and “Other” indices-all
purported measures of voluntary absenteeism-loaded heavily on the
same factor in each solution (coefftcient of congruence = .96). Again,
the Time Lost Index loaded on a separate factor, although one which was
not stable enough to warrant interpretation (coefficient of congruence =
.77). It can also be seen that there was little consistency in the loadings
of the Lateness index.
The analyses using the Nicholson and Goodge (1976) data resulted in
three factors for both 1971 and 1972, accounting for 89 and 84% of the
total variances, respectively. All three factors demonstrated high stability
across years. Specifically, the coefficients of congruence between the

TABLE 4
PROMAXROTATED FACTOR PATTERN MATRICES OF FOUR ABSENCE MEASURES
(HUSE & TAYLOR, 1962)

1957 Solution 1958 Solution

Factor: 1 2 ,,2” 1 2 h?” P


Frequency .30 .82 .95 .47 .65 .89 .61
Time Lost .94 .15 1.0 .94 .09 .97 .52
Attitudinal -.16 1.0 .98 -.12 1.0 .97 .23
Medical 1.00 -.13 1.00 1.oo - .I2 1.oo .19

Note. Reliabilities < communalities because correlations were corrected for attenuation
due to unreliability in measurement. Interfactor correlations: (1957) .35, (1958) .41.
(1Final communality estimates.
b Interyear reliabilities.
TABLE 5
PROMAX ROTATED FACTOR PATTERN MATRICES OF SIX ABSENCE MEASURES AND A LATENESS INDEX (CHADWICK-JONES ET AL., 1971)

1968 Solution 1969 Solution

Factor: 1 2 3 h*” 1 2 3 h*” P F

Frequency .91 -.08 .22 1.0 .89 -.05 .43 E


.29 1.00
Attitudinal 1.00 - .09 -.04 .99 .95 - .09 .lO .92 .38 2
Other .90 .03 -.18 1.00 .89 .22 -.05 .97 .27
Worst Day .71 .52 -.19 .a7 .44 -59 - .31 .7Q .21
zi
Time Lost .18 .16 .83 1.00 .44 .03 .69 .93 .19
Late .06 .93 - .02 .89 - .Ol .03 .97 .95 .16 z
Blue Monday -.15 29 .22 1.00 -.14 .92 .27 .a7 .Ol
E
Note. Reliabilities -C communalities because correlations were corrected for attenuation due to unreliability in measurement. Interfactor
correlations: (1968) 1 & 2 = .39, 1 & 3 = .49, 2 & 3 = .47; (1969) I & 2 = .40, 1 & 3 = .14, 2 & 3 = .42.
a Final communality estimates.
b Interyear reliabilities.
TABLE 6
PROMAX ROTATED FACTOR PATTERN MATRICES OF FOUR ABSENCE MEASURES AND A LATENESS INDEX
(NICHOLSON & GOODGE, 1976)
iii
1971 Solution 1972 Solution
E
Factor: 1 2 3 h2” 1 2 3 h2” P
s
t;
Time Lost -.04 - .os .99 .96 -.04 - .07 .98 .93 99 z
Frequency .91 .04 .20 .99 .78 - .08 .29 .85 .57
HolidayC .85 -.16 - .23 .64 .84 -.I8 -.24 .62 NG %
u
Unsanctionedd .89 .I1 .05 .86 .87 .10 .05 .83 NG
Latee - .02 .99 - .05 .97 - .02 ,953 - .07 .96 NG E
Nore. Reliabilities < communalities because correlations were corrected for attenuation due to unreliability in measurement. Interfactor g
correlations: (1971) 1 & 2 = .16, 1 & 3 = .28. 2 & 3 = .Ol; (1972) I & 2 = .l7, 1 & 3 = .29, 2 & 3 = .06. =!
0 Final communality estimates. Yi
b Interyear reliability.
[ A count of the total number of absences of less than 3-week duration occurring before. over. or immediately following a holiday, for the 2- 5
year period.
’ A count of the number of absences unaccompanied by a medical certificate over the 2-year period.
’ A count of the total number of lateness incidents over the 2-year period.
352 HACKETT AND GUION

first, second, and third factors across the two solutions were .99 in each
case. Factors 1 and 3 clearly separate the Time Lost Index from purported
measures of voluntary absence. It is also noteworthy that “Lateness”
defined its own factor.
Summary overview of validities. Overall, the results of these analyses,
taken together with the findings presented by Chadwick-Jones et al.
(1982), suggest that voluntary and involuntary absences can be distin-
guished empirically. Purported measures of voluntary absence loaded to-
gether on one factor, and the Time Lost Index (considered a measure of
involuntary absence) had its heaviest loading on another. This separation
was particularly evident in the Huse and Taylor (1962) and Nicholson and
Goodge (1976) solutions where the two factors demonstrated especially
high stability across years. In the two studies where both the Attitudinal
and Frequency indices were used (Huse & Taylor, 1962; Chadwick-Jones
et al., 1971), they had loadings of .90 or greater on the same factor and
may be considered measures of the same thing, corroborating the Chad-
wick-Jones et al. (1982) findings. These analyses, however, yielded little
support for considering the Worst Day Index and Blue Monday Index
useful measures of voluntary absenteeism. It is worth noting that the
separate lateness factor in the Nicholson et al. (1976) solution suggests
that absenteeism and lateness, rather than representing different levels
of the same withdrawal tendencies, measure empirically distinguishable
concepts (cf. Adler & Golan, 1981).
ABSENCE AS RELATED TO JOB SATISFACTION
Earlier reviews of the relationship of absence and job satisfaction have
typically given little attention to the size of the correlations reported or
to possible moderating effects of other “third factor” variables. We shall
consider both using meta analysis (Glass, McGaw, & Smith, 1981; Hunter
et al., 1982; Rosenthal, 1978). The method of meta-analysis chosen for
this paper is that introduced by Schmidt, Hunter, and their colleagues
under the rubric of “validity generalization” (Hunter et al., 1982;
Schmidt & Hunter, 1977; Schmidt, Hunter, Pearlman, & Shane, 1979).
Method
A major premise of the validity generalization model holds that a sub-
stantial proportion of variance in a set of correlations reported across
studies often can be explained when various statistical artifacts are con-
sidered. These sources of artifacts include (a) unreliable data becasue of
small sample sizes, (b) differences among studies in degree of range re-
striction, (c) differences among studies in the unreliability of each mea-
sure going into the correlation, (d) differences in factor structures be-
tween different tests thought to measure similar constructs, (e) differ-
ABSENTEEISM AND JOB SATISFACTION 353

ences among studies in amount and kind of criterion contamination and


deficiency, and (f) computational and typographical errors (Hunter et al.,
1982). Hunter et al. (1982) provide computational formulas for estimating
the variance attributable to the first three of these sources. To apply their
procedure to the absenteeism literature, a distribution of correlation coef-
ficients between similar measures of absence and similar measures of job
satisfaction must be compiled and the variance of this distribution
(weighted by sample size) computed. If this “observed variance” ap-
proaches zero after the estimated variance attributable to the first three
artifacts listed above is subtracted, the relationship under consideration
is the same across the various studies. The sample-size weighted mean,
corrected for restriction in range and measurement error, is the best es-
timate of the population value of the relationship between the two con-
structs as generally measured.
Sample. The studies included were restricted to those in which anal-
yses were done at the individual level. Since published research could be
biased in favor of statistically significant findings (Bakan, 1967; Mc-
Nemar, 1960; Rosenthal & Gaito, 1963; Sterling, 1959), an effort was
made to include the results from unpublished studies. Manual and com-
puter searches (keyword = employee absenteeism) of the Psychological
and Dissertation abstracts were done as well as complete searches of
references cited in relevant articles. Where studies did not specifically
report the magnitude of relationships found between measures of absence
and job satisfaction (i.e., Nicholson et al., 1976), letters were written
requesting this information. A total of 31 studies were found.
Procedure. The validity generalization procedure was applied to the
correlation coefficients presented in Tables 7 through 13, sets of corre-
lations based on the same kind of absence measure and job satisfaction
facet. Also, overall analyses were done for each facet of job satisfaction
across all absence measures, and for all correlations across all measures
of absence and all measures of job satisfaction.
Several researchers failed to report the reliabilities of the satisfaction
measures they used; corrections for restriction in range require that pop-
ulation variances in satisfaction scores be known or estimated and not
enough information was available to make these estimates. We decided,
therefore, not to correct the mean and variance of each of these distri-
butions for unreliability or restriction in range in job satisfaction; correc-
tions were made for sampling error and for unreliability in absence mea-
sure. Also, Terborg et al. (1982) cautioned that corrections made for un-
reliability and restriction in range in attitude measurement “might
actually be controlling for differences in the situation” (p. 447). They
noted that job attitudes are known to be correlated with features of the
work environment and that unreliability in job satisfaction measures and
354 HACKETT AND GUION

restriction in range could occur because of such things as changes in


group membership or increased awareness of alternatives. Accordingly,
corrections for these artifacts could result in the false conclusion that
there are no situational effects. Corrections to the sample-size weighted
mean and variance of each distribution, for unreliability of absence mea-
sures, were based on the means of the reliabilities reported for each index
(see Tables l-3). Corrections for sampling error only required knowledge
of the correlation coefficients reported and the sample sizes on which
they were based. Since some studies reported more than one correlation
there was some lack of independence in the data. Although the size of
this problem cannot be estimated quantitatively, its effect is to lead to
some underestimation of the adjustment for sampling error. Yet, Hunter
et al. (1982) stated that “if the number of values contributed by each
study is small in comparison to the number of studies, then there is little
error in the resulting cumulation” (p. 118). While the typical absenteeism
study presented several correlations between various measures of ab-
sence and job satisfaction, the problem of lack of independence among
coefficients was not serious since most of the analyses were done on
separate sets of absence-job satisfaction measurement combinations. In
none of the individual sets were more than two correlations contributed
by any one study (see Tables 7-12). Although Nicholson et al. (1976)
contributed more than two correlation coefficients, they were based on
data collected from independent samples. For the analysis across all mea-
sures of absence and all measures of job satisfaction, there is probably a
more substantial underestimation of the variance attributable to sampling
error.
Pearlman, Schmidt, and Hunter (1980) suggested as a “rule of thumb”
that the situational specificity hypothesis be rejected when 75% of the
variance can be accounted for by the four artifacts for which they pro-
vided computational formulas. Since with the present data set corrections
were made for only two of these artifacts, this 75% criterion was adjusted
downward. However, the adjustment should be small because sampling
error typically accounts for the largest proportion of artifactual variance,
with relatively small amounts of variance thereafter being accounted for
by each of the other artifacts (Pearlman et al., 1980). It was decided that
a 70% criterion would be used.
Results
The results of these analyses are summarized in Table 14. The best
estimate of the relationship between the constructs underlying various
measures of absence and job satisfaction is fi, which in this case is the
sample-size weighted mean corrected for unreliability in the measurement
of absenteeism. The highest 6 value obtained was - .23, between mea-
ABSENTEEISM AND JOB SATISFACTION 355

sures of overall job satisfaction and the Attitudinal Index. Of the job
satisfaction facets, j? values were highest for Work (- .16 with the Fre-
quency Index) and Supervision ( - . 11 with the Frequency Index). For the
combined distribution of correlations between all measures of absence
and all measures of job satisfaction, fi was - .09. This corrected mean
was based on 707 correlations and, treating each coefficient as having
come from an independent sample, an overall N of 62,308. The Nicholson
(1975) study, wherein 16 companies were sampled using three measures
of absence and 11 measures of job satisfaction (five JDI facets plus JDI-
Total, and five facets of the Hoppock scale), contributed 528 of these
correlations.
Using the 70% criterion the situational specificity hypothesis can be
rejected for 9 of the 25 relationships considered in Table 14; 4 involving
satisfaction with supervision and 2 involving satisfaction with work (see
Table 14). Yet, the fi values are so small that in all but three of these
cases (Attitudinal Index and Co-workers, Frequency Index and Super-
vision, Attitudinal Index and Supervision) the 95% confidence interval
included zero; this was also true for all the fi values for which the situ-
ational specificity hypothesis was not rejected. The mean proportion of
variance accounted for by artifacts, computed across the 18 separate
distributions of absence-job satisfaction correlations, was -69.
The mean SD computed across the individual distributions of absen-
teeism measurement combinations (. 142) approximated the mean SD
computed over the distributions for which absence measures had been
combined (. 144). This suggests that little variation in reported correlations
between absence and satisfaction can be attributed to type of absence
measure used, whether voluntary or involuntary. In none of the combined
distributions did corrections for differences between studies in unreli-
ability of absence measure account for more than 4% of the total esti-
mated artifactual variance.
Discussion
The results of these analyses have shown that typically less than 4%
of the variance in absence measures is associated with satisfaction scores.
These findings are consistent with those reported by Terborg et al. (1982)
in their application of the validity generalization model to data they col-
lected from replications across six retail stores of one national retail-sales
organization. Their highest mean correlation between absence and job-
facet satisfaction, corrected for unreliability of both absence and job sat-
isfaction measures, was - .37 (between Time Lost Index and JDI-Pay).
Corrected mean correlations associated with the other facets of job sat-
isfaction were all reported to be - .20 or less.
While both here and in Terborg et al. (1982) all corrected mean cor-
356 HACKETT AND GUION

TABLE 7
CORRELATIONS BETWEEN SATISFACTION WITH CO-WORKERS AND THE FREQUENCY,
TIME LOST, AND AITITUDINAL INDICES

Satisfaction
Study Subjects N measure r
Frequency
Hackman and Telephone repairmen 208 Single
.15*
Lawler (1971) 7-pt item
Waters and Female clerks of 131 JDI-Co”
- .18*
Roach (1971) insurance agency
Nicholson (1975) Female sewing 146 JDI-Co”; .02 - .Ol
machine operators 66 Hoppockb .06 .02
96 - .25* - .29**
103 .lO - .Ol
Male production 102 - .23* -.I5
workers 58 - .Ol .oo
62 .02 .06
20 .33 .16
Male chemical 73 .09 .15
process operators 61 .07 .18
85 - .02 -.20
80 .08 .06
Male bus drivers 76 - .08 - .04
and conductors 63 -.21 .09
58 -99 -.06
73 -.I0 -.19
Time Lost
Smith et al. Male technicians 98 JDI-Co
- .38**
(1969)
Watson (1981) Production workers 116 JDI-Co - .09
Adler and Female telephone 114 JDI-Co
- .05
Golan (1981) operators
Terborg et al. Retail-sales 23 JDI-Co .18
(1982) employees 41 -.15
(33% males) 46 - .38
42 .17
38 - .21
52 .04
Nicholson (1975) Female sewing 146 JDI-Co; .ll .03
machine operators 66 Hoppock .13 .14
96 - .24* -.20
103 - .03 -.lO
Male production 102 - .20* - .05
workers 58 - .02 .07
LIJob Descriptive Index (Smith er al., 1969). Co = co-workers.
b Self-developed Likert scales.
* p < .05.
** p -c .Ol.
ABSENTEEISM AND JOB SATISFACTION 357

TABLE I-Continued
Satisfaction
Study Subjects N measure r
62 - .07 .20
20 .08 - .Ol
Male chemical 73 .08 .I4
process operators 61 .09 .I6
85 .lO - .20
80 .I4 .I7
Male bus drivers 76 .03 .07
and conductors 63 - .46** -.20
58 - .25 - .0x
13 .03 - .20
Attitudinal
Nicholson (1975) Female sewing 146 JDI-Co; .oo - .O?
machine operators 66 Hoppock .03 - .O?
96 -.21* - .31**
103 .I0 .06
Male production 102 - .20* - .I7
workers 58 .oo - .02
62 - .03 - .09
20 .24 .09
Male chemical 73 .04 .05
process operators 61 .05 .20
85 - .04 -.I8
80 .06 .05
Male bus drivers 76 - .I0 .09
and conductors 63 -.I6 .I0
58 - .O? - .06
73 - .I0 ~ .?O

relations were negative, the best estimate of the strength of the relation-
ship is that is is very weak (Nicholson et al., 1976). Moreover, the highest
corrected means (those with overall measures of job satisfaction) were
associated with the highest standard deviations; on average, only 40% of
the variance of these distributions could be accounted for by the two
artifacts for which corrections were made. While the high residual vari-
ance might well be population variance, it may also be largely attributable
to the variety of measures of overall job satisfaction that have been used.
There has been less standardization in the measurement of overall job
satisfaction than in measuring the facets.
Chadwick-Jones et al. (1982) reported that the predictability of absence
from work attitudes varied considerably across the 16 plants they inves-
tigated; in most plants absence and attitudes were unrelated, while in
others attitudes were useful predictors. These authors, along with Johns
358 HACKETT AND GUION

TABLE 8
CORRELATIONS BETWEEN SATISFACTION WITH PAY AND THE FREQUENCY, TIME LOST, AND
ATTITUDINAL INDICES

Satisfaction
Study Subjects N measure r
Frequency
Hackman and Telephone repairmen 208 JDS” item
.02
Lawler (1971)
Waters and Female clerks of 131 JDI-Pay’
-.09
Roach (1971) insurance agency
Nicholson (1975) Female sewing 146 JDI-Payb; .13 .02
machine operators 66 Hoppockc .16 .12
96 -.1.5 -.lO
103 -.14 -.ll
Male production 102 .06 - .03
workers 58 .04 .15
62 -.20 -.24
20 .33 .08
Male chemical 73 .12 .0.5
process operators 61 .19 .02
85 -.lO -.14
80 -.07 -.19
Male bus drivers 76 .30** -.lO
and conductors 63 .18 .07
58 .16 .34*
73 - .25* -.06
Time Lost
Smith et nl. Male technicians 98 JDI-Pay - .24**
(1969)
Adler and Female telephone 124 JDI-Pay
-.I0
Golan (1981) operators
Watson (1981) Production workers 116 JDI-Pay .04
Terborg et al. Retail-sales 23 JDI-Pay - .49*
(1982) employees 41 - .Ol
(33% males) 46 -.21
42 .14
38 - .42*
52 -.14
Nicholson (1975) Female sewing 146 JDI-Pay; - .Ol -.13
machine operators 66 Hoppock .21 .21
96 .02 .04
103 -.04 -.14
Male production 102 -.05 -.07
workers 58 - .06 - .04

a Job Diagnostic Survey (Hackman & Lawler, 1971; Hackman & Oldham, 1975).
* Job Descriptive Index (Smith et al., 1969).
’ Self-developed Likert scales.
*p < .05.
**p < .Ol.
ABSENTEEISM AND JOB SATISFACTION 359

TABLE %-Continued
Satisfaction
Study Subjects N measure r
62 .lO - .09
20 .45* .05
Male chemical 13 .25* .07
process operators 61 - .09 -.19
85 -.Ol .03
80 -.I5 -.I7
Male bus drivers 76 - .07 - .24*
and conductors 63 - .I9 -.I4
58 -.13 .I9
73 - .09 - .05
Attitudinal
Nicholson (1975) Female sewing 146 JDI-Pay; .15 .08
machine operators 66 Hoppock .13 .09
96 - .18 - .13
103 - .09 - .03
Male production 102 .16 .03
workers 58 .03 .16
62 - .29* - .29*
20 .06 .I2
Male chemical 73 .06 .oo
process operators 61 .24 .I1
85 - .05 -.I8
80 .06 .05
16 .33* - .06
63 .23 .I?
58 .20 .32*
73 - .23 - .02

and Nicholson (1982), interpreted these findings as suggesting that the


plants differed in their “causal climates.” To the contrary, the results
reported here suggest that most of the variability can be accounted for
by statistical artifacts. Although situational specificity was rejected in
only nine sets of correlations, and although even these may be due to
low power (Osburn, Callender, Greener, & Asworth, 1983), there is little
in these data to support continued belief that job satisfaction is an im-
portant correlate of absenteeism even in special situations. The mean
proportion of variance accounted for by artifacts, computed across the
18 distributions, was .69; on average, not much variance is left over to
allow for situational moderators to operate-especially since corrections
were made for only two of the seven possible sources of artifactual vari-
ance.
The distributions of correlations for which we can have greatest con-
360 HACKETT AND GUION

TABLE 9
CORRELATIONS BETWEEN SATISFACTION WITH PROMOTION AND THE FREQUENCY, TIME
LOST, AND ATTITUDINAL INDICES

Satisfaction
Study Subjects N measure r
Frequency
Hackman and Telephone repairmen 208 Select JDS”
.09
Lawler (1971) items
Waters and Female clerks of 131 JDI-Promb
.05
Roach (1971) insurance agency
Nicholson (1975) Female sewing 146 JDI-Prom; - .32** -.12
machine operators 66 Hoppock’ -.lO .04
96 - .07 .ll
103 -.I5 -.ll
Male production 102 -.ll - .07
workers 58 .15 .I5
62 -.12 - .33
20 .25 .22
Male chemical 73 -.12 -.12
process operators 61 .18 .09
85 .08 - .08
80 - .09 -.14
Male bus drivers 76 .07 -.16
and conductors 63 .Ol .lO
58 -.09 .25
73 -.20 - .23
Time Lost
Watson (1981) Production workers 116 JDI-Prom .03
Adler and Female telephone 116 JDI-Prom - .15
Golan (1981) operators
Terborg et al. Retail-sales 23 - .03
(1982) employees 41 -.27
(33% males) 46 -.lO
42 .22
38 - .20
52 .08
Nicholson (1975) Female sewing 146 JDI-Prom; -.18* - .08
machine operators 66 Hoppock -.16 .09
96 .lO .18
103 - .20* - .31**
Male production 102 -.18 -.16
workers 58 .04 -.06
62 -.19 - .41**

0 Job Diagnostic Survey (Hackman & Lawler, 1971; Hackman & Oldham, 1975).
b Job Descriptive Index (Smith et al., 1969). Prom = promotion.
c Self-developed Likert scales.
*p < .05.
** p < .Ol.
ABSENTEEISM AND JOB SATISFACTION 361

TABLE !9-Continued

Satisfaction
Study Subjects N measure r

20 .59*+ .36
Male chemical 73 - .Ol - .05
process operators 61 .15 .02
85 .05 .05
80 - .24* -.20
Male bus drivers 76 ~ .07 - .31**
and conductors 63 - .Ol .07
58 .I3 .29*
73 .I4 - .08
Attitudinal
Nicholson (1975) Female sewing 146 JDI-Prom; - .25** - .09
machine operators 66 Hoppock - .08 .03
96 -.13 .08
103 -.I0 - .03
Male production 102 - .Ol .Ol
workers 58 .ll .I2
62 - .09 - .21
20 - .09 .08
Male chemical 73 -.I4 -.I4
process operators 61 .17 .II
85 .12 - .03
80 .06 .03
Male bus drivers 76 - .06 - .I0
and conductors 63 .05 .I2
58 - .09 .22
13 -.19 - .24*

fidence in rejecting the situational specificity hypothesis are those asso-


ciated with satisfaction with supervision. For these an average of 93% of
the variance could be accounted for by two artifacts (for the distributions
involving the Attitudinal and Frequency indices, 100% of the variance
could be accounted for). This is consistent with the report of Terborg et
al. (1982) that 100% of the variance in correlations between the Time
Lost Index and satisfaction with supervision could be accounted for by
four artifacts. The results of the analyses done here suggest that the
correlation between satisfaction with supervision and absence, indepen-
dent of other situational variables or absence measure used, will hover
around - .08, or perhaps lower. Terborg et al. (1982) reported a corrected
mean correlation of - .Ol between the Time Lost Index and satisfaction
with supervision which compares with a corrected mean correlation of
- .05 found between these same two indices in the present analysis.
Overall, the results of the meta-analyses are consistent with a growing
362 HACKETT AND GUION

TABLE 10
CORRELATIONS BETWEEN SATISFACTION WITH SUPERVISION AND THE FREQUENCY,
TIME LOST, AND ATTITUDINAL INDICES

Satisfaction
Study Subjects N measure r
Frequency
Hackman and Telephone repairmen 208 Select
- .07
Lawler (1971) JDS” items
Waters and Female clerks of 131 JDI-Supb
.05
Roach (1971) insurance agency
Bemardin (1976) Police officers 216 JDI-Sup -.lO
Johns (1978) Paper-products 208 JDI-Sup
plant employees -.14*
(24% females)
Breaugh (1981) Research scientists 112 Select
.oo
(42 females) JDS items
Nicholson (1975) Female sewing 146 JDI-Sup; - .08 -.14
machine operators 66 HoppockC .04 .16
96 - .05 -.I3
103 - .09 -.17
Male production 102 -.14 - .06
workers 58 .04 -.04
62 -.20 - .23
20 .oo -.12
Male chemical 73 .05 .02
process operators 61 .12 .lO
85 -.I3 - .09
80 -.lO -.I1
Male bus drivers 76 -.13 - .30**
and conductors 63 -.17 .04
58 -.18 .25
73 - .26 - .23
Time Lost
Smith er al. Male technicians 98 JDI-Sup
- .3t3**
(1969)
Adler and Female telephone 115 JDI-Sup
-.05
Golan (1981) operators
Breaugh (1981) Research scientists 112 Select
-.19*
(42 females) JDS items
Watson (1981) Production 116 JDI-Sup
employees .05
Terborg er al. Retail-sales 23 JDI-Sup .06
(1982) employees 41 - .08
(33% males) 46 - .26

(1Job Diagnostic Survey (Hackman & Lawler, 1971; Hackman & Oldham, 1975).
b Job Descriptive Index (Smith et al., 1969). Sup = supervision.
c Self-developed Likert scales.
* p < .05.
** p < .Ol.
ABSENTEEISM AND JOB SATISFACTION 363

TABLE IO-Continued
Satisfaction
Study Subjects N measure r
42 .Ol
38 .I1
52 .I2
Nicholson (1975) Female sewing 146 JDI-Sup; - .03 - .03
machine operators 66 Hoppock .I1 .I3
96 - .09 - .07
103 ~ .04 - .04
Male production 102 ~ .17 - .I4
workers 58 ~ .23 - .I3
62 -.18 -- .26*
20 .32 .I6
Male chemical 73 .08 .07
process operators 61 .I? .I5
85 -.Oi -- .o.s
80 -.I4 .02
Male bus drivers 76 .I2 -- .Ol
and conductors 63 -.I6 ~-.1x
S8 .08 .l.s
73 .02 -.lO
Attitudinal
Nicholson (1975) Female sewing 146 JDI-Sup: - .07 -.I1
machine operators 66 Hoppock .Ol .I4
96 - .03 -.15
103 -.05 - .I4
Male production 102 - .I0 - .02
workers 58 .08 - .04
62 -.I6 -.I2
20 -.31 -.31
Male chemical 73 .03 .02
process operators 61 .07 .oi
85 - .08 .03
80 - .08 ~ .09
Male bus drivers 76 -.IS -. .29*
and conductors 63 - .09 .I5
58 -.21 .?I!
73 - .27* -. .24”

body of literature questioning the posited link between absence and at-
titudes. Locke (1976), for example, said that the magnitude of the cor-
relation between job satisfaction and absenteeism seldom surpasses .40,
and is typically much lower. Vroom (1964) reported similarly weak find-
ings. In a more recent and comprehensive review of the literature,
Nicholson et al. (1976) concluded that “at best is seems that job satis-
faction and absence from work are tenuously related” (p. 734). Moreover
the relationship of attitudes to other indices of employee performance
364 HACKETT AND GUION

TABLE 11
CORRELATIONS BETWEEN SATISFACTION WITH WORK AND FREQUENCY, TIME LOST, AND
ATTITUDINAL INDICES

Satisfaction
Study Subjects N measure r

Frequency
Waters and Female clerks of 131 JDI-W” - .28**
Roach (1971) insurance agency

over time x settings


Waters and Female clerks of 62 Single -.32* -.40*
Roach (1973) insurance agency 117 12-pt item
JDI-W -.34* -.34'
Nicholson (1975) Female sewing 146 JDI-W; -.08 -.I1
machine operators 66 Hoppock* .07 -.03
96 -.25* -.I5
103 -.05 -.I2
Male production 102 -.26** -.27**
workers 58 .lI .08
62 .22 .16
20 .22 -.19
Male chemical 73 .I6 .13
process operators 61 -.I4 .OI
85 - .OI -.20
80 -.28* -.19
Male bus drivers 76 -.02 -.ll
and conductors 63 .08 .16
58 -.03 .17
73 -.17 -.09
Bemardin Police offtcers 216 JDI-W -.lO
(1976)
Johns (1978) Paper-products 208 JDI-W -.20*
plant employees
Breaugh (1981) Research Scientists 112 Select -.16
(42 females) JDS’
items
Time Lost
Nicholson (1975) Female sewing 146 JDI-W; -.05 -.I6
machine operators 66 Hoppock .08 .03
96 -.lO -.16
103 -.03 .03
Male production 102 -.04 -.I0
workers 58 -.09 .02
62 .08 .05
20 .47* -.20
Male chemical 73 .24* .I5
process workers 61 .Ol .Ol
85 .04 -.07
80 -.23* -.14
Male bus drivers 76 .oo -.35**
and conductors 63 -.09 .08
58 -.02 -.lO
73 -.07 -.02
Smith et al. Male technicians 98 JDI-W -.20
(1969)

’ Job Descriptive Index (Smith ef al., 1969). W = work.


’ Self-developed Likert scales.
’ Job Diagnostic Survey (Hackman & Lawler, 1971; Hackman & Oldham, 1975).
* p < .05.
**p < .Ol.
ABSENTEEISM AND JOB SATISFACTION 365

TABLE 1 I-Continued

Satisfaction
Study Subjects N measure r
Adler and Female telephone 106 JDI-W
Golan (1981) operators
Breaugh (1981) Research Scientists 112 Select
(42 females) JDS items
Watson (1981) 116 JDI-W -.08
Terborg er al. Retail-sales 23 JDI-W -.04
(1982) employees 41 -.43*
(33% males) 46 -.20
42 .23
38 - .Ol
52 -.22
Attitudinal
Nicholson (1975) Female sewing 146 JDI-W; -.04 -.07
machine operators 66 Hoppock .06 -.05
96 -.24* -.I2
103 -.02 -.I0
Male production 102 -.21* -.23*
workers 58 .I2 .07
62 .21 .17
20 .04 -.19
Male chemical 13 .03 .06
process operators 61 -.12 .02
85 .02 -.I5
80 p.15 -- .Ol
Male bus drivers 76 -.05 -.I1
and conductors 63 .09 .I4
58 -.03 .22
73 --.18 - .0x

appear to be just as tenuous (Brayfield & Crockett, 1955; Fisher, 1980;


Wicker, 1969). Three decades ago Brayfield and Crockett (1955) remarked
that “it is time to question the strategic and ethical merits of selling to
industrial concerns an assumed relationship between employee attitudes
and employee performance” (p. 421). This advice seems to be as appli-
cable to attendance as to performance.
Is there still room to argue reasonably in support of a significant link
between these two variables? Steers and Rhodes (1978), in proposing
their process model of attendance, interpreted the modest correlations
reported between absence and attitudes as suggesting that such a rela-
tionship may be moderated by other “third factor” variables. However,
the results of the meta-analyses argue against such an interpretation, and
accordingly undermine their attendance model, in which job satisfaction
is depicted as playing a major explanatory role. Too much of the variance
in the correlations reported across studies can be accounted for by sta-
tistical artifacts alone.
We recognize a general reluctance to abandon the notion that employee
attitudes play a primary role in explaining such “withdrawal” behaviors
as absenteeism and turnover. As Nicholson et al. (1976) noted, it is not
366 HACKETT AND GUION

TABLE 12
CORRELATIONS BETWEEN OVERALL-JOB SATISFACTION AND THE FREQUENCY, TIME LOST,
AND ATTITUDINAL INDICES

Satisfaction
Study Subjects N measure r

Frequency
Vroom (1962) 489 blue-collar 489 Sum three 5-pt items -.07
workers
Hackman and Telephone repairmen 208 Select JDS items” -.lO
Lawler (1971)
Waters and Female clerks of 131 Satisf/dissatisf - .23**
Roach (1971) insurance agency (12-pt item)
Satisf (12-pt item) - .28**
Dissatisf (12-pt item) .16
Stecker (1973) Govt. munitions 459 Factor score .07
mnfg. plant
Waters and Female clerks of 80 Satisf (12-pt item) - .34*
Roach (1973) insurance agency 117 - .38*
Johns (1978) Paper-plant 208 JDI-total -.14*
production workers
Rousseau (1978) Radio station 139 GM faces (Job)b - .24*
employees
(70% females)
Waters and Female clerks of 82 Single 12-pt item -.05 (Year 1)
Roach (1979) insurance agency 83 - .28* (Year 2)
82 -.26
Jamal (1981) Nurses (97% females) 431 Sum 18 items - .12**
and blue-collar
(70% male)
Clegg (1983) Blue-collar 325 Intrinsic & extrinsic -.14
workers of satisf items
engineering plant
F’opp and Male waste 206 Single 5-pt item .23*
Belohlav (1982) collection drivers
Nicholson (1975) Female sewing 146 JDI-Total’ - .08
machine operators 66 .08
% - .24*
103 -.09
Male production 102 - .25*
workers 58 .lO
62 -.06
20 .34
Male chemical 73 .lO
process workers 61 .10
85 -.06
80 -.17
Male bus drivers 76 -.04
and conductors 63 -.07
58 -.10
73 - .27
Hammer et al. Nonsupervisory 112 GM faces (overall) .05
(1981) furniture mnfg.
(15 females)
’ Job Diagnostic Survey (Hackman & Lawler, 1971; Hackman t Oldham, 1975)
b GM faces (Kunin, 1955).
’ Job Descriptive Index (Smith er al., 1969).
d Minnesotta Satisfaction Questionnaire (Weiss et al., 1%7).
* p < .05.
** p < .Ol.
ABSENTEEISM AND JOB SATISFACTION 367

TABLE 12-Continued

Satisfaction
Study Subjects N measure I
Time Lost
Smith et al. Male technicians 98 JDI-Total - .36**
(1969) GM faces (Job) - .22**
Ilgen and University clerical 165 MSQ-Total” - .09
Hollenback (1977) workers
Johns (1978) Paper-plant 208 JDI-Total - .03
production workers
Rousseau (1978) Radio station 139 GM faces (Job) -- .14*
employees
(70% females)
Adler and Female telephone 82 JDI-Total
Golan (1981) operators
Nicholson (1975) Female sewing 146 JDI-Total - .02
machine operators 66 .12
96 -.ll
103 p.10
Male production 102 -.21*
workers 58 -.12
62 -- .05
20 .62**
Male chemical 73 .16
process operators 61 .09
85 .OS
80 - .21
Male bus drivers 76 .03
and conductors 63 - .21*
58 .08
73 .Ol
Attitudinal
Nicholson (1975) Female sewing 146 JDI-Total - .05
machine operators 66 .05
% - .22f
103 -.04
Male production 102 -.16
workers 58 .I0
62 -.08
20 - .03
Male chemical 73 .02
process operators 61 .09
85 .02
80 - .04
Male bus drivers 76 - .06
and conductors 63 .oo
58 - .OR
73 - .27*

unusual to find review articles, or basic texts of industrial psychology,


citing an unrepresentative selection of studies in support of a primary
link between job satisfaction and absenteeism (e.g., Argyle, 1972; Beh-
rend, 1959; Katz & Kahn, 1980; Lawler, 1973; Maier, 1955; Nadler,
Hackman, & Lawler, 1979; Porter & Steers, 1973). Nicholson et al. (1976)
partly attributed such resistance to “an uncritical acceptance of the
TABLE 13
OTHERCORRELATIONS
BETWEENABSENCEAND JOB SATISFACTION 5

Study Subjects N Absence Satisfaction r


Noland (1945) Industrial workers Unspecified No. days/month Unclear -.67
Van Zelst and Blue collar workers -131 Unspefified Rvo items
Kerr (1953) - .31**
Newman (1974) Nursing home employees 108 Absent/not absent GM faces0 - .31**
(88% female) JDIb-Co -.ll
-sup .13
-Prom -.lO
-hY -.04
-Work - .19* F
Paid Unpaid a
Garrison and Mostly clerical 174 Paid/unpaid JDI-Co .02 -.12 2
Muchinsky (1977) (16 males) -sup -.06 -.12
-Prom - .08 -.04 %
u
-fiY .03 - .06
-Work -.12 - .26** 2
-Total .05 - .20**
Nicholson et a/. Male blue-collar 95 Self-report JDI-Co - .23** H
(1977) steelworkers -sup -.13
-Prom -.14
-fiY - .21
-Work - .37**
-Total - .30**
Firm - .24*
Cheloha and Clerical and supervisory 115 (1) Sick total MSQC (short) -- 1 2 -- 3 4
Farr (1980) personnel of govt. agency (2) Sick freq Intrinsic - .05 -.17 - .19* - .26
(54% females) (3) Vacation Total Extrinsic - .21* -.28** -.14 -.02
(4) Vacation Freq JDI-Co - .07 - .23** .03 -.13
-sup - .05 -.I0 -.03 .oo
-Prom .Ol -.I7 - .05 - .08
-fiY - .Ol - .20* .04 -.06
-Work -.18 - .38** - .25** - .30**
Time 1 Time 2
Fitzgibbons and Assembly plant workers 269 Excused Intrinsic - .23** -.12
Moth (1980) (78% nonsupervisory) Sickness Satisfaction - .22** -.I1
Unexcused -.ll - .21** b
in
(1) (11) 6
Adler and Female telephone operators 131 (1) Unexcused GM faces (Work) -.I8 * - .13
J
Golan (1981) 114 (2) Medical JDI-Co .05 .04
115 -sup - .Ol - .06 !!
117 -Prom -.14 .ll E
124 -hY -.12 - .06 k.
106 -Work - .20* -.04 i
u
82 -Total - .06 - .08
Breaugh (1981) Research scientists 112 Supervisory Select JDSd 2
-.lO
(42 females) rating items
Hammer et al. Nonsupervisory 112 Involuntary GM faces (Overall)
(1981) furniture mnfg. Hours - .08
(IS females) Days -.13
Freq - .20’
” GM faces (Kunin, 1955).
’ Job Descriptive Index (Smith et al., 1969).
c Minnesotta Satisfaction Questionnaire (Weiss, 1967).
d Job Diagnostic Survey (Hackman & Lawler, 1971; Hackman & Oldham, 1975).
e Scored such that higher scores indicate negative attitudes.
* p < .05.
** p < .Ol.
370 HACKETT AND GUION
ABSENTEEISM AND JOB SATISFACTION 371

Human Relations school’s oversimplistic prescriptions” (p. 728). That is,


it makes intuitive sense that job satisfaction would predict absence from
work. As expressed by Brayfield and Crockett (1955), “To the extent that
worker dissatisfaction indicates that the individual is in a punishing sit-
uation, we should expect dissatisfied workers to be absent more often”
(p. 415). The results of the current investigation, along with the findings
of Nicholson et al. (1976) and Terborg et al. (1982), certainly challenge
this theoretical orientation.
There are, however, alternative explanations that could be offered for
the tenuous correlations between absence and attitudes. Perhaps the re-
lationship is nonlinear and these low coefficients have resulted from
trying to fit a straight line to curved data (Hammer & Landau, 1981).
Hammer and Landau (1981) have also shown that, because of a large
number of zero values, the sample distribution of absenteeism tends to
be positively skewed and truncated-that in a given sample most people
are likely to be absent infrequently, and only a few are likely to be absent
a lot (cf. Walker, 1947). A skewed truncated distribution constricts the
value of the correlation coefficient; the more disparate the marginal dis-
tributions the lower will be the value of “r” (Carroll, 1961). Accordingly,
the typically low correlations found between employee absenteeism and
job satisfaction may in part be attributable to the skewed, truncated na-
ture of distributions of absences. Without the raw data from each study
it is not possible to test these alternative explanations. Future research
ought to report the characteristics of the distribution of absences used in
any correlational analysis.
SUGGESTIONS FOR FUTURE RESEARCH
Theory-Guided Multivariate Research
Most of the research on employee absenteeism has involved simple
bivariate correlational studies in which absences are correlated with some
other variable to which they are intuitively thought to be related (i.e.,
attitudes). Only recently have multivariate studies which attempt to take
into consideration the possible simultaneous influences of several vari-
ables on absence appeared in the literature (cf. Adler & Golan, 1981;
Breaugh, 1981; Cheloha & Fart-, 1980; Clegg, 1983, Fitzgibbons, & Moth,
1980; Frechette, 1981; Hammer et al., 1981; Ilgen & Hollenback, 1977;
Johns, 1978; Watson, 1981). Yet, such multivariate research needs direc-
tion. While some comprehensive models of absence behavior have been
developed (cf. Gibson, 1966; Nicholson, 1977; Steers & Rhodes, 1978),
research aimed at assessing these models has been scant. For example,
although the models have clear temporal implications, calling for a path-
analytic evaluation of their postulates, no such research has yet been
reported. The Steers and Rhodes (1978) model has received the most
TABLE 14
RESULTSFORVALIDITY GENERALIZATION
No. Observed Predicted” % var. Residual
N SD acct. for SD -b SDP 95% CI’
r’s cy SD P
i?
Promotion
Frequency 2783 34 - .046 .141 ,111 62 .086 -.066 .125 -.31 s p s .18
Attitudinal 2444 32 - .037 .120 .115 92* .034 - .060 .056 -.17sp=s.o5 5
Time Lost 2918 40 - ,057 .167 .117 50 .I19 - .071 .148 - .36 c p c .22
%
All 9390 116 - ,052 .170 .lll 43 .128 - .072 .150 - .37 s p s .22 u
Co-workers co
Frequency 2783 34 - .024 .134 .lll 69 .075 - .035 .109 - .25 s p c .I8
Attitudinal 2444 32 -.044 .116 .115 98* .015 - .072 .025 -.12 cp =s -.02
Time Lost 3014 41 - .039 ,162 .117 52 .112 - .049 .141 - .33 c p =s .23
All 9480 117 - .041 .136 .lll 67 .078 - .057 .108 -.27sps.15

hY
Frequency 2783 34 .ooo .141 .lll 62 .087 .ooo .I26 - .24 s p s .24
Attitudinal 2444 32 .030 .153 .llS 56 .lOl .049 ,165 - .27 s p s .37
Time Lost 3024 41 - .055 ,136 ,117 74* .069 - .069 .086 -.24 s p s .lO
All 9510 117 - .018 .141 .106 57 .093 - .025 .129 - .28 s p s .23
Work
Frequency 3469 40 -.112 .154 .llO 51 .108 -.162 .155 -.47 sp c .14
Attitudinal 2444 32 - $43 .119 .115 93* .031 - .069 .OSl -.17sps .03
Time Lost 3118 42 - ,054 .126 .116 m* ,049 - ,068 .062 -.19cp s .05
All 10254 124 - ,091 ,144 ,112 61 ,090 -.127 .053 - .23 s p =G- .02
Supervision
Frequency 3319 37 - ,074 ,106 ,106 100* .ooo -.107
Attitudinal 2444 32 - .062 ,113 ,113 loo* .ooo - .I01
Time Lost 3127 42 -.044 ,128 ,116 82* ,054 - .055 .067 -.19~~~.08
All 10131 121 - ,059 ,112 ,105 88% .038 - .082 .053 -.18 s p s .02
OveralP is
Frequency 4636 33 - .093 ,138 .087 40 .108 -.134 ,156 -.44sp=s .17
i4
Attitudinal 1382 17 - .143 ,258 ,117 21 ,230 - ,233 ,374 - .96 =Sp c .50
Time Lost 2012 22 - .076 .138 .106 59 *OS8 - ,096 ,111 -.31 <p =z .12 3
A11 9440 84 -.I02 .I59 ,098 37 .I26 -.141 ,175 - .48 s p c .20 i-4
G
All x All $
62308 707 - .067 .140 ,107 57 ,093 - ,093 .I26 -.33 s p =z .15
5
n When squared, this represents the estimated variance attributable to the two artifacts for which corrections were made (i.e., sampling error
and unreliability of absence measure). $
b Sample-size weighted mean corrected for unreliability of absence measure.
c 95% confidence interval. 5
d Includes JDI-Total, GM faces (Job), Sum of Intrinsic and Extrinsic Satisfaction scores, and Overall Likert items. t;
* Situational-specificity hypothesis rejected using 70% criterion. 2
2
374 HACKETT AND GUION

attention, yet only four studies providing partial tests of it have appeared
in the literature (cf. Frechette, 1981; Hammer et al., 1981; Terborg et al.,
1982; Watson, 1981); the implications that the results of these studies had
for various aspects of the model are discussed by Mowday et al. (1982,
pp. 99-103). Miller (1981) compared and assessed four models of with-
drawal behavior: (a) that absence and turnover ought to be modeled sep-
arately (i.e., Steers & Rhodes, 1978); (b) that absence represents an early
form of withdrawal behavior leading to final separation from the work
role via turnover (i.e., Herzberg et al., 1957; Hill & Trist, 1955); (c) that
absence and/or turnover behavior may result from negative job attitudes,
depending on the relative ease of engaging in these behaviors (i.e., March
& Simon, 1958); and (d) that absence behavior serves primarily to help
market-testing for alternative employment prospects (i.e., Mobley, Grif-
feth, Hand, & Meglino, 1979). According to Miller (1981) the first two of
these were not supported by the data, the third was “marginally sup-
ported,” and the fourth was found to be the most consistent with the
data. Further comprehensive model testing of this sort is badly needed.
However, models based on the assumption that dissatisfaction is a pri-
mary cause of absence do not seem the appropriate ones to test further.
Attitudes versus Values
There is a body of theoretical literature that suggests attitudes may be
less potent predictors of absence than the personal value systems of
workers (cf. Rokeach, 1973). In contrast to attitudes, which are consid-
ered to reflect a level of affect toward a specific object or situation, values
are thought to transcend objects and situations and be connected with
the satisfaction of higher order personal needs (Rokeach, 1973). Ac-
cording to Rokeach (1973), then, values are fewer in number and occupy
a more central position in an individual’s personality makeup and cog-
nitive system.
The two most consistent findings from absenteeism research show that
absences tend to be greater among females and negatively related to age
(cf. Porwoll, 1980). Both findings are consistent with the notions that (a)
a young mother’s primary values lie in the home and family and (b) for
females more than for males, work outside the home assumes a secondary
role. Moreover, higher absences among younger married men might be
attributable to a desire of young fathers to spend more time with their
families (Dekar, 1969). Morgan and Herman (1976) have shown that work
and nonwork values often conflict; that many employees can be drawn
away from their jobs on any given day by the anticipation of satisfaction
with off the job alternatives. Thus, any two persons reporting an equal
level of job satisfaction may experience different opportunities for ex-
ternal satisfactions on a day off from work. Too little research has sought
ABSENTEEISM AND JOB SATISFACTION 375

to investigate the role that personal values and non-work-related features


of an individual’s life play in absence. Rather than continuing to concep-
tualize absence primarily in terms of a process of withdrawal from a
negative work environment, which from a managerial perspective con-
veniently places such behavior under the direct control of organizations,
it is likely to prove more fruitful to view absence more in terms of a
process in which workers are drawn CX.Uof the workplace by valued
features of their nonwork environments.

A Cognitive Approach to Studying Absenteeism


The concept of “voluntary absenteeism” implies volition; that the in-
dividual worker makes a conscious decision on whether to attend work
on any given day. This is not to suggest that before starting off to work
each day the individual goes through some complex, conscious process
of evaluating the pros and cons of attending; attendance is likely to be
an established habit. Rather, when a potentially absence-inducing event
does arise, then this is likely to “trigger” a consideration of the costs
and benefits of absenting oneself for the day. Accordingly, efforts to un-
derstand voluntary absenteeism should be aimed at identifying those fac-
tors that are liklely to enter into a worker’s decision whether to be absent.
However, surprisingly few studies have attempted an investigation of the
“weightings” workers give to various factors when making these deci-
sions. Stagner and Rosen (1965, p. 30) illustrated how a decision analysis
involving the calculation of “subjective expected utilities” might be ap-
plied in efforts to understand and predict an employee’s absences.
Morgan and Herman (1976) used an expectancy-theory paradigm in hy-
pothesizing that “employees will decide whether or not to attend work
by weighing the deterrent and motivating consequences of being absent”
(p. 738). They found that absenteeism provides employees with the op-
portunity to experience desired outcomes (i.e., break from routine, per-
sonal business, family leisure) that are not offset by organizationally con-
trolled deterrent outcomes (i.e., loss in pay, supervisory reprimand, co-
workers work harder). One of the merits of their study is that a total of
22 possible outcomes (positive and negative, work and nonwork) of being
absent, all thought to influence an individual’s decision on whether to
attend work, were considered. In contrast to studies simply correlating
measures of absence with other demographic, social, organizational, and
attitudinal variables would be studies seeking an understanding of the
work and nonwork factors weighted most heavily in attendance or ab-
sence decisions. While the Morgan and Herman (1976) study clearly
moves in this direction, it poses a methodological problem. “importance
values” and “instrumentalities” were determined by asking the workers
themselves to rate each of the outcomes on these two dimensions (i.e.,
376 HACKETT AND GUION

“how important is it to you to be home working on the house?“; “how


instrumental would being absent from work be in attaining this out-
come?“). The problem with this approach is that workers may not be
very good at subjectively estimating the impact of these factors on their
decisions (cf. Slavic & Lichtenstein, 1971). An alternative approach
would be to do a policy-capturing study (cf. Brunswik, 1952, 1956; Ham-
mond, 1966, 1978) where a representative sample of the factors that enter
into decisions of absence would serve as the “cues” in the model.
Perhaps the voluntary-involuntary absence dichotomy is artilical and
misleading (cf. Nicholson, 1977); absenteeism might be better concep-
tualized on a continuum of justifiability. For example, is an absence as-
sociated with a 99” temperature on a particular workday to be classified
as voluntary or involuntary? How about an absence associated with an
unexpected visit by out-of-town relatives or friends? Such questions sug-
gest that work is needed to guide the development of new models of
absenteeism research. New models would surely be multivariate, since
there may be more than one dimension along which a worker could justify
an absence.
In view of the results reported here, it is clear that new research di-
rections are needed to determine the psychological processes, anteced-
ents, and consequences of staying away from work.
APPENDIX: STUDIES INCLUDED IN THE META-ANALYSES
Adler, S., & Golan, J. (1981). Lateness as a withdrawal behavior. Journnl of Applied Psy-
chology, 66, 544-554.
Bemardin, H. J. (1976). The influence of reinforcement orientation on the relationship be-
tween supervisory style and effectiveness criteria (Doctoral dissertation, Bowling
Green State University, 1976). Dissertation Abstracts International, 37 (2-B) 1018.
Bemberg, R. E. (1952). Socio-psychological factors in industrial morale: I. The prediction
of specific indicators. The Journal of Social Psychology, 36, 73-82.
Breaugh, J. A. (1981). Predicting absenteeism from prior absenteeism and work attitudes.
Journal of Applied Psychology, 66, 555-W.
Chadwick-Jones, J. K., Brown, C. A., Nicholson, N., & Sheppard, C. (1971). Absence
measures: Their reliability and stability in an industrial setting. Personnel Psychology,
24, 463-470.
Chadwick-Jones, J. K., Nicholson, N., & Brown, C. (1982). Social psychology of absen-
teeism. New York: Praeger.
Cheloha, R. S., & Farr, L. J. (1980). Absenteeism, job involvement, and job satisfaction in
an organizational setting. Journal of Applied Psychology, 65, 467-473.
Clegg, C. W. (1983). Psychology of employee lateness, absence, and turnover: A method-
ological critique and an empirical investigation. Journal of Applied Psychology, 68 (1)
88-101.
Farr, J. L., O’Leary, B. S., & Bartlett, C. J. (1971). Ethnic group membership as a mod-
erator of the prediction of job performance. Personnel Psychology, 24, 609-636.
Fitzgibbons, D., & Moth, M. (1980). Employee absenteeism: A multivariate analysis with
replication. Organizational Behavior and Human Performance. 26, 349-372.
Garrison, K. R., & Muchinsky, P M. (1977). Attitudinal and biographical predictors of
incidental absenteeism. Journal of Vocational Behavior, 10, 221-230.
ABSENTEEISM AND JOB SATISFACTION 377

Hackman, J. R., & Lawler, E. E., III (1971). Employee reactions to job characteristics.
Journal of Applied Psychology Monograph, 55, 2.59-286.
Hammer, T. H., & Landau, J. (1981). Methodological issues in the use of absence data.
Journal of Applied Psychology, 66, 574-581.
Hammer, T. H., Landau, J. C., & Stem, R. N. (1981). Absenteeism when workers have a
voice: The case of employee ownership. Journal of Applied Psychology, 66, 561-573.
Huse, E. F., & Taylor, D. K. (1%2). Reliability of absence measures. Journal of Applied
Psychology, 46, 159-160.
Ilgen, D. R., & Hollenback, J. H. (1977). The role of satisfaction in absence behavior.
Organizational Behavior and Human Performance, 19, 148-161.
Jamal, M. (1981). Shift work related to job attitudes, social participation and withdrawal
behavior: A study of nurses and industrial workers. Personnel Psychology, 34, 535-
547.
Johns, G. (1978). Attitudinal and nonattitudinal predictors of two forms of absence from
work. Organizational Behavior & Human Performance, 22, 431-444.
Latham, G. P., & Purse& E. D. (1975). Measuring absenteeism from the opposite side of
the coin. Journal of Applied Psychology, 60, 369-371.
Miller, H. E. (1981). Withdrawal behaviors among hospital employees (Doctoral Disserta-
tion, University of Illinois, 1981). Dissertation Abstracts International, 42 (6-B), 2586-
2587.
Mirvis, P. H., & Lawler, E. E. (1977). Measuring financial impact of employee attitudes.
Journal of Applied Psychology, 62, l-8.
Newman, J. D. (1974). Predicting absenteeism and turnover: A field comparison of Fish-
bein’s model and traditional job attitude measures. Journal of Applied Psychology. 59,
610-615.
Nicholson, N. (1975). Industrial absence as an indicant of employee motivation and job
satisfaction: A study of contrasting technologies in the North of England and South
Wales. Unpublished doctoral dissertation, University of Wales.
Nicholson, N., Brown, C. A., & Chadwick-Jones, J. K. (1976). Absence from work and
job satisfaction. Journal of Applied Psychology, 61 (6) 728-737.
Nicholson, N., & Goodge, P. M. (1976). The influence of social, organizational and bio-
graphical factors on female absence. The Journal of Management Studies, 13, 234-
254.
Nicholson, N., Wall, T., & Lischeron, J. (1977). The predictability of absence and pro-
pensity to leave from employees’ job satisfaction and attitudes toward influence in
decision-making. Human Relations, 30, 499-514.
Noland, E. W. (1945). Foreman and absenteeism. Personnel Journal, 45, 73-77.
Popp, P. O., & Belohlav, J. A. (1982). Absenteeism in a low status work environment.
Academy of Management Journal, 25 (3), 677-683.
Rosensteel, R. K. (1953). A validation of a test-battery and biographical data for machine
operators. Unpublished master’s thesis, Bowling Green State University, Ohio.
Rousseau, E. M. (1978). Relationship of work to nonwork. Journal of Applied Psychology.
63, 513-517.
Smith, P. C., Kendall, C., & Hulin, C. (1969). The measurement of satisfaction in work
and retirement. Chicago: Rand McNally.
Stecker, A. L. (1973). A two-factor behavioral theory of industrial absenteeism (Doctoral
dissertation, Indiana University School of Business, 1973). Dissertation Abstracts Zn-
ternational, 34 (3-A), 95 1.
Terborg, .I. R., Lee, T. W., Smith, F. J., Davis, G. A., & Ilnbin, M. S. (1982). Extension
of the Schmidt and Hunter validity generalization procedure to the prediction of ab-
senteeism behavior from knowledge ofjob satisfaction and organizational commitment.
Journal of Applied Psychology, 67 (4). 440-449.
378 HACKETT AND GUION

‘Buner, W. W. (1960). Dimensions of foremen performance: A factor analysis of criterion


measures. Journal of Applied Psychology, 44, 216-223.
Van Zelst, R. H., & Kerr, W. A. (19.53).Workers attitudes toward merit rating. Personnel
Psychology, 6, 159-172.
Vroom, V. H. (1962). Ego involvement, job satisfaction, and job performance. Personnel
Psychology, 15, 159- 177.
Waters, L. K., & Roach, D. (1971). Relationship between job attitudes and two forms of
withdrawal from the work situation. Journal of Applied Psychology, 55, 92-94.
Waters, L. K., & Roach, D. (1973). Job attitudes as predictors of termination and absen-
teeism: Consistency over time and across organizational units. Journal of Applied
Psychology, 57, 341-342.
Waters, L. K., & Roach, D. (1979). Job satisfaction, behavioral intension, and absenteeism
as predictors of turnover. Personnel Psychology, 32, 393-397.
Watson, C. J. (1981). An evaluation of some aspects of the Steers and Rhodes model of
employee attendance. Journal of Applied Psychology, 66, 385-389.

REFERENCES
Adler, S., & Golan, J. (1981). Lateness as a withdrawal behavior. Journal of Applied Psy-
chology,
66, 544-554.
Argyle, M. (1972). The social psychology of work. Harmondsworth, Middlesex: Penguin.
Arglye, M., Gardner, G., & Ciofft, E (1958). Supervisory methods related to productivity,
absenteeism, and labour turnover. Human Relations, 11, 23-40.
Bakan, D. (1967). On method. San Francisco: Jossey-Bass.
Behrend, H. (1959). Voluntary absence from work. International Labour Review. 79, 109-
140.
Bemardin, H. J. (1976). The influence of reinforcement orientation on the relationship be-
tween supervisory style and effectiveness criteria (Doctoral dissertation, Bowling
Green State University, 1976). Dissertation Abstracts International, 37 (2-B), 1018.
Bernberg, R. E. (1952). Socio-psychological factors in industrial morale: I. The prediction
of specific indicators. The Journal of Social Psychology, 36, 73-82.
Brayfield, A. H., & Crockett, W. H. (1955). Employee attitudes and employee performance.
Psychology Bulletin, 52, 396-424.
Breaugh, J. A. (1981). Predicting absenteeism from prior absenteeism and work attitudes.
Journal of Applied Psychology, 66, 555-560.
Bnmswik, E. (1952). The conceptual framework of psychology. Chicago: Univ. of Chicago
Press.
Brunswik, E. (1955). Representative design and probabilistic theory in a functional psy-
chology. Psychological Review, 62, 193-217.
Brunswik, E. (1956). Perception and the representative design of experiments. Berkeley:
Univ. of California Press.
Carroll, J. B. (1961). The nature of the data, or how to choose a correlation coefficient.
Psychometrica, 26, 347-372.
Cattell, R. B. (1966). The Scree test for the number of factors. Multivariate Behavioral
Research, 1, 245-276.
Chadwick-Jones, J. K., Brown, C. A., Nicholson, N., & Sheppard, C. (1971). Absene mea-
sures: Their reliability and stability in an industrial setting. Personnel Psychology, 24,
463-470.
Chadwick-Jones, J. K., Nicholson, N., & Brown, C. (1982). Social psychology of absen-
teeism. New York: Praeger.
Cheloha, R. S., & Farr, L. J. (1980). Absenteeism, job involvement, and job satisfaction in
an organizational setting. Journal of Applied Psychology, 65, 467-473.
Clegg, C. W. (1983). Psychology of employee lateness, absence, and turnover: A method-
ABSENTEEISM AND JOB SATISFACTION 379

ological critique and an empirical investigation. Journnl of Applied Psychology. 68 (1).


88-101.
Dekar, A. D. (1%9, November) Absenteeism: A fact of life. Personnel Journal, 881-888.
Fair, R. C. (1977). A note on the computation of the Tobit estimator. Econometrics, 45,
1723-1127.
Farr, J. L., O’Leary, B. S., & Bartlett, C. J. (1971). Ethnic group membership as a mod-
erator of the prediction of job performance. Personnel Psychology, 24, 609-636.
Fisher, C. D. (1980). On the dubious wisdom of expecting job satisfaction to correlate with
performance. Academy of Management Review, 5, 607-612.
Fitzgibbons, D., & Moth, M. (1980). Employee absenteeism: A multivariate analysis with
replication. Organizational Behavior and Human Performance, 26, 349-372.
Fox, J. B., & Scott, J. R. (1943). Absenteeism: Managements problem. Boston: Graduate
School of Business Administration, Harvard University.
Frechette, H. (1981, August). An investigation ofthe utility ofSteers and Rhodes’ process
model of attendance behavior. Paper presented at the 41st Annual Meeting of the
Academy of Management, San Diego.
Gaudet, F. J. (1%3). Solving the problems of employee absence. New York: American
Management Association.
Gibson, J. (1966). Toward a conceptualization of absence behaviour of personnel in orga-
nizations. Administrative Science Quarterly, 11, 107- 133.
Glass, G. V., McGaw, B., & Smith, M. L. (1981). Meta-analysis in social research. Beverly
Hills. CA Sage.
Hackman, J. R., & Lawler, E. E., III (1971). Employee reactions to job characteristics.
Journal of Applied Psychology Monograph, 55, 259-286.
Hackman, J. R., & Oldham, G. R. (1975). Development of the job diagnostic survey.
Journal of Applied Psychology, 60, 159- 170.
Hammer, T. H., & Landau, J. (1981). Methodological issues in the use of absence data.
Journal of Applied Psychology, 66, 574-581.
Hammer, T. H., Landau, J. C., & Stem, R. N. (1981). Absenteeism when workers have a
voice: The case of employee ownership. Journal of Applied Psychology, 66, 561-573.
Hammond, K. R. (1966). Probabilistic functionalism: Egon Brunswik’s integration of the
history, theory, and method of psychology. In K. R. Hammond (Ed.), The psychology
of Egon Brunswik. New York: Holt, Rinehart, & Winston.
Hammond, K. R. (1978). Toward increasing competence of thought in public policy for-
mation. In K. R. Hammond (Ed.), Judgment and decision in public policy formation.
Boulder. CO Westview Press.
Harman, H. H. (1967). Modern factor analysis. Chicago: Univ. of Chicago Press.
Helwig, J. T., & Council, K. A. (1979). SAS user’s guide. Raleigh, NC: SAS Institute.
Hendrickson, A. E., & White, P. 0. (1964). Promax: A quick method for rotation to oblique
simple structure. British Journal of Statistical Psychology, 17, 65-70.
Herzberg, F., Mausner, B., Peterson, R., & Capwell, D. (1957). Job attitudes: Review of
research and opinion. Pittsburgh: Psychology Service of Pittsburgh.
Hill, J., & Trist, E. (1955). Changes in accidents and other absences with length of service.
Human Relations, 8, 121-152.
Hunter, J. E., Schmidt, F. L., & Jackson, G. (1982). Advanced meta-analysis: Quantitative
methods for cumulating research findings across studies. Beverly Hills. Sage.
Huse, E. F., & Taylor, D. K. (1962). Reliability of absence measures. Journal of Applied
Psychology, 46, 159-160.
Ilgen, D. R. (1977). Attendance vs absenteeism as performance criteria, criticism of G. P.
Latham & E. D. Pursell’s conclusions. Journal of Applied Psychology, 62, 230-233.
Ilgen, D. R., & Hollenback, J. H. (1977). The role of satisfaction in absence behavior.
Organizational Behavior and Human Performance, 19, 148- 161.
380 HACKETT AND GUION

Johns, G. (1978). Attitudinal and nonattitudinal predictors of two forms of absence from
work. Organizational Behavior and Human Performance, 22, 431-444.
Johns, G., & Nicholson, N. (1982). The meanings of absence: New strategies for theory
and research. Research in Organizational Behavior, 4, 127-172.
Katz, D., & Kahn, R. (1980). The social psychology of organizations (2nd ed.). New York:
Wiley.
Kunin, T. (1955). The construction of a new type of attitude measure. Personnel Psy-
chology, 8, 65-78.
Latham, G. P, & Pursell, E. D. (1975). Measuring absenteeism from the opposite side of
the coin. Journal of Applied Psychology, 60, 369-371.
Latham, G. P., & Purse& E. D. (1977). Measuring attendance: A reply to Ilgen. Journal of
Applied Psychology, 62 (2), 234-236.
Lawler, E. E. (1973). Motivation in work organizations. Monterey, CA: Brooks/Cole.
Locke, E. A. (1976). The nature and cause of job satisfaction. In M. D. Dunnette (Ed.),
Handbook of industrial and organizational psychology. Chicago: Rand McNally.
Lyons, T. E (1972). Turnover and absenteeism: A review of relationships and shared cor-
relates. Personnel Psychology, 25, 271-281.
Maier, N. R. E (1955). Psychology in industry. Boston: Houghton Mifflin.
March, J., & Simon, H. (1958). Organizations. New York: Wiley.
McNemar, Q. (1960). At random: Sense and nonsense. American Psychologist, 15, 295-
300.
Meehl, P E. (1978). Theoretical risks and tabular asterisks: Sir Karl, Sir Ronald, and the
slow progress of soft psychology. Journal of Consulting and Clinical Psychology, 46
(No. 4), 806-834.
Miller, H. E. (1981). Withdrawal behaviors among hospital employees (Doctoral disserta-
tion, University of Illinois, 1981). Dissertation Abstracts International, 42 (6-B), 2586-
2587.
Mirvis, P. H., & Lawler, E. E. (1977). Measuring financial impact of employee attitudes.
Journal of Applied Psychology, 62, l-8.
Mobley, W., Griffeth, R., Hand, H., & Meglino, B. (1979). Review and conceptual analysis
of the employee turnover process. Psychological Bulletin, 86, 493-522.
Montanelli, R. G., & Humphrey% L. G. (1976). Latent roots of random data correlation
matrices with squared multiple correlations in the diagonal: A Monte Carlo study.
Psychometrika, 41 (3), 341-348.
Morgan, L., & Herman, J. (1976). Perceived consequences of absenteeism. Journal of
Applied Psychology, 61, 738-742.
Mowday, R. T., Porter, L. Y., & Steers, R. M. (1982). Employee-organization linkages:
The psychology of commitment, absenteeism, and turnover. New York: Academic
Press.
Muchinsky, P. (1977). Employee absenteeism: A review of the literature. Journal of Vo-
cational Behavior, 10, 316-334.
Nadler, D. A., Hackman, R. J., & Lawler, E. E. (1979). Managing organizational behavior.
Boston: Little, Brown.
Nicholson, N. (1975). Industrial absence as an indicant of employee motivation and job
satisfaction: A study of contrasting technologies in the North of England and South
Wales. Unpublished doctoral dissertation, University of Wales.
Nicholson, N. (1977). Absence behaviour and attendance motrivation: A conceptual syn-
thesis. Journal of Management Studies, 14, 23l-252.
Nicholson, N., Brown, C. A., & Chadwick-Jones, J. K. (1976). Absence from work and
job satisfaction. Journal of Applied Psychology, 61 (6), 728-737.
Nicholson, N., & Goodge, P. M. (1976). The influence of social, organizational and bio-
graphical factors on female absence. The Journal of Management Studies, 13, 234-
254.
ABSENTEEISM AND JOB SATISFACTION 381

Osbum, H. G., Callendar, J. C., Greener, J. M., & Asbworth (1983). Statistical power of
tests of the situational specificity hypothesis in validity generalization studies: A cau-
tionary note. Journal of Applied Psychology, 68, 115-122.
Pearlman, K., Schmidt, E L., & Hunter, J. E. (1980). Validity generalization results for
tests used to predict job proficiency and training success in clerical occupations.
Journal of Applied Psychology, 65, 373-406.
Porter, L. W., & Steers, R. M. (1973). Organizational, work and personal factors in em-
ployee turnover and absenteeism. Psychology Bulletin, 80, 151- 156.
Porwoll, F! J. (1980). Employee absenteeism: A summary of research. Arlington, VA: Re-
search Brief, Educational Research Service.
Rokeach, M. (1973). The nature of human values. New York: Free Press.
Rosensteel, R. K. (1953). A validation ofa test-battery and biographical data for machine
operators. Unpublished master’s thesis, Bowling Green State University, Ohio.
Rosenthal, R. (1978). Combining results of independent studies. Psychological Bulletin, 85,
185-193.
Rosenthal, R., & Gaito, J. (1963). The interpretation of levels of significance by psycho-
logical researchers. Journal OfPsychology, 55, 33-38.
Rummel, R. J. (1970). Appliedfactor analysis. Evanston: Northwestern Univ. Press.
Schmidt, F. L., & Hunter, J. E. (1977). Development of a general solution to the problem
of validity generalization. Journal of Applied Psychology, 62, 529-540.
Schmidt, E L., Hunter, J. E., Pearlman, K., & Shane, G. S. (1979). Further tests of the
Schmidt-Hunter Bayesian validity generalization procedure. Personnel Psychology,
32, 257-281.
Slavic, P., & Lichtenstein, S. (1971). Comparison of Bayesian and regression approaches
to the study of information processing judgment. Organizational Behavior and Human
Performance, 6, 649-744.
Smith, P. C., Kendall, C., & Hulin, C. (1969). The measurement of satisfaction in Mjork
and retirement. Chicago: Rand McNally.
Stagner, R., & Rosen, H. (1%5). The psychology of union-management relations. Bel-
mont, Calif.: Wadsworth.
Steers, M. R., &Rhodes, S. R. (1978). Major influences on employee attendance: A process
model. Journal of Applied Psychology, 63, 391-407.
Sterling, T. D. (1959). Publication decisions and their possible effects on inferences drawn
from tests of significance-or vice versa. Journal of the American Statistical Associ-
ation, 54, 30-34.
Terborg, J. R., Lee, T. W., Smith, F. J., Davis, G. A., & Turbin, M. S. (1982). Extension
of the Schmidt and Hunter validity generalization procedure to the prediction of ab-
senteeism behavior from knowledge of job satisfaction and organizational commitment.
Journal of Applied Psychology, 67 (4), 440-449.
Turner, W. W. (1960). Dimensions of foremen performance: A factor analysis of criterion
measures. Journal of Applied Psychology, 44, 216-223.
Vroom, V. H. (1964). Work and motivation. New York: Wiley.
Walker, K. (1947). The application of the J-curve hypothesis of conforming behavior to
industrial absenteeism. Journal of Social Psychology, 25, 207-216.
Waters, L. K., & Roach, D. (1979). Job satisfaction, behavioral intension, and absenteeism
as predictors of turnover. Personnel Psychology, 32, 393-397.
Watson, C. J. (1981). An evaluation of some aspects of the Steers and Rhodes model of
employee attendance. Journal of Applied Psychology, 66, 385-389.
Weiss, D., Dawis, R., England, G., & Lofquist, L. (1967). Manual for the Minnesota
Satisfaction Questionnaire. Minneapolis: Univ. of Minnesota Press.
Wicker, A. W. (1969). Attitudes vs actions: The relationship of verbal and overt behavior
responses to attitude objects. Journal of Social Issues, 25, 41-78.
RECEIVED: October 26, 1983

You might also like