Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Omega, Int. J. Mgmt Sci. Vol. 25, No. 3, pp.

285-299, 1997
© 1997ElsevierScienceLtd. All rights reserved
~ Pergamon Printed in Great Britain
PII: S0305-0483(96)00063-1 0305-0483/97$17.00+ 0.00

Overtime Schedules for Full-time


Service Workers

FF EASTON
Syracuse University, Syracuse, NY, USA

DF ROSSIN
University of Michigan-Dearborn, Dearborn, MI, USA

(Received September 1995; accepted after revision November 1996)

Part-time employees help service organizations extend their operating hours and provide extra
capacity for peak demand periods. While this strategy tends to increase the number of employees
needed to staff the system, part-timers usually earn less per week than full-timersbecause they don't
work as many hours. However, escalatingper capita labor expenseshave increasedthe effectivehourly
wages for part-timers, threatening one of their key advantages. According to governmentstatistics,
service sector employersnow obtain more labor from overtime work than they do from part-timers.
Although the benefitsof part-time schedulingpoliciesare wellunderstood,comparativelylittle research
has focused on overtime schedulingpolicies. Typically, we think of overtime as a means of buffering
service systems against supply and demand uncertainty. In this study, however, we demonstrate that
scheduled overtimeprovidesmany of the same operational advantagesof part-time schedulingpolicies.
We evaluate the effects of alternative overtime staffing and scheduling policies on important
performance measures such as total labor expense, labor utilization, and workforce size. Compared
with standard (40 hours per week) employee schedules, we find that even small amounts of
premium-pay overtime work provide significant savings. We also find that the ideal workforce size
and proportion of overtime work for a given scheduling policy seem to be relatively insensitiveto
changes in per capita labor expenses. This means that employers may need much more aggressive
overtime schedulingpoliciesto mitigate the effects of risingper capita labor expenses. © 1997 Elsevier
Science Ltd

Key words--labor scheduling, human resources, mathematical programming

1. INTRODUCTION c u s t o m e r retention rates, a n d generate favorable


w o r d - o f - m o u t h advertising. These factors in-
T o VARYING DEGREES, most o r g a n i z a t i o n s crease v o l u m e a n d help provide scale economies,
embrace high quality service as a c o m p o n e n t o f widening profit m a r g i n s even further [3, p. 1 ~].
their overall competitive strategy. C u s t o m e r T o e n h a n c e c u s t o m e r perceptions a b o u t
perceptions of service quality are influenced access, m a n y o r g a n i z a t i o n s extend their service
by several factors [1, pp. 15-32], i n c l u d i n g hours b e y o n d the c u s t o m a r y eight-five work
reliability (the ability to p e r f o r m the expected day, to weekends, a n d increasingly, a r o u n d the
service d e p e n d a b l y a n d accurately), responsive- clock. D u r i n g these times, the d e m a n d for ser-
ness (the willingness to help customers a n d vice m a y vary significantly [4-7]. Usually, the
provide p r o m p t service), a n d access ( m a k i n g the p o r t i o n of the service that involves contact
service available at c o n v e n i e n t times). By im- with the c u s t o m e r c a n n o t be p o s t p o n e d for
p r o v i n g c u s t o m e r perceptions o f these a n d other long w i t h o u t degrading c u s t o m e r perceptions
service quality attributes, service o r g a n i z a t i o n s a b o u t reliability a n d responsiveness [8,9].
can c o m m a n d p r e m i u m prices [2], i m p r o v e T o score high o n these dimensions, service

285
286 Easton, Rossin--Overtime Schedules

Table 1. Service sector employee benefits as percent of total compensation [17]


Mo/ Total Legally[*] Paid Healthandli~ Reti~mentand Other
year benefits required leave insurance savings benefits
3/88 25.7% 8.5% 7.1% 5.0% 3.0% 2.1%
3/89 25.5% 8.7% 7.0% 5.2% 2.7% 1.9%
3/90 26.0% 8.7% 6.9% 5.6% 2.8% 2.0%
3/91 26.0% 8.8% 6.8% 6.0% 2.5% 1.9%
3/92 26.5% 8.9% 6.7% 6.3% 2.5% 2.1%
3/93 26.9% 9.0% 6.5% 6.5% 2.5% 2.4%
* Legallyrequired:SoeialSecurity, Stateand FederalUnemployment, and WorkersCompensation

organizations must have the capacity to satisfy the Employee Retirement Income Security Act,
customer demand soon after it is presented. and the 1986 Tax Reform Act, has greatly
In the short run, the capacity of a service expanded the scope of employment arrange-
organization is largely determined by the ments subject to such payments [16]. Since that
number of workers it employs and the times at time, legally required payments and employer
which these employees are scheduled for duty. subsidized benefits have increased steadily [17].
Roughly two-thirds of all full-time employees As shown in Table 1, these expenses now exceed
are scheduled to work 5 days per week, 8 hours 26% of the total US service sector payroll
per day [10]. Most (80%) full-time employees expense.
start work between 07:00 and 09:00 and finish Where applicable, these per capita expenses
between 16:00 and 18:00, according to Mellor increase the effective hourly wage rates for
[11]. To help extend their operating hours part-time employees, threatening one of their
beyond these times and to respond to demand key advantages. As indicated in Table 2, service
peaks that occur during the normal work day, organizations have started reducing the pro-
many service organizations rely on part-time portion of total labor obtained from part-
(less than 35 hours per week) employees [12, 13]. timers. Part-time labor appears to have been
Total labor expenses tend to be lower when replaced by full-time employees working over-
some employees have part-time schedules [14]. time (more than 40 hours per week) schedules.
Frequently, however, part-time staffing and Employers exploit overtime work schedules
scheduling policies require more employees to get for a variety of reasons. For example, although
the job done than might be necessary with full- overtime work usually requires premium pay it
time employees alone [14]. Some labor-related can sometimes be used to help reduce the total
expenses, including pension and retirement bene- number of permanent employees needed to staff
fits, life insurance and health care benefits, unem- the system, and thus total per capita expenses.
ployment insurance premiums, paid vacations, Overtime may also be used to respond to
sick leave, paid holidays, as well as recruiting and unforeseen events such as absenteeism or higher
training costs, vary with the number of than expected demand. In effect, overtime can
employees rather than the number of hours each function as a 'safety stock' that buffers the
employee works [15]. Recent US legislation, such system against uncertainties in either supply or
as the Tax Equity and Fiscal Responsibility Act, demand. Finally, overtime may be used on a

Table 2. Average weekly work hours for service sector* employees [18]
Percent of service sector employees working
Number of employees Average weekly Part-time Full-Time
Date ( x t000) h/employee ( < 35h/week) (35 < h/week < 40) Overtime ( > 40h/week)
3/84 58,349 36.9 24.98% 51.84% 23.18%
3/85 60,670 37.4 24.51% 50.19°/0 25.300/0
3/86 62,645 37.4 24.41% 50.22% 25.370/0
3/87 64,917 37.6 23.85% 49.71% 26.44%
3/88 66,482 37.9 23.49% 49.55% 26.96%
3/89 68,181 37.8 23.98 % 48.92% 27.10%0
3/90 69,227 38.0 23.32% 48.83% 27.84%
3/91 69,649 37.8 23.91% 49.65% 26.44%
3/92 70,855 38.1 23.15% 48.98% 27.88%
3/93 72,174 38.1 23.24% 49.15% 27.61%
* Includes transportation, communications, public utilities, wholesale trade, retail trade, finance, insurance, and real estate, and services.
Omega, Vol. 25, No. 3 287

more prolonged basis to help meet seasonal or Z'/40 = 0.0065 Z') is only about 5-7 of the
cyclical demand peaks, especially when work- customary 50% hourly wage premium for
force change costs are high. overtime work (0.5 × [0.74 Z'/40] = 0.00925
From a macro-economic perspective, the Z'). Therefore, the observed growth in overtime
decision to use overtime seems straightforward. use (Table 2) is not adequately explained by our
The amount of overtime labor that a service simple macro-economic model (equation (3)).
firm can use is typically constrained by the This leads us to conclude that the growth of
number of employees available to work overtime work is being driven by factors other
overtime schedules [19, 20]. To illustrate, sup- than the simple avoidance of per capita
pose each full-time employee works at least H expenses.
hours per week at an average wage of C/h, with Unfortunately, comparatively little is known
fringe benefits and legally required payments about the operational benefits of overtime
totaling F/employee. Assume that employees scheduling flexibility. To better understand the
can work up to G x H additional hours each effects of overtime scheduling policies on
week (0 < G < 1), earning premium wage P/h. important performance measures like total
If the firm needs W hours of labor, its total payroll expense and workforce size, we adapted
labor expense for the period (Z') is determined a tour scheduling model [21] to incorporate
by the number of employees (K) and the overtime scheduling options for hypothetical
number of overtime hours worked. Total labor service firms. Our initial goal was to determine
expense for the firm can be approximated with whether service firms can achieve performance
the relationship: advantages (compared to standard 8 hours per
day, 5 days per week schedules) with the use of
Z' = K(F + HC) + P ( W - - KH) , (1) planned overtime work (at premium pay). Since
overtime work may disrupt employees' personal
where lives, we investigated both conservative and
aggressive overtime scheduling policies. We
W hoped to find a few relatively benign overtime
K> H(1 + G ) (2) scheduling policies that could produce signifi-
cant performance improvements (again, relative
Each additional employee increases total labor to standard 8 hours per day, 5 days per week
expense by F + HC, but eliminates H hours of schedules). Finally, since per capita payroll
overtime expense. Assuming (1) is differentiable expenses are likely to continue their upward
in K, the ideal workforce size (K*) is: trend, we wanted to project how increases are
likely to affect employment levels and overtime
W F W use.
K * = ~ if ~ < (P -- C), H(1 + G ) The results of our experiments confirmed that
practitioners can realize considerable savings by
otherwise. (3) utilizing overtime scheduling policies. The
results also yielded some useful managerial
That is, when the incremental hourly insights. For example, we found that in some
premium paid for overtime work ( P - C) operating environments, about 95% of the
exceeds the pro-rated hourly per capita expense maximum attainable benefit from overtime
per employee (F/H), overtime work should be work can be achieved with employee schedules
avoided. When the premium is less than the that limit overtime work to just 6 hours per
prorated hourly per capita expense, the firm employee per week. Contrary to our expec-
should maximize the amount of overtime labor tations, however, we found that the ideal
it uses. workforce size and the ideal amount of
As shown in Table 1, the average per capita scheduled overtime tend to be relatively
expense for a full-time employee in the US is insensitive to per capita expense rates. We
now more than 26% of total wages. Assuming discuss these and other findings, along with
an average work week of H = 40 h and total other experimental details, in Sections 3 and 4.
weekly wages of Z ' per employee, the current In Section 2, we present a critical review of the
average prorated per capita expense (0.26 relevant literature.
288 Easton, Rossin--Overtime Schedules

2. FLEXIBLE LABOR STAFFING A N D of staffing and scheduling decisions, labor


SCHEDULING POLICIES staffing and scheduling decisions are frequently
modeled as tour scheduling mathematical
The contractual relationships between service programs [5, 6, 13, 14, 26-35]. Their purpose is
organizations and employees are by no means to simultaneously determine the number of
standard. The labor once provided by tra- employees, and the way they should be
ditional permanent, full-time, 40 hours per week scheduled, to satisfy the predicted demand for
employees is increasingly supplied under flexible each period in the planning horizon at minimum
or contingent staffing arrangements such as cost. In this problem, each feasible employee
leased employees, consultants, freelancers, self- schedule (tour) is represented by a separate
employed workers, temporary employees, and integer variable, whose value is the number of
part-timers [22]. Many of these non-traditional employees that should be assigned to that
employment arrangements have been shown to schedule. Each tour must conform to the
yield important organizational benefits for the scheduling constraints that are explicitly or
firm, including reduced absenteeism and implicitly embodied in the contractual relation-
turnover [23]. Despite the popularity of these ships between the service firm and its workers.
arrangements, however, a significant and in- We use the following notation to characterize
creasing fraction of all US service workers this complex, economically important schedul-
(more than 27%) are full-time employees ing problem.
assigned to overtime (more than 40 hours per
week) schedules. 2.1. Model parameters
In the US, at least, there are few regulatory
N The number of different feasible employee
impediments to the use of overtime scheduling
schedules, indexed j = 1..... N.
policies. The US Fair Labor Standards Act,
Cj The weekly wages (including overtime pay)
which covers about 80 million employees, places
for one employee assigned to schedule j.
no ceiling on the number of hours that adult
The fixed or per capita labor expense for an
employees can be required to work [24]. US
employee assigned to schedule j.
labor laws generally regard the establishment of
T The number of periods in the planning
employee work schedules as a basic managerial
horizon, indexed t = 1..... T.
prerogative. As long as they are not excessive or
R, The minimum labor needed for period t to
unsafe, and advance notice is given, employees
provide satisfactory service, for t = 1.... T
must generally accept overtime schedules or risk
atj [1 if period t is a work period in tour j, 0
discharge for insubordination [25]. While many
otherwise], indexed t = 1..... T; j = 1..... N.
collective bargaining agreements between em-
ployers and employees provide very specific 2.2. Decision variables
guidelines about overtime pay and overtime
allocation, the public and private sector Xj The number of employees assigned to
contracts that we reviewed did not impose schedule j, defined for j = 1..... N.
specific limits on the maximum amount of
overtime employees could be required to work 2.3. Problem statement
each week. A plausible explanation is that
employers are reluctant to surrender the staffing N

flexibility provided by federal law, and em- Minimize Z = ~ (Cj + ~ ) ~ (4)


j=l
ployees are reluctant to surrender opportunities
to substantially increase their income.
N
No matter what kind of contractual relation-
Subject to: ~ a~j~ > R,,Vt (5)
ship exists between a service firm and its j=l
workers, however, the firm's labor resources
must still be deployed (scheduled) overtime to ~ {0,1,2,...),Vj (6)
satisfy temporal service demand. Although
there are many ways to characterize the staffing The objective (equation (4)) is to minimize total
and scheduling goals of an organization, and labor expense, or the wages and per capita
many different ways to measure the effectiveness expenses for each tour multiplied by the number
Omega, Vol. 25, No. 3 289

Table 3. Labor scheduling constraints [33]


Constraint category Example Definition
Shift constraints Shift start time flexibility The set of allowable times at which shifts normally begin.
Break placement flexibility The allowable range of times, from shift start, during which a break must be initiated.
Shift length flexibility The range of allowable shift lengths.
Days-off constraints Number of off-days The minimum and maximum number of rest days per week.
Timing of off-days Whether rest days must be consecutive, on weekends, etc.
Tour constraints Shift start time float The allowable range between the earliest and latest shift start times in a particular tour.
Break float The maximum allowable difference between the earliest and latest scheduled break times
(from shift start), for a particular tour.
Shift length float The maximum allowable difference between the length of the shifts in a particular tour.
Tour length flexibility The allowable range of scheduled work periods in a tour.

of employee assigned to that tour. Constraints widely studied. In some cases the effects are
(5) insure that enough employees are scheduled straightforward and can be assessed analytically
for work during period t to satisfy the minimum [37]. Jacobs and Bechtold [33] studied more
staffing requirements for that period. Con- complex operating environments and found
straints (6) are the integrality requirements. very little difference in schedule performance
Other performance measures, such as labor when days-off were required to be consecutive,
utilization or schedule efficiency (required labor rather than an arbitrary pair. However,
E,R, + scheduled labor E,E~a,iXj), workforce schedule performance often improves when
size, and the amount of scheduled overtime, can employees are allowed to work fewer than 5
be easily extracted from a solution to this days per week and/or 8 hours per day
problem. [13, 14, 33].
In general, better solutions can be achieved While these studies have greatly increased our
when there are few temporal constraints on the understanding of the operational advantages of
deployment of labor resources [14]. These part-time scheduling flexibility, they tell us very
constraints tend to fall into three broad little about the effectiveness of overtime
categories [28]: shift constraints (the set of scheduling policies (e.g., employee schedules
allowable shift start times, the range of with more than 8 hours per shift or more than
allowable shift lengths, and break timing); 40 hours per tour). In their aggregate planning
day-off constraints (the number and timing of study, Hancock et al. [38] demonstrated that
rest days); and tour constraints (shift start time planned overtime work in health care services
float, shift length float, and break float). More could reduce labor costs and improve pro-
complete definitions for these scheduling con- ductivity. However, aggregate plans must
straints can be found in Jacobs and Bechtold usually be disaggregated to determine the
[33]. A brief summary of each type of constraint precise timing and quantity of overtime usage
appears in Table 3. [39]. Since overtime labor must be drawn from
Several recent studies have explored the the pool of off-duty employees, and may be
relationship between shift scheduling con- subject to other scheduling constraints, aggre-
straints and scheduling performance measures gate plans may overstate the attainable benefits
(labor expense, workforce size, labor utilization, of planned overtime use.
or other indices of the quality of the match Tour scheduling models are based on explicit
between scheduled and required capacity). 'legal' employee schedules, and thereby avoid
Many researchers [28, 29, 33, 36] have found such limitations. While these models can
that flexibility in the timing of meal breaks, or provide more precise estimates of the benefits of
'break flexibility,' yields significant savings. alternative staffing and scheduling policies, the
Schedule performance also tends to improve number of feasible employee schedules that the
with increasing cardinality in the set of model must consider tends to increase rapidly
allowable shift start times [21,33]. Better when scheduling constraints are relaxed. In
performance is also likely when the minimum Henderson and Berry's [32] research with
allowable shift length is reduced to less than 8 telephone operator schedules, for example, meal
hours per day [13, 14, 26, 27, 33]. break windows increased the number of integer
The relationship between schedule perform- variables in their mathematical program for a
ance and days-off constraints has also been labor scheduling problem by a factor of fifty
290 Easton, R o s s i n - - O v e r t i m e Schedules

[33]. Scheduling flexibility options like overtime ant operational performance measures such as
have similar effects. To illustrate, consider a labor utilization and payroll expense. However,
convenient source of overtime labor: those employment statistics indicate that service
employees who have just completed their sector firms are steadily replacing part-time
'regular shift'. They're already on-site and labor with overtime work. While this change
therefore incur minimal additional costs (make- may be motivated in part by rising per capita
ready time, transportation to the job site) to costs, the prospect of productivity gains [38]
extend their shifts. Suppose overtime work is offers another compelling reason for the growth
restricted to those employees just finishing their of overtime work.
regular shifts. Further suppose we wish to Of course, the benefits of overtime scheduling
restrict the total amount of planned overtime an flexibility may depend on the operating environ-
employee is allowed to work in a given time ment where it is employed. To illustrate,
period, say the next N shifts, to at most Y hours. consider the simple shift scheduling examples
The number of different ways to allocate exactly depicted in Table 4. Assuming each full-time
Y hours of overtime work to the end of up to employee works an 8 hour shift (with a
N different shifts is the number of different scheduled 1 hour rest break after the fourth
compositions of the integer Y into N parts [40], hour), at least two employees are needed to staff
or: the system when it operates for 12 hours per
day. Note that even with up to 3 hours of
[ ( Y + N - - 1)!]/[Y!(N-- 1)!] (7)
overtime appended to the end of Shift 1
For example, there could be as many as (covering all periods except hour 5), at least two
1 + 5 + 15 ÷ 35 + 70 = 126 different ways to full-time employees are still needed to staff the
append 1 < Y < 4 hours of overtime work to system. With a slightly longer operating day,
the end of the N = 5 'regular' shifts in a single however, overtime scheduling policies can
standard 5-day schedule. provide significant savings. The second half of
The combinatorial growth in the number of Table 4 illustrates a system that operates for 16
feasible schedules makes the precise comparison hours per day. In this case, at least three
of alternative scheduling flexibility options like different employees to cover all 16 periods if
overtime difficult. However, several techniques standard 8 hour shifts are used. However, by
to improve the tractability of large-scale labor appending 3 hours of overtime to the end of
scheduling problems have been proposed. An Shift 5, the 16 hour per day system can be
excellent review of these procedures can be staffed with a minimum of two employees.
found in Jarrah et al. [34]. Based on these examples and current US
employment statistics, we speculate that over-
time scheduling flexibility has the potential to
3. EVALUATING OVERTIME SCHEDULING
FLEXIBILITY favorably affect total labor costs, workforce
size, and schedule efficiency (total labor
Previous studies have confirmed that part- r e q u i r e m e n t s - total scheduled hours), com-
time scheduling flexibility helps improve import- pared with traditional, 5 day, 40 hours per week

T a b l e 4. Effects o f o v e r t i m e shifts o n p r o d u c t i v i t y
12 o p e r a t i n g h o u r s / d a y
Standard 8-hour shifts 1 2 3 4 5 6 7 8 9 10 11 12
Shift 1 1 1 1 1 0 1 1 1 1
Shift 4 1 1 1 1 0 1 1 1 1
Overtime shifts
Shift 1 + OT 1 1 1 1 0 1 1 1 1 O1 02 03
Shift 4 1 1 1 1 0 1 1 1 1
16 o p e r a t i n g h r s / d a y
Standard 8-hour shifts 1 2 3 4 5 6 7 8 9 10 ll 12 13 14 15 16
Shift 1 1 1 1 1 0 1 1 1 1
Shift 5 1 I 1 1 0 1 1 I 1
Shift 8 1 1 1 1 0 1 1 l 1
Overtime shifts
Shift 1 1 1 1 1 0 1 1 1 1
Shift 5 + O T I 1 1 1 0 1 1 1 1 O1 02 03
Omega, Vol. 25, No. 3 291

Table 5. Experimental design


Factors Levels
Schedule policy .factors
Maximum overtime hours/shift 0, 2, or 4 h/shift
Maximum overtime hours/tour 0, 2, 4, 6 ..... or 20 h/tour
Operating environment .factors
Length of operating day 12, i6, or 20 h/day
Within-week demand pattern Level, trend + , concave, convex, sinusoidal, or bimodal,
Within-day demand pattern Level, trend + , concave, convex, sinusoidal, or bimodal,
Per capita expense 0% or 100% of base wage

staffing and scheduling policies. Depending on To avoid deploying overtime in a manner that
the operating environment, we expect to find might be confused with other kinds of labor
better scheduling performance when shift scheduling flexibility, we focused our investi-
lengths are allowed to exceed 8 hours (at gation on a fairly narrow range of overtime
premium pay, of course), and as tours are scheduling options. Specifically, we assumed
allowed to increase beyond the traditional 40 that scheduled overtime only occurs at the end
working hours per week. We also expect to find of a standard 9 hour (including meal break)
that overtime scheduling policies are more shift. Employees who have just completed their
effective in environments with longer operating 'regular shift' are a convenient source of
days, since there are more opportunities to use overtime labor, since they're already on-site and
overtime. Finally, we expect to find that higher therefore incur minimal additional costs (make-
per capita labor expense rates increase the ready time, transportation to the job site) to
attractiveness of overtime schedules. extend their shifts. We did not evaluate other
To test these hypotheses, we devised a set of reasonable options for scheduling overtime,
tour scheduling experiments for hypothetical such as starting work earlier than normal or
service firms. Our goal was to compare various working during off days. Scheduling overtime
performance measures for alternative overtime work during an employee's regular day off is
scheduling policies with those attained with a very similar to part-time scheduling flexibility,
standard 40 hours per week scheduling policy especially when per capita labor expenses are
(our base case). The experimental factors and low. Similarly, scheduling overtime before an
levels that we used in our study are summarized employee's normal shift starting time may
in Table 5. F o r the base case, we assumed that produce effects that are more easily attained
employees work 9 hour shifts on each of 5 with break placement flexibility.
consecutive days, with an unpaid 1 hour meal We defined overtime in terms of hours per
break in the middle of each shift. Shifts can start shift and hours per tour, and therefore devised
at any hour of the day provided enough time two types of interacting scheduling policy
remains to complete the shift by the end of the constraints to limit the amount of scheduled
operating day. However, the shifts within a overtime in a particular tour. First, we
particular tour must start at the same time each constrained the m a x i m u m allowable amount of
day. Similar full-time work rules have been overtime per employee per shift to 0, _< 2, or
assumed in m a n y previous labor scheduling < 4 hours. Second, we limited the total number
studies [13, 29, 31, 33]. of overtime hours per employee per week to 0,
We used the Fair Labor Standards Act and _< 2, _< 4 ..... or _< 20 hours. These constraints
the W a l s h - H e a l y Act, which require overtime yielded a total of 16 different overtime policies:
pay for work in excess or 40 hours per week or one for zero overtime, five for the case when
8 hours per day, respectively, to define overtime overtime is limited to a maximum of 2 hours per
work. We assumed that overtime work is shift, and 10 for the case when overtime is
remunerated at 1.5 times the standard hourly limited to a m a x i m u m of 4 hours per shift.
pay rate. Although tour scheduling models Theoretically, the most aggressive overtime
can easily accommodate productivity differ- policy (up to 4 hours per day, up to 20 hours per
ences [6, 31] we assumed that employee pro- tour) could provide up to 33% of the total
ductivity does not decay over the course of a scheduled labor.
shift. We also examined three different operating

OME 25/3--C
292 Easton, Rossin--Overtime Schedules

environments, assuming the service firm would would be active in the optimal continuous
operate for 12, 16, or 20 hours per day, 7 days solution for each problem (a sufficient working
per week. To simulate the demand character- subset for the linear program) by using IBM's
istics for a variety of service industries, we chose Optimization Subroutine Library [42] and a
a set of well-known synthetic demand patterns custom dynamic program as solvers. In this
[29, 41]. We combined one of six different stage of our heuristic procedure, the tour
day-of-the week demand patterns (level, increas- scheduling problem is initially formulated with
ing trend (trend + ), concave, convex, sine, and a small working subset of tours and solved as a
bimodal) with one of six time-of-day demand linear program. The optimal dual prices (E) for
patterns (level, trend + , concave, convex, sine, the current formulation are passed to the
and bimodal), forming a total of 36 different dynamic program, which implicitly evaluates
demand patterns for each operating environ- the reduced costs (i.e. Q-Z,r=~a,jY,) for all
ment. As in Bechtold and Showalter [29], we other feasible tours. The tour with the smallest
assumed that the average hourly demand for negative reduced cost is appended to the
each of the 36 different service demand working subset to create a new linear program.
distributions was 50 employees, with a maxi- The process then repeats until the standard
mum amplitude of 16 employees. simplex optimality criterion is satisfied (i.e. all
Finally, we evaluated overtime scheduling remaining tours have reduced costs > 0).
effectiveness with per capita labor expense fixed We used a dynamic program adapted from a
at either 0 or 100% of the standard weekly procedure described in Easton and Rossin [31]
wage. Result (3) predicts that overtime work to determine the reduced costs for feasible tours
should be avoided when the prorated hourly per not yet included in the working subset. The
capita expense is less than the hourly overtime procedure efficiently constructs and evaluates
premium. However, as shown in the second complete tours one shift at a time. It begins by
example in Table 4, overtime also improves generating the set of all feasible shifts (indexed
schedule efficiency (i.e. total labor require- by day of week, shift start time, and shift
ments - total scheduled hours). Therefore, length). For each shift, the program computes a
overtime schedules could prove beneficial even partial reduced cost, or the hourly labor expense
when the per capita labor expense is low. On the for the shift (wages plus 1/5 of the per capita
other hand, result (3) suggests that overtime cost of a tour) less the dual price for each work
work would become very attractive to employ- period in the shift. Next, it constructs the set of
ers when the overtime premium is less than the partial tours with a single shift. However, only
prorated hourly per capita expense. With a per those partial tours with negative partial reduced
capita labor expense of 100% of the standard costs need to be retained (see Theorem 1 in the
weekly wage, the prorated hourly per capita Appendix), reducing state space and compu-
expense is exactly double the hourly premium tational effort. At subsequent stages, single
for overtime work. At this level, the ideal shifts are appended to the partial tours
amount of scheduled overtime work should be generated during the previous stage to produce
far greater than the amount used when per feasible partial tours with one more shift. Again,
capita expenses are comparatively low. only the feasible partial tours with negative
In all, we generated and solved a total of 3 partial reduced costs are retained. The dynamic
operating environments x 16 overtime policies program continues in this manner until either
x 2 per capita expense levels x 36 demand (1) at some stage the set of feasible partial tours
distributions = 3456 test problems. Our as- all have reduced costs of zero or more, or (2) the
sumed work rules and operating environments final stage is reached and one or more complete
resulted in test problems with between 28 and tours with negative reduced costs have been
184,100 unique employee schedules (or tours, formed. If the former occurs, the simplex
each represented by a separate integer variable). optimality criterion is satisfied and the pro-
With no practical method to obtain exact cedure terminates. If the latter occurs, the tour
solutions for the larger problems, we adopted a with the smallest reduced cost is added to the
heuristic solution strategy based on column current L.P. formulation and the new linear
generation [31] and simulated annealing [30]. program is solved again.
We first identified the set of feasible tours that Finally, after forming the sufficient working
Omega, Vol. 25, No. 3 293

subset for the linear program, we solved the Table 6, we present various performance
problem as an integer program using Brusco statistics for alternate overtime policies when
and Jacobs' [30] simulated annealing heuristic. the per capita labor expense is zero. Although in
In earlier studies, this procedure was shown to theory low per capita expense rates should result
produce optimal or near optimal results with in minimal overtime work, every overtime
similar kinds of problems. All of the computer scheduling policy we studied outperformed, on
programs were coded in F O R T R A N and average, the standard, 40 hours per week
implemented on an IBM RS/6000 Model 570 schedules. This demonstrates that overtime
computer. For the largest test problems (20 scheduling policies provide a cost-effective
operating hours per day, with tours that allow means for reducing some of the overstaffing that
up to 4 hours overtime per shift and up to 60 occurs when all employees must work standard
hours of work per tour), our solution procedure 40 hours per week schedules. Overtime sched-
required an average of 3 minutes of CPU time ules generally resulted in lower total labor
to produce a sufficient working subset. The expense, a smaller workforce, and greater
simulated annealing heuristic was allowed to schedule efficiency than standard full-time
run for 30 seconds to produce an approximate schedules.
integer solution. The annealing heuristic pro- As expected, the advantages of overtime
duced feasible integer solutions that were, on staffing and scheduling policies were more
average, about 1.17 % greater than the pronounced in environments with longer
corresponding optimal L.P. solutions; quite business days. Under the most aggressive
similar to the performance reported by Brusco overtime policy (up to 4 hours of overtime
and Jacobs [30] allowed per shift, with tours of up to 60 hours
Our test results are summarized below. In per week), labor costs were 95.3, 86.6, and

Table 6. Relative schedule performance for overtime policies: per capita expense = 0.0
Maximum allowable overtime per shift
Up to 2 hours per shift Up to 4 hours per shift
Operating Maximum Average Average Average Average Average Average
hours per day hours per tour total cost workforce schedule total cost workforce schedule
size efficiency size efficiency
12 40 100.0% 100.0% 69.9% 100.0% 100.0% 69.9%
12 42 98.4 97.8 71.2 98.3 97.7 71.3
12 44 97.3 96.2 72.2 97.1 95.6 72.4
12 46 97.0 95.8 72.4 95.4 92.1 74.1
12 48 97.1 95.8 72.4 95.3 91.9 74.1
12 50 97.0 95.8 72.5 95.2 91.8 74.2
12 52 95.4 92.0 74.1
12 54 95.4 92.0 74.1
12 56 95.3 92.0 74.1
12 58 95.3 92.0 74.1
12 60 95.3 92.0 74.1
16 40 100.0% 100.0% 64.9% 100.0% 100.0% 64.9%
16 42 99.2 98.9 65.5 99.2 98.9 65.5
16 44 98.8 98.2 65.8 97.8 95.5 67.0
16 46 98.6 97.9 66.0 93.8 86.6 71.0
16 48 98.6 97.9 66.0 93.4 85.5 71.5
16 50 98.5 97.8 66.0 91.0 79.8 74.4
16 52 88.4 73.8 77.7
16 54 88.3 73.4 77.9
16 56 86.6 69.6 80.2
16 58 86.5 69.3 80.3
16 60 86.6 69.5 80.2
20 40 t00.0% 100.0% 62.2% 100.0% 100.0% 62.2%
20 42 94.3 92.8 66.3 94.3 92.8 66.3
20 44 90.4 86.9 69.7 88.8 83.1 71.5
20 46 88.7 81.2 72.1 86.2 78.2 74.4
20 48 88.2 79.2 73.0 86.0 77.1 74.9
20 50 87.7 77.5 73.7 85.4 76.3 75.5
20 52 85.5 76.4 75.5
20 54 85.5 76.4 75.4
20 56 85.5 76.4 75.5
20 58 85.4 76.3 75.5
20 60 85.4 76.3 75.5
Note: Each cell represents the average value for 36 demand distributions.
294 Easton, Rossin--Overtime Schedules
18%

16%

14%

12%

~ ~o%

6%

4%

2%

0%

Demand Pattern (within week/within day)


Fig. 1. Labor expense reduction with aggressive overtime policy.

85.4% of the base case for the 12, 16, and 20 For the 16 hour per day environment, one of
hours per day operating environments, respect- the standard 8 hour shifts can be eliminated
ively. Longer business days increase the number with 3 hours of strategically placed overtime
of possible shifts and provide more opportuni- (see Table 4). Since full-time employees work
ties for scheduled overtime. In addition, a five shifts per tour, this effect is not fully realized
greater amount of overtime can be appended to until the allowable overtime per tour is
the end of many of these shifts (work rules increased to 15 hours. The results in Table 6
permitting, of course) when the firm extends its confirm the existence of this phenomena. Like
operating hours. the 12 hour per day and 20 hour per day
Scheduling performance generally improved environments, schedule performance in the 16
with the amount of overtime allowed per shift hour per day environment improves steadily as
and per tour. As expected, the most aggressive maximum tour length is increased from 40 to 46
overtime policies produced the greatest savings. hours. After plateauing briefly, however, sched-
In all three environments, policies that allowed ule performance continues to improve until
up to 4 hours of overtime per shift outper- maximum tour length is increased to 56 hours
formed those that allowed a maximum of 2 per week. However, since extensive overtime
hours per shift. Savings also tended to increase may lead to employee fatigue, absenteeism, and
with greater limits on maximum tour length, accidents, as well as a decline in output quality
although at a decreasing rate. For the 12 hours [43], it may still be prudent to limit overtime to
per day and 20 hours per day operating a maximum of 4 hours per shift and 6 hours per
environments, 95% of the average savings tour. For the 16 hours per day operating
provided under the most aggressive overtime environment, this limit provides about one-half
policy can be achieved with employee schedules of the benefit of the most aggressive overtime
that are limited to a maximum of 6 hours of policy.
overtime per tour and up to 4 hours of overtime On average, the overtime scheduling policies
per shift. we studied significantly outperformed the
Omega, Vol. 25, No. 3 295

standard full-time scheduling policy. However, again, we display the results for the most
overtime scheduling policies seemed to work aggressive overtime policy. Figure 2 reveals that
better with some service demand patterns than on average, most employees were assigned to
they did with others. In Fig. 1, we show the standard, 40 hours per week schedules. With the
savings achieved for each demand pattern with exception of the 16 hours per day operating
our most aggressive overtime policy (employee environment, very few employees were sched-
schedules with up to 4 hours of overtime per uled to work more than 45 hours per week, even
shift and up to 20 hours of overtime per week). under the most aggressive overtime policy. The
This policy was most effective when service large fraction of employees assigned to 55 hour
demand increased steadily during the day per week tours in the 16 hour per day operating
(trend + within day pattern), especially for the environment is probably due to the effect
12 hours per day operating environment. With illustrated in Table 4.
this pattern, some employees who start work Figure 3 shows the distribution of Shift
early in the day can work an extra hour or two lengths for the most aggressive overtime policy.
beyond the end of their normal shift and Once again, most shifts were 8 hours long.
eliminate the need for other full-time workers Where it was needed, employees were rarely
who would otherwise have to be present just to scheduled for more than 3 hours of overtime
meet the end of day demand peak. work per shift. These results suggest that
As Fig. 1 also shows, overtime scheduling effective overtime policies may not require
policies failed to improve schedule performance broad employee participation, and that signifi-
with some of the 12 hours per day demand cant savings are possible when even a few
patterns, particularly those where peak demand employees work overtime schedules. However,
occurs near the meal break for Shift 1 (level and tour scheduling problems frequently have
sinusoidal). In these cases the need to cover the alternate optima [44]. Therefore, it is possible
meal break for Shift 1 makes it difficult to that similar savings could be achieved with
substitute overtime work for full-time em- other allocations of overtime. This existence of
ployees (see Table 4). alternative optima may help employers dis-
In Figs 2 and 3, we show how overtime was tribute overtime work among their employees in
deployed to accomplish these savings. Once a fair and equitable manner.

90% --~

80% --

70% -
L!

60% --~
o

~" 50% - /

~'o 40% - -

30%--/
m

20% --/

10% - - /

0% _L~

40 41
43 44 45

t
55 56 57 58
Hours per Tour
59 6oL °-'~~

F i g . 2. A v e r a g e tour length distribution for aggressive overtime policy.


296 Easton, Rossin--Overtime Schedules

100%-

90% -

80%
/
70%

-~ 60%
/
50%
8 4o~
/
~- 30%
/
20%-

10% - /

0%'~

:g H o u r s
O v e r t i m e per Shift (hrs) Day

Fig. 3. Shift length distribution for aggressive overtime policy.

In Table 7, we compare the average solutions and huge increases in overtime work, unless
for low and high per capita rates. Of course, they are willing to adopt far more aggressive
total labor expense increases dramatically with overtime scheduling policies.
per capita labor expense. However, the most
startling result of this study is that effective 4. FURTHER RESEARCH
overtime policies seem to be relatively insensi-
tive to large changes in per capita labor expense. The prospect of higher per capita labor
Generally, the workforce size and the ratio of expenses now looms over all US employers.
overtime hours to total scheduled hours for a Strategies to limit the impact of these increases,
given policy was about the same for both low such as overtime work for full-time employees,
and high per capita labor expense rates. In most have been widely adopted but are not yet well
cases, the differences were less than 1%. The understood. This research examined the impact
exceptions, marked in bold in Table 7, occur of scheduled overtime following the completion
with the more aggressive overtime policies in the of an employee's standard 8 hour shift. We
16 hours per day operating environment. found that even when per capita expenses are
The various overtime policies that we low, comparatively small amounts of overtime
investigated allowed between 2 and 33% of total can help employers achieve significant savings.
labor to be overtime work. However, Table 7 We suspect, but have not confirmed, that similar
shows that the average amount of scheduled benefits are likely in other operating environ-
overtime was far less than the m a x i m u m ments, including systems that operate around
permitted, even with high per capita expenses. the clock and those with service demand
This suggests that while overtime scheduling patterns other than those examined in our
policies can help improve scheduling perform- study.
ance, employers who already use overtime Of course, there are other ways to form
effectively should not expect increased per capita overtime schedules, such as starting work before
expenses to result in large workforce reductions an employee's normal shift start time or
Omega, Vol. 25, No. 3 297

scheduling workers for duty during their normal demand. In this role, overtime functions as a
off days. Provided employees are willing to 'safety stock' that buffers the system against
accept less pay for fewer scheduled hours per uncertainties in either supply or demand. In this
week, undertime scheduling options (weekly research, we examined overtime's role as a
tours with between 35 and 40 hours) may also substitute for additional full-time staff. Clearly,
prove beneficial. To enhance our understanding if an employee is already working overtime, the
of the effects of scheduling flexibility, these organization must look to other workers to
scheduling options should be explored. fulfill the safety stock role. In effect, the use of
An implicit assumption of our experiments a planned overtime strategy reduces the
was that the pattern of service demand would maximum size of the firm's emergency reserves.
remain the same week after week. With We are unaware of any study that addresses
pronounced demand seasonality, the ideal the question of an appropriate size for the
workforce size may change over time, resulting unplanned overtime buffer, so this would be an
in additional costs for recruiting, training, and interesting avenue for future research. Since
severance pay. Our model does not account for planned and unplanned overtime strategies
these costs. Further research is needed to might compete for the same resources (in our
understand the implications of seasonal and case, overtime potential, or the labor from
cyclical demand variations on staffing and employees who have just completed their
scheduling policies. regular shift), it may be necessary to evaluate
Service managers often deploy overtime on an the opportunity costs for alternative uses of
ad hoc basis in response to unforeseen events potential overtime labor.
such as absenteeism or higher than expected Finally, most previous labor scheduling

Table 7. Performance comparison for low and high per capita labor expense
Up to 2 hours overtime per shift Up to 4 hours overtime per shift
100% Per capita
0% Per capita expense expense 0% Per capita expense 100 Per capita expense
Maximum Mean Mean Mean Mean Mean Mean
Operating hours percent OT workforce percent workforce percent OT workforce Workforce
hours per day per tour Labor size OT labor size labor size Percent OT size
12 40 NA 152.8 NA 152.5 NA 152.8 NA 152.8
12 42 0.37% 149.5 0.39% 149.4 0.39% 149.4 0.39% 149.3
12 44 0.77% 147.0 0.73% 147.2 1.01% 146.1 1.07% 145.8
12 46 0.81% 146.4 0.80% 146.4 2.35% 140.7 2.36% 140.6
12 48 0.85% 146.5 0.82% 146.5 2.38% 140.5 2.38% 140.4
12 50 0.82% 146.4 0.83% 146.3 2.46% 140.3 2.40% 140.3
12 52 2.36% 140.6 2.42% 140.3
12 54 2.36% 140.6 2.42% 140.4
12 56 2.33% 140.7 2.37% 140.4
12 58 2.33% 140.7 2.37% 140.4
12 60 2.33% 140.7 2.37% 140.4
16 40 NA 216.7 NA 216.7 NA 216.7 NA 216.7
16 42 0.23% 214.2 0.24% 214.1 0.22% 214.3 0.21% 214.4
16 44 0.40% 212.8 0.41% 212.9 1.57% 207.0 1.64% 206.8
16 46 0.47% 212.1 0.42% 212.3 5.27% 187.6 5.41% 186.9
16 48 0.47% 212.1 0.45% 212.2 5.77% 185.3 6.11% 184.9
16 50 0.47% 212.0 0.46% 212.1 8.51% 172.9 8.60% 172.8
16 52 11.63% 160.0 11.96% 159.3
16 54 11.93% 159.0 12.04% 158.9
16 56 14.00% 150.8 15.34% 147.6
16 58 14.15% 150.2 15.38% 147.4
16 60 14.13% 150.2 15.39% 147.6
20 40 NA 282.7 NA 282.7 NA 282.7 NA 282.7
20 42 1,06% 262.4 1.17% 261.2 1.06% 262.4 1.17% 261.3
20 44 2,65% 245.6 4.40% 239.2 4.40% 234.9 4.61% 234.5
20 46 5,83% 229.5 7.66% 223.6 6,41% 221.0 6.60% 220.1
20 48 7,03% 223.9 8.33% 220.7 7.11% 218.0 7.64% 216.6
20 50 8.04% 219,2 8.14% 218.8 7.36% 215.8 7.35% 215.6
20 52 7.34% 215.9 7:35% 215.4
20 54 7.34% 216.1 7.30% 216.0
20 56 7.33% 216.0 7.35% 215.6
20 58 7.38% 215.6 7.30% 216.0
20 60 7.36% 215.7 7.33% 215.7
Note: Each cell represents the average value for 36 demand distributions.
298 Easton, Rossin--Overtime Schedules

research has ignored the effects o f d e m a n d below, if a partial tour with k shifts and negative
m a n a g e m e n t on labor staffing and scheduling partial reduced costs exists, there is a proper
decisions. However, practitioners frequently subset o f this partial tour, with k - 1 shifts, that
utilize d e m a n d m a n a g e m e n t techniques to also has negative partial reduced costs.
attenuate or stimulate demand, thereby improv- T h e o r e m 1: Let Tk = {$1, S2,...Sk} be a feasible
ing the utilization o f their capacity. If service partial tour with k different shifts. Let
d e m a n d is elastic, marketing efforts could C(Tk) = Zsj~r~C(S/) be the partial reduced cost
mitigate the need for extraordinary degrees o f for this partial tour, where 6'(St) represents the
scheduling flexibility. We encourage future labor expense for shift S~ (wages and the per
researchers to attempt to link these two problem shift per capita expense), less the dual prices for
domains. the w o r k periods in shift Sj If C(Tk) < 0, there
must be at least one proper subset o f partial tour
ACKNOWLEDGEMENTS Tk, the feasible partial tour Tk_l = T k - Si,
Partial support for this research was furnished by the Sje{T~}, such that C ( T k _ I ) < 0 . Therefore,
Robert H. Brethen Operations Management Institute at partial tours with partial reduced costs o f zero
Syracuse University. The authors gratefully acknowledge
the contribution of two anonymous referees, who provided or more m a y be ignored.
suggestions to improve an earlier version of this paper. Proof." We shall assume the converse. That is,
C(Tk) < 0 but C'(Tk._ 1) _-_ 0 for any Sje{Tk}. F o r
APPENDIX both assumptions to hold, there must be at
least one Sj~{Tk} such that C(Tk_ 1) + C(Sj) < 0,
COLUMN GENERATION DYNAMIC PROGRAM where Sj~{Tk 1}. Therefore, C(Sfl < 0_<
Overtime scheduling options greatly increase C(Tk-I). Since we assumed that C(Tk 1)--> 0,
the n u m b e r o f feasible tours and therefore, the there must also be at least one shift S,e Tk_ ~such
n u m b e r o f integer variables in tour scheduling that C(Tk _ ~) _> 0. By replacing shift S, in partial
models. In an optimal solution to such a tour Tk_ 1 with shift St, we form a new partial
problem, only a few o f these tours are likely to tour with k - 1 shifts. The reduced cost for this
be active (i.e. ~ > 0). Using column generation, tour will be C'(Tk_ 1) + C(Sj) -- C'(Sj), which
we construct a working subset o f feasible tours must be less than zero. This contradicts the
that includes all active tours f r o m the optimal premise and completes the proof.
solution for the closely related tour scheduling
linear program. Starting f r o m an initial working REFERENCES
subset that includes e n o u g h tours to allow a 1. Zeithaml, V., Parasuraman, A. and Berry, L., Delivering
feasible solution, new tours that violate the Quality Service: Balancing Customer Perceptions and
Expectations. The Free Press, New York, 1990.
simplex optimality criterion (i.e. tours with 2. Buzzell, R. and Gale, B., The PIMS Principles. The
negative reduced costs) are added to the Free Press, New York, 1987, p. 107.
working subset [23]. F o r some overtime policies, 3. Heskett, J., Sasser, E. and Hart, C, Service Break-
throughs: Changing the Rules of the Game. Free Press,
the reduced costs for hundreds o f thousands o f New York, 1990.
tours might have to be c o m p u t e d at each 4. Andrews, B. and Parsons, H., L.L. Bean chooses a
iteration. We use a dynamic p r o g r a m to telephone agent scheduling system. Interfaces, 1989, 19,
1-9.
efficiently evaluate prospective new tours for the 5. Holloran, T. and Byrn, J., United Airlines Station
working subset. At each iteration, dual prices manpower planning system. Interfaces, 1986, 16, 39-50.
obtained f r o m the current linear p r o g r a m m i n g 6. Li, C., Robinson, E. and Mabert, V., An evaluation of
tour scheduling heuristics with differences in employee
solution are passed to the dynamic program. It productivity and cost. Decision Sciences, 1991, 22,
then constructs and prices all feasible shifts, and 70(~718.
uses them to assemble feasible partial tours with 7. Taylor, P. and Huxley, S., A break from tradition for
the San Francisco police: patrol officer scheduling using
first one shift, then two shifts, and so on. A an optimization-based decision support system. Inter-
feasible partial tour is an incompletely specified faces, 1989, 19, 4-24.
employee schedule that could, with the addition 8. Chase, R., The customer contact approach to services:
theoretical bases and practical extensions. Operations
o f m o r e shifts, form a tour that satisfies all work Research, 1981, 29, 698-706.
rules. To reduce state space and improve 9. Thompson, G., Improving the utilization of front-line
c o m p u t a t i o n a l efficiency, the dynamic p r o g r a m service delivery personnel. Decision Sciences, 1992, 23,
1072-1098.
ignores partial tours (states) with partial 10. Smith, S., The growing diversity of work schedules.
reduced costs of zero or more. As we show Monthly Labor Review, 1986, 109, 7-13.
Omega, Vol. 25, No. 3 299

11. Mellor, E., Shift work and flexitime: How prevalent are 29. Bechtold, S. and Showalter, M., A methodology for
they? Monthly Labor Review, 1986, 109, 14-21. labor staffing in a service operating system. Decision
12. Lovelock, C., Strategies for managing capacity-con- Sciences, 1987, 18, 89-107.
strained service organizations. Service Industries Jour- 30. Brusco, M. and Jacobs, L., A simulated annealing
nal, 1984, 4, 12 30. approach to the solution of flexible labour scheduling
13. Showalter, M. and Mabert, V , An evaluation of problems. Journal of the Operational Research Society,
full-/part-time tour scheduling methodology. Inter- 1993, 44, 1191 1200.
national Journal of Operations and Production Manage- 31. Easton, F. and Rossin, D., Sufficient working subsets
ment, 1989, 8, 54-71. for the tour scheduling problem. Management Science,
14. Mabert, V. and Showalter, M., Measuring the impact of 1991, 37, 1441-1451.
part-time workers in service organizations. Journal of 32. Henderson, W. and Berry, W., Heuristic methods
Operations Management, 1990, 9, 209-229. for telephone operator shift scheduling: an experimental
15. Ehrenberg, R., Fringe Benefits and Overtime Behavior. analysis. Management Science, 1976, 22, 1372-1380.
Lexington Books, Lexington, MA, 1989, pp. 51-54. 33. Jacobs, L. and Bechtold, S., Labor utilization effects of
16. Murthy, B. and Murrmann, S., Employee leasing: an labor scheduling flexibility alternatives in a tour
alternative staffing strategy. The Cornell HRA Quar- scheduling environment. Decision Sciences, 1993, 24,
terly, (June), 1993, pp. 18-23. 148-166.
17. US Census Department, Statistical Abstract of the 34. Jarrah, A., Bard, J. and deSilva, A., Solving large-scale
United States. Washington DC Government Printing tour scheduling problems. Management Science, 1994,
Office, 1993. 40, 1 1 2 4 - 1 1 4 4 .
18. US Department of Labor, Bureau of Labor Statistics, 35. Malhotra, M. and Ritzman, L., Scheduling flexibility in
Table A29: Persons at work in non-agricultural the service sector: a postal case study. Production and
industries by class of worker on f u l ~ or part-time Operations Management, 1994, 3, 100 117.
status. Employment and Earnings. Washington DC 36. Thompson, G., Shift scheduling in services when
Government Printing Office, April, 1993. employees have limited availability: an L.P. approach.
19. Kao, E. and Tung, G., Aggregate nursing requirements Journal of Operations Management, 1990, 9, 352-370.
planning in a public health care delivery system. 37. Burns, R. and Koop, G., A modular approach to
Socio-Economic Planning Sciences, 1981, 15, 119-127. optimal multiple-shift manpower scheduling. Oper-
20. Kao, E. and Queyranne, M., Budgeting costs of nursing ations Research, 1987, 35, 100-110.
in a hospital. Management Science, 1985, 31, 608-621. 38. Hancock, W., Pollock, S. and Kim, M., A model to
21. Mabert, V. and Watts, C., A simulation analysis of determine staff levels, cost, and productivity of hospital
tour-shift construction procedures. Management Sci- units. Journal of Medical Systems, 1987, 11, 319-330.
ence, 1982, 28, 520-532. 39. Ritzman, L., Krajewski, L. and Showalter, M., The
22. Svatko, J., Contingent employment. Small Business disaggregation of aggregate manpower plans. Manage-
Reports, (November), 1988, p. 53. ment Science, 1976, 22, 1204-1214.
23. Cohen, A. and Gadon, H., Alternative Work Schedules: 40. Nijenhuis, A. and Wilf, H., Combinatorial Algorithms.
Integrating Individual and Organizational Needs. Ad- New York, Academic Press, 1978.
dison-Wesley, Reading, MA, 1978. 41. Brusco, M. and Johns, T., The effect of demand
24. U.S. Department of Labor, Employment Standards characteristics on labour scheduling methods. Inter-
Administration, Fact Sheet No. ESA 91-7 Federal national Journal of Operations and Production manage-
Minimum Wage And Overtime Pay Standards. Wash- ment, 1995, 15, 74-88.
ington D.C. Government Printing Office, 1991. 42. IBM Corporation, Optimization Subroutine Library:
25. Fortado, B., Travis, T. and Jennings, K., Refusal to Guide and Reference (Release 2), 1991.
accept a work assignment: how arbitrators rule in 43. Thomas, R., Effects of scheduled overtime on labor
discharge cases. Employee Relations L. J., 1990, 16, productivity. Journal of Construction Engineering and
205-216. Management., 1992, 118, 60-76.
26. Bailey, J., Integrated Days off and shift personnel 44. Easton, F. and Rossin, D., Equivalent alternate
scheduling. Computers and Industrial Engineering, 1985, solutions for the tour scheduling problem. Decision
9, 395~404. Sciences, 1991, 22, 985 1007.
27. Bailey, J. and Field, J., Personnel scheduling with
flexshift models. Journal of Operations Management,
1985, 5, 327-338. ADDRESS FOR CORRESPONDENCE: Fred F Easton, Robert H
28. Bechtold, S. and Jacobs, L., Implicit modeling of Brethen, Operations Management Institute, School of
flexible break assignments in optimal shift scheduling. Management, Syracuse University, Syracuse, N Y 13244-
Management Science, 1990, 36, 1339-1351. 2130, USA.

You might also like