ILLINOIS LIBRARY
at URBANA-CHAMPAIGN
w STACKS
Wow, Ss SSw -* ,,>5, f~ —
JUN051938
Digitized by the Internet Archive
in 2011 with funding from
University of Illinois Urbana-Champaign
http://www.archive.org/details/automationemploy592moch
Faculty Working Papers
College of Commerce and Business Administration
University of Illinois at U rba n a - Cha m pa ig n
FACULTY WORKING PAPERS
College of Commerce and Business Administration
University of Illinois at Urbana-Champaign
August 9, 1979
DRAFT: DO NOT CITE OR QUOTE.
AUTOMATION, EMPLOYEE CENTRALITY IN THE
PRODUCTION PROCESS, THE EXTENT TO WHICH
ABSENCES CAN BE ANTICIPATED, AND THE RE-
LATIONSHIP BETWEEN ABSENTEEISM AND
OPERATING EFFICIENCY: AN EMPIRICAL
ASSESSMENT
Michael K. Moch, Assistant Professor,
Department of Business Administration
Dale E. Fitzgibbons, Graduate Student
Department of Business Administration
#592
Summary (Abstract) is on the next page.
AUTOMATION, EMPLOYEE CENTRALITY IN THE PRODUCTION
PROCESS, THE EXTENT TO WHICH ABSENCES CAN BE
ANTICIPATED, AND THE RELATIONSHIP BETWEEN ABSENTEEISM
AND OPERATING EFFICIENCY: AN EMPIRICAL ASSESSMENT
ABSTRACT
Despite almost universal agreement that employee absenteeism leads
to decreased production efficiency, there is little documentation of a
negative relationship between these variables. Recently, Staw and
Oldham (1978) have even argued that absenteeism might lead to increased
productivity, at least at the individual level of analysis. The present
study hypothesizes and demonstrates that absenteeism and plant level ef-
ficiency are negatively associated 1) when production processes are not
highly automated, 2) when those who are absent are central to the pro-
duction process, and 3) when the absences cannot be anticipated in ad-
vance. Despite these limitations, the costs attributable to the impact
of absenteeism on plant operating efficiency were substantial. If ab-
senteeism is a function of employee satisfaction, quality of work pro-
grams designed to increase satisfaction are likely to result in con-
siderable savings by increasing operating efficiency. This will be
particularly true to the extent that they are conducted 1) in organi-
zations in which human input is central and 2) on employees within
these organizations who are central to the production process.
Proponents of programs designed to improve employee satisfaction and
the quality of working life often claim that such efforts, if successful,
will improve operating efficiency and effectiveness (e.g., Likert, 1961,
1967; Mills, 1975). An increasing amount of effort is being directed
toward determining whether or not this is true and, if true, toward mea-
suring the actual costs and benefits. Mirvis and Macy (1976) divide these
efforts into asset models and cost models. Asset models consider human
resources as corporate assets. Pyle (1970) suggests that assets invested
in human resources may be treated like any other assets and evaluated on
a cost-return basis. Asset models, however, do not deal adequately with
employee effectiveness and performance on-the-job. Consequently, they
are limited in their ability to assess the impact of employee morale on
plant effectiveness and efficiency. Cost models, however, focus directly
on employee behavior and attempt to evaluate the costs of behavior in
dollar amounts. When these behaviors are associated with employee sat-
isfaction and morale — e.g., absenteeism, turnover, tardiness — these
costs can be viewed as variable, amenable to reduction with increased
quality of work life.
Mirvis and Macy (1976) and Macy and Mirvis (1976) provide a good
deal of evidence documenting the costs of behaviors often associated with
employee morale, and they suggest that their evidence should stimulate
quality of work programs. Recently, however, Staw and Oldham (1978)
have argued that behaviors traditionally considered to be costly can,
under some conditions, be beneficial to the organization. They report
that for employees who are relatively dissatisfied with their jobs,
absenteeism is positively not negatively associated with productivity.
-2-
Presumably, absenteeism relieves dissatisfied employees of job-related
stress and allows them to be more productive when they return to work.
Staw and Oldham point out that efficiency at the individual level may
or may not be associated with organizational level performance. Ab-
senteeism may be negatively rather than positively associated with over-
all organization effectiveness and efficiency. However, as will be noted
below, even this is not as obvious as it may seem at first glance. The
relationships between absenteeism, turnover, tardiness and other be-
haviors thought to be costly to organizations must be established through
empirical research. This is particularly true when the assumption that
the costs of these behaviors outweight their benefits guides policy
decision-making (Willems, 1973). The present study, therefore, attempts
to empirically assess the impact of one type of behavior, employee ab-
senteeism, on organizational efficiency, the pounds produced and wasted
per labor hour. In addition to Staw and Oldham (1978), several studies
have failed to document a negative relationship between these variables
(e.g., Seashore, Indik, & Georgopoulos, 1960; Turner, 19bO; Argyle et
al., 1958; Ronan, 1963). Most of these, however, focus on individual
rather than organizational performance. The present study, in addition
to considering organizational level performance, uses "hard" criterion
measures and is based upon time series data gathered over a period of
two years.
Absenteeism and Organization Efficiency;
Some General Considerations
Macy and Mirvis (1976) identify several types of costs which can be
associated with absenteeism. Fringe benefits or salary paid to absent
-3-
personnel, supervision time spent finding replacements or training new
personnel, and unabsorbed overhead are just a few. These authors, like
many others (Metzner & Mann, 1953; Morgan & Herman, 1976; Steers &
Rhodes, 1978), also argue that absenteeism hinders operating effective-
ness and efficiency. Mirvis and Lawler (1977) report that employee
attitudes were associated with absenteeism in a midwestern bank. They
also were associated with the frequency of teller shortages or over-
payments to customers. While these authors did not report the relation-
ship between absenteeism and teller shortages or overpayments, the co-
variation of absenteeism with attitudes and of attitudes with shortages
suggests a positive relationship. Yet three considerations indicate
that a positive association between absenteeism and lost effectiveness
may not occur under all, and perhaps many, circumstances.
First, production methods used by industry often are designed pre-
cisely to avoid uncertainties associated with human operators. "Idiot
proof" jobs deplored by many job design researchers may succeed in making
many jobs "absentee proof." If standardized repetitive behaviors are all
that is required and if sufficient control and "fail safe" systems are
implemented, it may make little difference who is doing the job. So
long as someone is present, the product may be produced effectively and
efficiently. Organizations using highly automated technology requiring
little human intervention and discretion may be relatively immune to
negative effects of absenteeism among production personnel.
Such immunity, however, is not likely to extend to all personnel,
and this suggests a second consideration. Who is absent is likely to
be at least as important as the degree to which production processes
-4-
are automated. For example, the technical personnel required to keep
automatic processes operating must be present if the processes are to
be serviced and maintained. To generalize, absenteeism among personnel
who are central to the production process is likely to have a greater
impact on operating efficiency and effectiveness than is absenteeism
among those who are less central.
The third consideration is that absenteeism is likely to have a
negative effect on operating efficiency to the extent that it cannot be
anticipated and planned for in advance. Illnesses, family problems, or
other events seldom can be anticipated, and the organization has little
time to find adequate replacements or to reschedule production. Other
absences, such as vacations, however, can be anticipated well in advance
and the organization can schedule staff and production to minimize if
not eliminate production losses due to these absences.
These three considerations suggest that, despite almost universal
agreement that absenteeism is associated with decreased production ef-
ficiency, the relationship between these variables is far from obvious.
Organizations employing automated technology may suffer no losses from
absences of production personnel. Organizations may suffer little loss
from absences of peripheral personnel regardless of the technology they
use, and the costs of absenteeism may be minimized or eliminated to the
extent that it can be anticipated. These considerations do not go so
far as Staw and Oldham (1978) in suggesting that absenteeism can, under
some circumstances, be beneficial; however, they do indicate that the
relationship between absenteeism and efficiency should be assessed
empirically rather than simply assumed. Likewise, dollar costs of
-5-
absenteeism, to the extent that they exist, should be estimated em-
pirically rather than simply relying on estimates by experts.
Study Site and Method
Absenteeism and efficiency data were gathered from an assembly and
packaging plant employing approximately 750 persons. Well over half of
the floor space and personnel were devoted to packaging. Assemblers
prepared the material for packing, and maintenance personnel serviced
the extensive conveying and packaging machinery. Approximately 450
employees were assigned to the assembly department. The plant employed
about 90 maintenance personnel, including machinists, electricians,
and other skilled tradespersons. There were about 130 assemblers. By
far the most central activity, however, was packaging.
The packaging department was organized around several conveyors.
Product being carried by these conveyors was packaged more or less auto-
matically, depending upon the product and upon the line. Some products
were packaged almost totally by hand. Other product was packaged almost
totally by machines, never once coming in contact with a human hand.
Conveyors often were converted to allow for packaging different products.
There were, however, very few conversions on the most automated line.
Production on this line came close to Woodward's (1965) description of
continuous process flow.
Production plans were made weekly. Department superintendents at-
tended weekly planning meetings where product goals would be established.
They then would make personnel assignments. In the packaging depart-
ment, most employees were assigned to different lines, depending upon
-6-
the production plan. While some people had more or less permanent posi-
tions, most fell into a common labor pool and would be assigned to dif-
ferent lines on different weeks. A high proportion of absenteeism in
the packaging department as a whole therefore could affect production
efficiency on all operating lines. The lines were, in this sense, char-
acterized by pooled interdependence (Thompson, 1967) ; they all depended
upon the common labor pool. It therefore was possible to select products
for the efficiency analysis rather than rely upon an overall estimate
which would include variance due to product mix.
Two products were selected on the basis of their comparability and
the frequency — number of weeks — with which they were produced over the
two year period of the study. The first product was a speciality item
for the plant. It was produced almost continuously (N=103 weeks) on
the most automated line described earlier. The second product, a non-
specialty item, also was produced almost every week (N=101 weeks) ; how-
ever, production was less continuous than on the specialty line. Down-
time due to packaging department product changes on the line which ran
the non-specialty product averaged 3.35 hours per week. In contrast,
an average of 0.20 packaging department hours were spent in downtime
changing products on the line which ran the specialty product. The
difference between these averages is statistically significant (p < .05),
indicating that it was not due to vagaries of time sampling. While
these two products differed substantially on the basis of the extent
to which their production was automated, they were quite similar in
other respects. In fact, when conversions were made on the automated
line, the non-specialty product was one of the few which could be (and
-7-
was) produced on the line which usually ran the specialty product. Any
differences in the relationship between pounds producted or wasted per
labor hour and employee absenteeism for these two products therefore
could plausibly be attributed to the difference in production automa-
tion rather than to differences in the nature of the product.
Data and Measures
Data on plant operating efficiency were gathered separately for
each of the two products. Plant records were available for two one-
year periods, January 1 to December 31, 1977 and April 1, 1978 to
March 31, 1979. The data gathered included the number of direct
labor hours for both assembly and packaging which were allocated to
each of the two products under investigation. They also included the
number of pounds of each product produced for each week as well as the
number of pounds refuse. Refuse included material which was broken
2
and rejected; it was for all intents and purposes waste product.
Refuse pounds, pounds produced, and labor hours allocated for each
product provided the efficiency measures used in this study. Data on
the downtime due to changeovers discussed earlier were gathered from
copies of the weekly production plans.
Absenteeism data were available for each of the weeks for which
efficiency data had been gathered. It was possible to distinguish ab-
sences for each of the three major departments — packaging, assembly,
and maintenance. It also was possible to distinguish among several
different reasons for absences. Sickness, excused absences, and vaca-
tions were chosen for analysis, because they were both frequent and
varied in terms of the degree to which they allowed the organization
-8-
to anticipate employees ' lost time. Sicknesses were very difficult to
anticipate. Excused absences infrequently were arranged in advance.
Vacations could be anticipated weeks and sometimes months in advance.
The number of days absent for each reason for each employee were summed
for each week under study. These sums then were added to reflect the
number of absent days for all employees for each of the three major de-
3
partments for each of the three reasons for absences.
Insert Table 1 about here
Means and standard deviations for the days absent for each reason
for each department are presented in Table 1. Differences in averages
across departments generally reflect differences in department size;
however, it appears that maintenance personnel have a relatively higher
incidence of excused absences and packaging personnel have a relatively
lower incidence of vacation absences. The latter difference is probably
due to the fact that vacation time was associated with seniority and
packagers tended to be the least senior employees. Correlations among
the absence variables are presented in Table 2. Here it is clear that
vacation absences are associated across departments. While this reflects
the fact that the organization can plan for and make adjustments to re-
duce costs associated with vacation absences, the magnitude of these
correlations will make it difficult to separate out any effects of vaca-
tion absences in different departments.
Other patterns are evident in the data. Sickness and excused ab-
sences are significantly associated but only in the packaging department.
Sickness absences are positively associated across departments, perhaps
reflecting seasonal illnesses. Excused absences in packaging also covary
with sickness absences in assembly and in maintenance, due in part to
-9-
Table 1
Descriptive Statistics Reflecting Number of Absent Days
by Department and by Reason for Absence
(N = 103 Weeks)
X Days Absent
Packaging Department
Sickness Absences 176.6
Excused Absences 12.8
Vacation Absences 138.4
Assembly Department
Sickness Absences 22.6
Excused Absences 1.2
Vacation Absences 39.3
Maintenance Department
Sickness Absences 13.1
Excused Absences 4.0
Vacation Absences 27.0
Standard
Deviation
44,
.9
5.
.1
52,
.7
6,
.7
1,
.1
17,
.4
6,
.2
3,
,0
16,
.7
-10-
the association between illnesses and excused absences in packaging.
When maintenance personnel are sick, relatively few take vacations
(r = -.40*, p < .05). This suggests that vacations may be distributed
to adjust for illness absences; however this association was not evident
in the packaging or assembly departments. None of these correlations,
however, are so large as to preclude discriminating among the effects
of absences by department or by reason. Only vacation absences across
departments correlated highly enough to compromise clear cut discrimi-
nation.
Insert Table 2 about here
The Operational Model and Hypotheses
The general considerations discussed earlier guided the data gather-
ing. The efficiency measures, pounds produced and pounds refuse per
labor hour, assess organizational rather than individual level produc-
tivity. Likewise, the absenteeism data are aggregated to reflect de-
partmental absenteeism. The packaging department is clearly the most
central, at least for the less automated line, the line producing the
non-specialty product. The maintenance department was more central
for the more automated line because of the high degree of mechaniza-
tion and because of the relatively minor — almost observer — role played
by packaging personnel on this line. It was possible to anticipate and
therefore plan for vacation absences; this was less true for excused
absences and not at all the case for sicknesses. It therefore was ex-
pected that absences due to sicknesses and, perhaps, excused absences
would have a greater impact on plant efficiency than would absences due
to vacations.
-11-
Table 2
Pearson Product-Moment Correlations Among Measures of
Absenteeism by Department and by Reason for Absence
(N = 103 Weeks)
Packaging Department
1. Sickness Absences
2. Excused Absences
3. Vacation Absences
Assembly Department
4. Sickness Absences
5. Excused Absences
6. Vacation Absences
Maintenance Department
7. Sickness Absences
8. Excused Absences
9. Vacation Absences
<p < .05
.23*
.03
-.08
.31*
.16*
.00
.02
-.05
-.08
.02
-.01
-.01
.77*
.02
-.04
.34*
.21*
-.26*
.12
-.09
-.19*
.08
-.03
.17*
-.10
-.01
.12
-.12
-.07
-.11
.83*
-.11
-.10
.71*
-.40*
.17*
1
2
3
4
5
6
7
8
-12-
These considerations specify a rather complex model describing the
relationship between absenteeism, labor hours, and pounds of product
and refuse produced. Specifically, the impact of labor hours on pounds
produced or rejected is seen to be a function of the level of absenteeism.
Under conditions of high absenteeism, the slope reflecting the number of
pounds produced per labor hour is expected to be greater under conditions
(weeks) of low relative to high absenteeism. For pounds refuse, the re-
lationship is expected to be reversed. The slope reflecting the number
of pounds refuse per labor hour is viewed as being greater under con-
ditions (weeks) of high as opposed to low absenteeism.
It also is expected that the impact of absenteeism will vary de-
pending upon (1) the degree of automation in the production process,
(2) the relative centrality (department) of those who are absent, and
(3) the extent to which the absences can be anticipated. The expected
interaction between absenteeism and labor hours as they effect pounds
of product and refuse produced is therefore expected primarily,
(1) when production procedures are not automated and the absences are
those of central personnel (i.e., for the non-specialty product when
packaging department absences are high), (2) when production procedures
are automated and the absences are those of central personnel (i.e.,
for the specialty product when maintenance department absences are
high) , and (3) when the absences cannot be anticipated and planned for
(i.e., when absences are due to sicknesses or excused reasons rather
than vacations) .
Tests for these interactions involved comparisons between models
which did and did not allow for a differential impact of labor hours on
pounds product and pounds refuse produced. To get a baseline assessment,
-13-
a model which did not specify absenteeism- labor hours interactions was
estimated. This model took the following form:
Y = aX1 + gX2 + yX3 + C (1)
where,
Y = pounds product or refuse produced
X- = labor hours assigned in packaging
X„ = labor hours assigned in assembly
X„ = number of days absent for one of the three reasons
in one of the three departments
C = a constant
Once the coefficients in equation (1) had been estimated and a
2
value for variance explained, R (1), had been obtained, the data were
run again. This time, however, they were run against a model allowing
for separate estimates of the impact of labor hours for conditions of
high versus low absenteeism. This model took the following form:
where,
Y = alXla + a2Xlb + BX2 + ^X3 + C (2)
Y = pounds product or refuse produced
X.. = labor hours assigned in packaging for weeks exper-
iencing greater than average absenteeism for one of the
three reasons for one of the three departments (low
absenteeism weeks were coded zero on this variable)
X-. = labor hours assigned in packaging for weeks exper-
iencing less than average absenteeism for one of the
three reasons for one of the three departments (high
absenteeism weeks were coded zero on this variable)
X0 = labor hours assigned in assembly
X_ = number of days absent for one of the three reasons
in one of the three departments
C = a constant.
-14-
Regression coefficients a., and a„ in equation (2) provided independent
estimates of the number of pounds product or refuse produced per labor
hour under conditions of high versus low absenteeism for each department
for each reason. Differences in estimates of the variance explained by
2 2
equation (1), R (1), and by equation (2), R (2), provided a means for
assessing the significance of the differences in slopes. Since,
R2(2) - R2(l)
[1 - R (2)] /(number of weeks - 5)
has an F distribution with 1 and the number of weeks - 5 degrees of
freedom, the statistical significance of the difference in slopes for
high versus low absenteeism could be assessed (Nie et al., 1975:389).
Initial regressions based upon equation (1) revealed substantial
amounts of autocolinearity. Values of the Durbin-Watson d statistic
tended to be very close to 1.0. Accordingly, the Cochrane-Orcutt tech-
nique was used to transform the measures to reduce correlation among
first-order residuals (Johnston, 1963:192ff). The Durbin-Watson d es-
timated using the transformed data seldom was less than 2.0 for the
subsequent regressions. Even then, d tended to be very close to 2.0.
RESULTS
Regression coefficients measuring pounds product produced per
packaging department labor hour for weeks with high and with low ab-
senteeism are presented in Tables 3 and 4. It had been expected that
sicknesses and excused absences in packaging would decrease pounds pro-
duced per packaging labor hour, especially for the non-specialty item.
The coefficients in Tables 3 and 4, however, show considerable stability
-15-
across conditions of low versus high packaging absenteeism. For what-
ever reason they are absent, absences in the packaging department do not
appear to constrain pounds produced per labor hour for either product.
The trend, in fact, is in the opposite direction. Perhaps packaging
personnel who report in sick or have an excused absence are not replaced.
Any losses in production, therefore, may be balanced by savings in terms
of allocated labor hours. This, however, is unlikely. The plant main-
tains a pool of personnel from which replacements can be made on any
Q
particular line. A more likely possibility is that production pro-
cedures for both products studied are sufficiently standardized that
almost anyone can perform the production tasks and produce the prescribed
amount of product. As will be seen below, however, they may not do so
with equal amounts of refuse.
Insert Tables 3 & 4 about here
The results are somewhat different for absences in the maintenance
department. The trend is reversed. As expected, with the exception of
vacation absences, absences in the maintenance department are associated
with higher levels of production per packaging labor hour. While these
differences are substantial for the non-specialty product, they are
statistically significant only in the case of the specialty product and
for excused absences.
Since maintenance personnel are more central to production of the
more automated specialty product, this relationship was expected. How-
ever, the effect of sickness absences for maintenance personnel also
was expected. The relatively small difference in production per labor
hour for high and low maintenance sickness absences may be due to dif-
ferences in the quality of those who are absent. Those responsible
-16-
Table 3
Pounds Product Produced Per Labor Hour (Regression Coefficients)
by Absenteeism Level and Department for
Three Reasons for Absenteeism
(Specialty Product, N = 103 Weeks)
Pounds Product Produced Per Labor Hour
Packaging
As
sembly
Maintenance
Reason for
Absent
eeism
Abs
enteeism
Absenteeism
Absenteeism
Low
High
Low
High
Low
High
Sickness
107
108
111
109
108
107
Excus ed
108
111
108
114
118 *
106
Vacations
109
114
110
112
107
112
^Difference between coefficients significant p < .05. Two tailed tests
were used, because it was possible for absenteeism to actually increase
pounds produced per labor hour. If employees were not replaced and
those remaining performed the duties of those missing, pounds per hour
would have increased rather than decreased as a result of absenteeism.
-17-
Table 4
Pounds Product Produced Per Labor Hour (Regression Coefficients)
by Absenteeism Level and Department for
Three Reasons for Absenteeism
(Non-Specialty Product, N = 101 weeks)*
Pounds Product Produced Per Labor Hour
Reason for
Absenteeism
Packaging
Absenteeism
Low High
Assembly
Absenteeism
Low High
Maintenance
Absenteeism
Low High
Sickness
297
295
297
311
317
295
Excused
294
315
297
307
300
282
Vacations
287
294
271
302
297
295
^Difference between coefficients significant, p < .05, two-tailed
test.
-18-
enough to secure excuses — a procedure which, in this plant, can require
documentation — may also be those who more responsibly fulfill their work,
duties. There is no documentation for this possibility, however, and it
must remain speculative.
In sum, absenteeism among packaging personnel did not seem to af-
fect the number of pounds of product per labor hour the department was
able to produce. Absences among maintenance personnel, however, appear
to be associated with fewer pounds of product produced per packaging
labor hour. As expected, this effect is pronounced for the more auto-
mated specialty product; however, this is true only for excused absences.
Vacation absences can be anticipated. There is no relationship between
vacation absences and pounds product produced for either packaging or
maintenance personnel. There is no relationship between absences in
the assembly department and pounds product per labor hour for either
product or for any of the three absence reasons. The absence of any
relationship for assemblers provides support for the contention that
centrality to the production process affects the impact of absenteeism
on plant efficiency.
Insert Tables 5 & 6 about here
The number of pounds refuse per packaging labor hour for high and
for low absence weeks by department and reason for absence are presented
in Tables 5 and 6. Coefficients presented in these tables provide con-
siderable support for the hypotheses. Trends for both products show
the pounds refuse produced per packaging labor hour to be greater under
conditions of high as opposed to low packaging absenteeism, except, as
expected, for vacation absences. Also as expected, the relationship
between absenteeism and pounds refuse per labor hour is particularly
-19-
Table 5
Pounds Refuse Produced Per Labor Hour (Regression Coefficients)
by Absenteeism Level and Department for
Three Reasons for Absenteeism
(Specialty Product, N = 103 weeks)*
Pounds Refuse Produced Per Labor Hour
Reasons for
Absenteeism
Packaging
Absenteeism
Low High
Assembly
Absenteeism
Low High
Maintenance
Absenteeism
Low High
Sickness
7.00
7.67
8.95
7.68
7.51
8.01
Excused
7.42
7.95
7.35
7.41
6.91
7.83
Vacations
7.42
7.76
8.36
6.61
8.05
6.96
*Difference between coefficients significant at p < .05 (one-tailed
test).
-20-
Table 6
Pounds Refuse Produced Per Labor Hour (Regression Coefficients)
by Absenteeism Level and Department for
Three Reasons for Absenteeism
(Non-Specialty Product, N = 101 weeks)
Pounds Refuse Produced Per Labor Hour
Reasons for
Absenteeism
Packaging
Absenteeism
Low High
Assembly
Absenteeism
Low High
Maintenance
Absenteeism
Low High
Sickness
1.00
*
2.93
0.93
C.40
-2.27 *
0.91
Excused
1.81
*
3.99
1.57
0.37
2.71
1.37
Vacations
1.88
1.85
1.57
1.71
1.81
0.80
*Difference between coefficients significant at p < .05 (one-tailed
test).
-21-
pronounced and statistically significant only for the less automated
non-specialty product. This relationship, as anticipated, holds only
for sicknesses and excused absences. These absences were shown in
Table 2 to be slightly but significantly associated (r = .23). It is
unlikely that this small an association could completely account for
the similarity in the patterns of coefficients; however, some confound-
ing is possible. In any case the effect of these two absences is not
seen in the case of vacations. In fact the direction is slightly re-
versed. It appears that while absences in the packaging department do
not affect the pounds product produced per invested labor hour, they
do increase the pounds refuse produced by the replacements for those
who are absent. Vacation absences can be anticipated and planned for.
These absences have little or no effect on the amount of product or
refuse produced.
It had been expected that sicknesses and excused absences in the
maintenance department would be associated with greater refuse per labor
hour, particularly for the specialty product. The trends support the
expectation that maintenance absences increase waste; however the only
statistically significant relationship occurs with the non-specialty
y
rather than with the specialty product. While production of the non-
specialty product is less automated than that of the specialty product,
the non-specialty line is nonetheless highly mechanized. Maintenance
personnel are required on both lines. It is possible that automation
minimizes human errors and therefore that the impact of absenteeism on
waste per labor hour will be minimal, even for those responsible for
maintaining the equipment. None of the differences in Table 5 attain
-22-
statistical significance. The impact of absenteeism of both packaging
and maintenance personnel, however, is evident in Table 6. Sickness
absences across these departments were associated (r = .34), so some
confounding is possible. However, packaging sickness also was asso-
ciated with sickness absences in the assembly department and, as ex-
pected, the relationship between assembly department sickness or ex-
cused absences and waste per labor hour was not pronounced. In fact,
the direction tends to be reversed, with low absence weeks showing
greater refuse than high absence weeks. Also as expected, no substan-
tial differences were evident for vacation absences.
SUMMARY AND DISCUSSION
It appears that the impact of absenteeism on operating efficiency
occurs primarily to the extent that production is not automated. Ex-
cused absences in the maintenance department appeared to decrease pounds
product per labor hour for the specialty product; however, maintenance
absences were not significantly associated with refuse per labor hour
for this more automated product. Besides reducing vulnerability to ab-
senteeism of production personnel, automation may limit the impact of
any absenteeism on waste. The only way absenteeism may affect operating
efficiency when production is highly automated may be in terms of pounds
product produced per invested labor hour and then only when maintenance
personnel are absent for reasons which cannot be anticipated.
Less automated production seems to be more vulnerable to efficiency
losses traceable to absenteeism. This vulnerability, however, seems
focused on refuse rather than on pounds product per labor hour. This
finding is similar to that reported by Seashore, Indik, and Georgopoulous
-23-
(1960). This plant, like many others, based its planning on weekly pro-
duction goals. Because plans for product distribution required the plant
to meet these goals, management had little freedom to vary the number
of pounds product produced. The same number of pounds had to be pro-
duced, even if many key personnel were absent. The only degrees of free-
dom left to management, therefore, involved the speed of the line, the
number of line-interruptions, or the number of personnel assigned to
produce the product. Assigning additional personnel is costly. The
labor union in this plant, as in others, was very sensitive about line
speedups. The line of least resistance, therefore, may be to decrease
the number of line interruptions. Lines may be stopped to make adjust-
ments for product quality. They may be stopped, because mechanical
problems may be decreasing the percentage of packageable product.
These problems create refuse. Yet stopping the line would reduce the
number of pounds of product produced and possibly keep the plant from
meeting its production goals. The only cost-effective alternative may
be to absorb costs in terms of higher refuse in order to avoid costs
associated with the underutilization of distribution facilities and,
perhaps, decreased sales and a permanent loss of customers.
If this sort of tradeoff occurs, it would explain why absenteeism ap-
pears to be more highly related to the production of waste than to the num-
ber of pounds produced. Relatively inexperienced personnel must produce the
same amount of product as their more proficient counterparts. In the pro-
cess, quality is likely to suffer and refuse accumulate. This will happen,
however, only to the extent that production is not highly automated and the
-24-
absences cannot be anticipated. The absences also must be those of per-
sonnel central to the production process. More refuse per packaging
labor hour was produced during weeks of high packaging absences for sick-
nesses and for excused reasons than during weeks of relatively low ab-
senteeism. Sicknesses in the maintenance department were also associated
with more refuse produced. These effects, however, occurred only for
the less-automated non-specialty product. No relationship between ab-
senteeism and refuse was documented for the more automated specialty
product, and no relationship was evident between absenteeism in the less
central assembly department and refuse.
Production of both the products studied here was highly mechanized.
Production of the specialty product is best characterized as continuous
process flow (Woodward, 1965) . Even a high degree of mechanization, how-
ever, does not appear to have insulated production of the non-specialty
product from the effects of absenteeism on operating efficiency. For
example, the difference in pounds of non-specialty product refuse pro-
duced during high versus low packaging sicknesses is 1.93 pounds per
labor hour. An average of 462 packaging department hours were allocated
to this product every week. This means that 1.93 x 462 = 892 more pounds
of this product were lost to refuse during high as opposed to low ab-
senteeism weeks. Exactly half of the weeks studied were above average
in packaging sickness absenteeism. Sickness absenteeism, therefore,
may be held responsible for 892 x 50.5 = 45,046 pounds of this product
lost due to sickness absenteeism during the course of the study. This
product retails for approximately $.85 for a half pound container. If
5% of this cost represents retail mark-up and 25% represents transpor-
tation and the costs of packaging materials, then each pound lost to
-25-
refuse represents (.85 x 2). 70 = $1.19 lost income. This totals
45,046 x $1.19 = $53,605 lost due to sicknesses in the packaging depart-
ment during the course of the study. This does not include the effect
of excused absence or illnesses in the maintenance department; although
the effects of these factors are confounded with those of packaging
sicknesses. It also does not include the costs of these absences which
were absorbed by other products and other lines. When variable costs
of absenteeism other than production efficiency such as fringe benefits
paid out, costs associated with maintaining a labor pool sufficient to
provide replacements, etc., the costs of absenteeism total much higher
than the $26,803 annually lost on the non-specialty product for sick-
ness absences in packaging. Aggregated to the national level, the esti-
mate of $26.4 billion (Steers & Rhodes, 1978) annually therefore may not
be out of line. These calculations indicate that, if they are success-
ful, programs such as quality of work life experiments designed to reduce
absenteeism will result in considerable savings.
The findings suggest some strategies which could increase the im-
pact of quality of working life programs on operating efficiency. For
example, by increasing cooperativeness and trust, a greater proportion
of absences might be anticipated and planned for in advance. Employees
who know they will be absent may be willing to communicate this in ad-
vance to the extent that they are concerned with plant efficiency and
feel they will not be punished for being absent for what may be a reason
which is difficult to justify. The data suggest that knowing about
absences in advance is just as effective as preventing them, at least
in terms of minimizing their impact on operating efficiency. Secondly,
-26-
quality of work programs might usefully begin by focusing upon those who
are central to the production process. Certainly, programs directed pri-
marily toward more peripheral personnel cannot hope to secure the rate
of benefits suggested here. Third, programs directed toward reducing
absenteeism in order to increase operating efficiency are likely to be
more effective in less automated settings where human input explains a
substantial portion of the variance in efficiency.
The present study documents some gains and losses attributable to
absenteeism of employees who perform different functions in the organi-
zation. The approach avoids problems associated with supervisor ratings
of effectiveness by using time series data and "hard" criterion measures.
However, there is no reason why this approach cannot be used to assess
the impact of a less tangible factor, that of employee satisfaction on
production efficiency. It has often been argued that such an associa-
tion exists; however, it has been very difficult to document (Brayfield
& Crockett, 1955; Vroom, 1964; Schwab & Cummings, 1970). It is possible
that the degree to which work procedures are automated or even stan-
dardized, the extent to which the satisfied employees play a central role
in the production process, and other constraints which limit the extent
to which employees' can affect overall production efficiency may account
for some of the inconsistent or inconclusive findings. There is no
reason, however, why time series data cannot be used to compare plant
level performance for weeks when central personnel who report high
levels of satisfaction are present versus weeks when they tend to be
absent. Such a comparison may show greater efficiency for weeks when
the satisfied personnel are present rather than absent. At least to
-27-
the extent that production is not highly automated. The same procedure
may be employed to investigate the effects of other factors, such as
role stress, intrinsic motivation, job involvement, etc. This proce-
dur would more closely tie productivity to employee attitudes through
employee behaviors than has generally been done in the past (e.g.,
Likert and Bowers, 1973). Until this is done, however, the present
study indicates that, to the extent that they lead to lower sickness
and excused absence rates, employee attitudes will increase organiza-
tional performance.
-28-
FOOTNOTES
records were kept for 12 months, then discarded. When the re-
searchers arrived at the site in mid April, 1979, to gather the second
year data, data from January 1 to March 31, 1978 had been destroyed.
2
''Some material was recyclable. These pounds were not included as
either production or refuse.
3
In studies relating absenteeism to employee attitudes, the num-
ber of absences rather than the number of days absent is usually pre-
ferred (Chadwick- Jones et al., 1971; Kuse & Taylor, 1962; Metzner &
Mann, 1973). However, lost production efficiency, if there is any, is
likely to be due to the fact that the needed individual is absent on a
particular day. The frequency of sicknesses, excused absences, or
vacations is likely to be less salient. Accordingly, the number of
days absent rather than the frequency of absences was calculated.
4
Due to the limited number of weeks available for analysis, it was
not possible to assess the absenteeism-labor hour interaction simul-
taneously for all absence types (reasons by departments) . Testing for
interactions one absence type at a time is justified, given the gen-
erally low correlations among absenteeism measures. Possible limita-
tions of this procedure are further minimized by the fact, to be shown
later, that the only absence types to be highly correlated (vacations
across departments) did not affect pounds product or refuse per labor
hour. Where confounding is possible, however, it is discussed in the
text.
The constant was included for three reasons. First, there ap-
peared to be a minimum number of pounds — product and refuse — that could
be produced with maximum (or minimum) feasible absenteeism. Second,
it was necessary to focus on what Macy and Mirvis (1976) call variable
costs, those which can be affected by individual behaviors. These
costs in pounds would be those above the minimum, given maximum (or
minimum) feasible absenteeism. Third, it is conceivable that competitors
of the plant could identify it as the subject of this research. They
also might identify the products under study. If this were to happen,
competitors would obtain valuable information concerning operating
efficiency. Addition of the constant term precludes this possibility,
and for this reason the values of these terms will not be reported
here.
All regresesions were statistically significant, p < .05. The
number of days absent seldom had a significant main effect on pounds
product or refuse produced. Only coefficients for two absence types
were statistically significant (p < .05; one positive the other nega-
tive), just about the number expected by chance -1.8.
-29-
Differences in regression coefficients across the tables are due
to differences in the nature (e.g., weight) of the products and in pro-
duction methods.
g
This is a cost of absenteeism which is not under investigation
here.
9
It is particularly interesting that increasing the number of
labor hours actually may reduce pounds refuse under conditions of low
sickness absenteeism in the maintenance department. If refuse is not
created by workers operating faulty machines, it seems plausible that
the addition of workers to tend properly functioning equipment could
actually reduce errors and waste.
M/C/144
-30-
REFERENCES
Argle, M., G. Gardner, and F. Cioffi
1958 "Supervisory Methods Related to Productivity, Absenteeism,
and Labor Turnover." Human Relations, 11:23-40.
Brayfield, A. H. and W. H. Crockett
1955 "Employee Attitudes and Employee Performance." Psycho-
logical Bulletin, 52:396-424.
Chadwick- Jones, J. K. , C. A. Brown, N. Nicholson and C. Sheppard
1971 "Absence Measures: Their Reliability and Stability in
an Industrial Setting." Personnel Psychology, 24:
463-470.
Huse, E. F., and E. K. Taylor
1962 "The Reliability of Absence Measures." Journal of Applied
Psychology, 46:159-160.
Johnston, J.
1963 Econometric Methods. New York: McGraw-Hill.
Likert, Rensis
1961 New Patterns of Management. New York: McGraw-Hill.
Likert, Rensis
1967 The Human Organization: Its Management and Value.
New York: McGraw-Hill.
Likert, Rensis and D. Bowers
1973 "Improving the Accuracy of P/6 Reports by Estimating
the Change in Dollar Value of the Human Organization."
Michigan Business Review, 25:15-24.
Macy, B. A., and P. H. Mirvis
1976 "Measuring Quality of Work and Organizational Effective-
ness in Behavioral-Economic Terms." Administrative
Science Quarterly, 21:212-226.
Metzner, H. , and F. Mann
1953 "Employee Attitudes and Absences." Personnel Psychology,
6:467-485.
Mills, Ted
1975 "Human Resources — Why the New Concern." Harvard Business
Review, 53: .
Mirvis, P. H., and E. E. Lawler III
1977 "Measuring the Financial Impact of Employee Attitudes."
Journal of Applied Psychology, 62:1-8.
-31-
Mirvis, P. H., and Barry Macy
1976 "Human Resource Accounting: A Measurement Perspective."
Academy of Management Review, 1:74-83.
Morgan, L. G., and J. B. Herman
1976 "Perceived Consequences of Absenteeism." Journal of
Applied Psychology, 61:738-742.
Nie, Norman H., C. Hadlai Hull, J. Jenkins, K. Steinbrenner, and
D. H. Bent
1975 Statistical Package for the Social Sciences, 2nd edition.
New York: McGraw-Hill.
Pyle, W.
1970 "Monitoring Human Resources — On Line." Michigan Business
Review, 19-32.
Ronan, W. W.
1963 "A Factor Analysis of Eight Job Performance Measures."
Journal of Industrial Psychology, 1:107-112.
Schwab, D. P., and L. L. Cummings
1970 "Theories of Performance and Satisfaction: A Review."
Industrial Relations, 9:408-430.
Seashore, S. E., B. P. Indik, and B. S. Georgopoulos
1960 "Relationships Among Criteria of Job Performance."
Journal of Applied Psychology, 44:195-202.
Staw, Barry M. , and Greg R. Oldham
1978 "Reconsidering Our Dependent Variables: A Critique and
Empirical Study." Academy of Management Journal, 21:
539-559.
Steers, Richard M. , and Susan R. Rhodes
1978 "Major Influences on Employee Attendance: A Process
Model." Journal of Applied Psychology, 63:391-407.
Turner, W. E.
I960 "Dimensions of Foreman Performance: A Factor Analysis
of Criterion Measures." Journal of Applied Psychology,
44:216-223.
Vroom, Victor
1964 Work and Motivation. New York: Wiley.
Williems, E. P.
1973 "Go Ye Into All the World and Modify Behavior: An Ecolo-
gist's View." Representative Research in Social
Psychology, 4:93-105.
Woodward, Joan
1965 Industrial Organization: Theory and Practice. London:
Oxford University Press.
t\v