Professional Documents
Culture Documents
B.G. Raghavendra
Indian Institute of Science,
Bangalore-560012, India
Abstract: The chronological age of members in an organisation has been found
to be one of the determinants of their effectiveness, efficiency and
performance, and organisational planners find it advantageous to monitor the
age distributions in the hierarchical grades of the organisation, to maintain
homogeneity and proper flow of personnel up the hierarchy. In this paper,
methods have been devised to maintain specified age distributions for
hierarchical systems in general, and also for a class of hierarchical
organisations which follow proportionality policies. Proportionality policies
are those that restrict the recruitment to every level of the hierarchy to be in
proportion to the promotions into that level, and are of relevance to
organisations which outsource a part of their work, the outsourced workforce
being notionally viewed as recruits to the system, and also to those which
ostensibly aim to protect the career interests of their existing employees
through such policies. The required or specified age characteristics are
maintained by selecting recruits of the appropriate ages to the system. The
theoretical analyses have been illustrated with numerical examples, and
validated with real-world data. The results extend in an identical and simple
manner to aggregate tenure distributions also.
Keywords: manpower planning; Markov models; manpower systems; age
distributions.
Reference to this paper should be made as follows: Nilakantan, K. and
Raghavendra, B.G. (2011) Determination of recruitment age in Markov
manpower systems, Int. J. Applied Management Science, Vol. 3, No. 1,
pp.7296.
Biographical notes: K. Nilakantan holds a Bachelors degree in Chemical
Engineering from the Indian Institute of Technology, Madras, India, a
PhD from the Indian Institute of Science, Blore, India, and a Masters in
Mathematics from the University of South Florida. He was with the National
Institute of Industrial Engineering, Mumbai, India for a large part of his
teaching career. He has published papers in many reputed international
journals. His research interests are in the applications of dynamic optimisation
in industrial, business, environmental and biotech systems.
Copyright 2011 Inderscience Enterprises Ltd.
73
Introduction
74
The approaches to the modelling of manpower systems have been through the use of
mathematical programming and network models (Aronson and Thompson, 1985; Chew
and Khoong, 1994; Glen and Yang, 1996; Grinold, 1978; Khoong, 1999), Monte-Carlo
simulation models (Bazargan Lari et al., 2003; Chattopadhyay and Gupta, 2007;
Standridge, 1979), multi-criteria models (Klingman and Phillips, 1984; Silverman et al.,
1988), and Markov models.
The most widely encountered technique in the mathematical modelling of hierarchical
manpower systems is through the use of Markov models (De Feyter and Guerry, 2009).
Hierarchical manpower systems facilitate their being modelled as Markov systems, and
thereby lend themselves to tractable quantitative analysis as Markov Chains. The
rationale is that although individual members can vary greatly in their performance and
capabilities, however, when taken collectively, the expected aggregate behaviour of the
whole grade can be described by the average behaviour patterns of its individual
members (Bartholomew, 1982). This rationale behind the modelling of hierarchical
manpower systems as Markov models is explained further and justified in Bartholomew
(1982) and Vajda (1978).
The literature on Markov manpower systems is both rich and extensive, spanning a
period of more than 65 years. A comprehensive treatment of the earlier work can be
found in the books by Bartholomew (1982) and Vajda, (1978). A number of other aspects
of hierarchical Markov manpower systems have been analysed by subsequent
researchers, like sensitivity of the system structure, and variance-covariance of the grade
sizes in Vassiliou and Gerontidas (1985), periodic systems in Vassiliou (1986), control of
system structure in Kalamatianiou (1987), Gaimon and Thompson (1984), Poornachandra
Rao (1991), and Zanakis and Maret (1981), multi-characteristic systems in Hayne and
Marshall (1977), and Chandra (1988, 1989), and a bivariate model in Raghavendra
(1991). Other significant aspects covered are in the papers of Gerontidis (1995) on
periodicity in Markov manpower systems and the problem of re-attainability, Georgiou
and Vassiliou (1997) on cost models in non-homogenous Markov manpower systems,
Popova and Morton (1998) on the use of Bayesian prediction with stochastic
programming in stochastic manpower scheduling, Carette (1999) on the use of two-wave
panel data in the development of hierarchical manpower systems models, Tsantas (2001)
on the limiting behaviour of the means and covariances of the grades, and Ugwuowo and
McClean (2000) on modelling heterogeneity in Markov manpower systems. The most
recent papers are on fire-fighter staffing under conditions of temporary absences and
wastage (Fry et. al., 2003), analysis of semi-Markov systems and their related reward
paths (Papadopoulou and Tsaklidas, 2004), and an application of Markov manpower
modelling in the armed forces (Skulj et al., 2008).
Papers pertaining to the length of service (tenure) and age distributions and their use
in manpower planning are those of Glen (1977) on the derivation of the steady-state
tenure distributions and their means and variances, Sirvanci (1984) on forecasting
manpower losses with tenure distributions estimated from past data, and Mukherjee and
Chattopadhyay (1990) on the computation of the maximum ages for promotions, in a
system with fixed structure and pre-specified ages of entry and retirement. Also relevant
are those of Nehra and Khurana (1990) on manpower planning in the Navy, using the
Markov model in conjunction with chronological age distributions, Guerry (1999) on the
use of fuzzy sets in manpower systems through the introduction of the grade of
75
The primary variable of interest is the age [denoted by physiological age (PA)] of a
member in grade j at time T of the system, and is denoted by the variable PAj(T) which
can take integer values of 0,1,2, ..., the unit of measurement of PA being the system
transition period.
Rj (t, x) denotes the number of recruits to grade j in (t, t + 1] with a PA of x
uj(T, x) denotes the number of members in grade j at time T with a PA of x periods, and
plays a central role in the characterisation of age distributions.
U(T, x) is the row vector whose elements are the uj(T, x) defined above, and is expressed
in terms of the fundamental parameters, P and R(), as
k
U ( , x ) =
R ( ) P
l =1
( x al )
(3.1)
76
Under stochastic recruitment ages, the probability mass function of the recruitment age
distribution for grade j (or the proportion of recruits to grade j with PA = w) is another
fundamental parameter of the system, and is denoted by:
pr ( w) j = Pr ob [ (age of recruit to grade j ) = w] , for w = 0,1, 2,...
U ( , x ) =
[ R() * P (s)]P
r
xs
(3.2)
s =0
The quantities (uj (, x) / nj()) are the steady state proportions of members in grade j for
j = 0, 1, 2, ..., k with PA of x, for x = 0, 1, 2, ., and are the PA distributions or
characteristics of the system.
The model extends in an identical way to aggregate tenure (the total tenure over all
organisations that an employee has been in) distributions also, if in place of age, PA is
replaced by, or interpreted as, aggregate tenure (TA) instead.
The notations above are basic to the study of recruitment age taken up below.
One of the decisions an organisation is faced with during a recruitment planning process
is that of specifying the age for personnel proposed to be recruited to the organisation at
various levels. In this context, it is pertinent to note that in many developing and Eastern
countries, recruitment notifications frequently are found to mention desirable age
specifications for recruits.
The rationale for, and advantages of, such recruitment age specifications are usually
three fold. The first is that of ensuring the availability of persons with desired experience
and expertise at all levels of the organisation, (particularly in the higher levels). The
second is that of ensuring smooth flow of personnel up the hierarchical levels within their
working lives and avoiding stagnation. And the third is that of maintaining a degree of
homogeneity of the work force in the grades.
The rationale for using the PA as a measure of the experience level is that an
employee with sufficiently high experience or tenure levels, during the course of his
working life both within the organisation and outside of it, would also have attained a
high PA, as well as the vice-versa. This is particularly true of the middle and higher
grades of the hierarchy, wherein the experience and tenure across all organisations put
together will be of good value. This is captured in the PA. (Alternatively, the aggregate
tenure (TA) can also be directly used instead).
These objectives can be translated down to the achievement and maintenance of
desired PA (or TA) distributions within the system. The recruitment ages (or TA) can
then be so chosen, to be able to achieve the desired PA(TA) distributions. In our analysis,
as in the literature, by distribution, we mean the proportions of members with different
PA(TA).
77
(4.1)
(4.2)
In equation (4.2), Da is a diagonal matrix whose elements are the aj, and * and % are two
array operators defined as below:
x * y = ( x1 y1 , x2 y2 ,....) and x% y = ( x1 / y1 , x2 / y2 ,...).
Post-multiplication of both sides of equation (4.2), first by n() using the array *
operator, and then by (I P) yields
R () Da = [ M d * n()]( I P) R() P ( I P )1
(4.3)
(4.4)
which yields an expression for A, the required recruitment age at each level of the system.
The method yields a single set of values of the recruitment ages. Adhering strictly to
these would not be feasible in practice, and what is intended in this method is that the
planners try to keep the recruitment ages close to the specified values. Alternatively, the
planner may specify the desired range for the MeanPA (minimum and maximum values),
for which, the method will then yield the desired range of recruitment ages (minimum
and maximum ages). This is illustrated in the examples subsequently. The method is
simple and operationally easy to apply in practice.
Identical results hold for TA also.
78
u ( , x ) / n ( ) x
j
for j = 1, 2, ., k
(4.5)
x =b j
And the problem of determining the recruitment ages can be stated as that of determining
the recruitment age vector, A, to satisfy the condition (4.5) above.
Equation (4.5) can be written in the following more convenient form
u ( , x ) x n
l
for l = 1, 2.., k
l l
(4.6)
x =bl
Now, using the notation of Section 3, equation (4.6) can be written in vector form as
( R()e )(e P
'
l
x =bl
( x al )
)] n() * X
(4.7)
l =1
( R()e )(e P
'
l
( x al )
)] n() * X
(4.8)
l =1 x =bl
( R()e ')(e P
l
( bl al )
l =1
( R()e ')(e P
l
(bl al )
)( I P) 1 n() * X
(4.9)
l =1
Equation (4.9) can then be used to determine the desired recruitment ages.
For super-diagonal P, equation (4.9), written in scalar form for grade j is
R j () p jj
(b j a j )
j 1
[ R (){ p
l
(bl al )
. p j 1, j
l , j 1
l =1
n j () x j (1 p jj )
/ (1 p j 1, j 1 ) + plj (bl al ) }]
(4.10)
for j = 1, 2,....., k
(4.11)
j 1
(b j a j )
[ R (){ p
( bl al )
. p j 1, j
l , j 1
/ (1 p j 1, j 1 ) + plj (bl al ) }]
l =1
79
(4.12)
+ n j () x j (1 p jj )
In equation (4.12), all terms except the first term involving aj are known from the
previous computations. Notably, it may be observed that since P is upper triangular,
Pjj(bj aj) = (pjj)bj aj, where the RHS is simply the scalar diagonal element of P raised to
the power bj aj, which simplifies the computation considerably.
This method yields the upper limits on the PA of recruits to the grades, as it specifies
the maximum permissible ages of recruits to the various grades; and it is not uncommon
to find recruiters specifying such maximum age limits.
This method of restricted tails can as well be applied to achieve the first objective of
the planner, i.e., that of ensuring that the grades comprise of members of adequate
experience, and hence of adequate PA/TA. In this method, as an alternative to specifying
the MeanPA/TA, the planner might require that the grades should not contain too many
members of insufficiently low PA/TA. This specification can again be translated down to
a restricted tails criterion, wherein, the area under the tail of the distribution at the low
PA/TA end (lower tail) is restricted to specified low values. In the case of age, let hj be
the desired cut-off level of PA for grade j, and qj the maximum permissible proportion of
grade j that can have PA below hj, and let Q be the row vector whose elements are the qj.
Then the criterion of restricted tails can be stated as
h j 1
u ( , x ) / n ( )
j
qj
for j = 1, 2, ., k
(4.13)
x =0
And the problem of determining the recruitment ages can be stated as that of determining
the recruitment age vector, A, to satisfy the condition (4.13) above, which can be written
in the following more convenient form as
n j ( )
u ( , x ) n ( ) q
j
for j = 1, 2, , k
(4.14)
x=hj
and equivalently as
u (, x) n ()(1 q )
j
(4.15)
x=hj
[ R ( )e ] [e P
1
l =1
( hl al )
(4.16)
80
(h j a j )
j 1
[ R (){ p
( hl al )
. p j 1, j
l , j 1
/ (1 p j 1, j 1 ) + plj ( hl al ) }]
l =1
(4.17)
n j ()(1 q j )(1 p jj )
(4.18)
From which a1 can be determined. And for the higher and subsequent grades
R j () p jj
(h j a j )
j 1
[ R (){ p
l
( hl al )
. p j 1, j
l , j 1
l =1
/ (1 p j 1, j 1 ) + plj ( hl al ) }]
(4.19)
+ n j ()(1 q j )(1 p jj )
81
(4.20)
And thus,
x
U ( , x ) =
P (w) D P
r
xw
for x = 0,1, 2,
(4.21)
w=0
where DR is a diagonal matrix whose diagonal elements are the elements of R().
Noting that the term Pr(x) in the equations above is present only in the equations for
U(, s) for s x, equation (4.21) yields an expression for Pr(x), the vector of
probabilities (or proportions) of recruits to the grades with PA = x, as under
Pr ( x) = (U (, x) U (, x 1) P ) DR 1
for x = 1, 2, ,
(4.22)
and
Pr (0) = U (, 0) DR 1
for x = 0
(4.23)
The set of equations above can be used to determine the vector of the probabilities of
recruits to the grades with different ages (the recruitment age distributions).
However, since these values have necessarily to be non-negative numbers between
zero and unity, and should sum to unity, it is evident that not every arbitrary PA
distribution will be compatible with the system P (and F) matrix, thereby showing that
only certain PA distributions can be attained and maintained for a given set of promotion
(and proportionality) policies of the system. We can hence define the maintainability of a
collection of PA distributions across the grades of a system as under:
Definition 1: A collection of PA distributions across the grades of a system, one for each
grade, is maintainable if there exists a collection of recruitment age distributions, from
equations (4.22) and (4.23,) across the grades, one for each grade, which satisfies the
following conditions:
1
Pr ( x)
(4.24)
Pr ( x ) l
(4.25)
P ( x) = l
(4.26)
x =0
These are the conditions of non-negativity, upper bound of unity, and unity sum, of the
probability mass points of the collection of recruitment age distributions.
In order to test the compatibility and maintainability of a given collection of PA
distributions of a system with the given system parameters, P (and F), simple tests can
thus be devised, and are stated in theorem 4.1 below.
Theorem 4.1: For any given set of system parameters P (and F), with P upper-triangular,
a collection of PA distributions for the grades of the system, is a maintainable collection
of distributions, if its probability mass points satisfy the following inequalities
simultaneously
1
PPA ( w) PPA ( w 1) P
for every w = 1, 2,
(4.27)
82
for every w = 1, 2,
(4.28)
PPA ( w)
for w = 0
(4.29)
PPA ( w) R ()%n()
for w = 0
(4.30)
PPA (0) +
= R ()%n()
(4.31)
(P
PA ( x ) PPA ( x 1) P )
x =1
(4.32)
Pr (0) = U (, 0)% R ()
(4.33)
and
respectively, and substituting equations (4.32) and (4.33) into (4.24) and (4.25), and
making use of equation (4.20), yields after some rearrangement, conditions (1) to (4)
above. Condition (5) is established by substituting equations (4.32) and (4.33) into
equation (4.26) and again making use of equation (4.20) yielding
(P
PA ( x ) PPA ( x 1) P ) * n()]% R ( )
x =1
=l
(4.34)
from which, condition (5) follows upon post multiplication by and , using the * and %
array operators respectively.
This establishes the results of the theorem.
The existence of a collection of PA distributions satisfying the conditions of
theorem 4.1 is shown below in theorem 4.2
Theorem 4.2: There exists a collection of PA distributions across the grades, one for each
grade, which satisfies the conditions of theorem 4.1
Proof: The existence of a collection of PA distributions satisfying the conditions of
theorem 4.1 is proved by showing the non-negativity, and unity sum of such PPA(x) (from
which the condition of unity upper bound automatically follows).
The non-negativity of the PPA(x) which satisfies the conditions of theorem 4.1 is
ensured by conditions 1 and 3 of the theorem itself (since the elements of P are all
non-negative).
That the PPA(x) which satisfy the conditions of theorem 4.1 will sum to unity, is
proved by writing equation (4.31) as
PA ( x )( I
x =0
P ) = R()%n()
(4.35)
83
which yields
1
PPA ( x) = R ()( I P ) %n() = l
x = 0
(4.36)
where, we have made use of the identity: R() = n()(I P), which holds for all Markov
manpower systems, thereby proving the theorem.
The related and consequent problem of rectification of a collection of PA
distributions if found non-maintainable, is essentially that of constructing suitable PA
distributions that would not violate the conditions of theorem 4.1.
Identical results hold for TA distributions also.
(4.37)
(4.38)
(4.39)
(4.40)
and then to
x 1
PPA ( x) j p j 1, j (
PA ( x 1 l ) j ( p jj )
) + PPA (0) j ( p jj ) x
(4.41)
l =0
which again shows that it is possible to have a wide variety of shapes of the PPA(x)j
curves, however, with the tail decaying to zero asymptotically, in view of equation
(4.41).
84
c
PPA ( x) j p j 1, j (
l =0
x amin, j 1 1
(4.42)
which again permits a wide variety of shapes for the PA curves, with the tail again
decaying to zero asymptotically in view of equation (4.42).
d
It is conceivable that the planner might want the following conditions to hold for PA
proportions: PPA(x)j = 0 for x amin,j, where amin,j is the desired minimum PA below
which range the PA proportions in grade j are zero.
Such a distribution is maintainable, as can be seen from the above equations.
However, a condition of the type: PPA(x)j = 0 for x amax,j is not maintainable, as can
be seen from equation (4.42).
Thus within the interval [amin,j, ), the PA proportions can have any desired profile,
provided they do not violate equations (4.40) to (4.42) , and thus, the maintainable set
admits a wide variety of PA distributions.
A more detailed characterisation of the maintainable set offers scope for further
research.
If a collection of PA distributions is maintainable, then the collection of recruitment
age distributions which would maintain it can be obtained from the working equations
below.
Pr (0) = PPA (0) * n()% R()
(4.43)
(4.44)
in practice, the values are restricted to integer values, which can cause the PA
distributions to fail the test conditions,
Nevertheless, these equations can still be used in practice, although with a certain degree
of approximation, as is shown in the example in Section 5.2 below.
All the results above extend in an identical manner to TA distributions also.
85
Examples
0.2
0.8
0
0.05
0.65
0
0
0.2
0.85
0
0
F =
0
1
0
0
2
0
0
0
and
R1 ( ) = 100,
which yield, R() = (100, 50, 12.5, 5.37). (In general systems R() can be chosen
arbitrarily).
To illustrate the determination of recruitment ages by the first method of MeanPA, we
fix the desired MeanPA vector as [20, 28, 36, 45].
The desired recruitment age vector is then readily calculated using equation (4.4), as
A = (18.5, 27, 44.4, 48.5).
If, instead, the desired range of MeanPA is fixed as [20, 28, 36, 45] to
[25, 33, 40, 50], the range of recruitment ages comes out as (18.5, 27, 44.4, 48.5) to
(23.5, 32, 46.4, 54.7), thereby providing a measure of flexibility to the planner.
We next illustrate the determination of the recruitment age vector A using the
restricted tails criterion. The system we use for this computation is one with the same P,
with F = diagl(0,1,1,2). The other parameters are as follows:
a
the vector X, of maximum permissible proportion of the grades with PA above the
cut off PA, is taken as (.05, .10, .20, .15).
a1 < 24.14
86
P=
0
0
0.85 0.067
0
0
0.84
0
The company followed proportionality policies which yielded a value of the F matrix as:
F = diagl{0, 2, , ), i.e., the recruitment to grade two was equal to half the number of
promotions to it, and recruitment to grades three and four were proscribed. The steady
state recruitment vector for the system was: R() = (30, 11, 0, 0).
The actual prevailing PA proportions across the grades are given in terms of the
uj(, x) values in Table 1, which give very jagged PA curves, and the smoothed-out
version, in Table 2, wherein we allow the values to be fractional. Using equations (4.43)
and (4.44), the recruitment age distributions were calculated and are tabulated in Table 3
and plotted in Figure 1 and Figure 2, for grades one and two respectively, which yield
approximate bimodal distributions for the recruitment ages. (There was no recruitment to
grades three and four). The curves show some very small negative values, which in
practice, are to be interpreted as zero, since exact zero values are usually never achieved,
due to the reasons given in Section 4.2. For the same reasons, the negative values of the
proportions at the tail of the distributions are also neglected, and the values then
normalised. Thus, the recruitment age distributions which would maintain the prevailing
PA distributions are arrived at with a degree of approximation.
Table 1
PA
u1
u2
u3
u4
20
22
24
26
28
30
32
34
36
38
40
42
44
46
48
50
52
54
56
58
60
0
0
6
18
20
13
7
8
7
15
11
16
27
29
17
19
19
18
5
8
0
0
0
2
2
9
11
9
3
7
4
10
7
7
4
9
16
14
16
4
8
0
0
0
0
0
0
0
0
0
3
1
2
4
8
10
1
6
3
5
8
2
0
0
0
0
0
0
0
0
4
8
3
4
4
3
0
1
3
3
5
0
1
0
87
PA
u1
u2
u3
u4
24
26
28
30
32
34
36
38
40
42
44
46
48
50
52
54
56
58
60
4.5
15
16.5
14.5
12.5
10.5
8
6.5
6
12
23.5
26.5
27.5
21
17
14
11
9
0
0
1
3.5
8
9.5
9
7.5
6.5
5.5
6
7
10.5
13.5
14.5
14
13
11
8
0
0
0
0
2
2.5
3.5
4
4.5
4
3.8
3.7
3.5
3.7
4
4
4
4
4
0
0
0
0
0
0
1.5
1.8
2
2.5
2.7
3
3
3
3
3
3
3
3
0
Table 3
PA
pr1
pr2
24
26
28
30
32
34
36
38
40
42
44
46
48
50
52
54
56
58
60
0.1754
0.4269
0.117
0.0136
0.0214
0.0292
0.0565
0.0273
0.0058
0.2573
0.4951
0.2086
0.1423
0.1462
0.0741
0.0507
0.0624
0.0351
0.3158
0
0.0714
0.176
0.3901
0.1893
0.0364
0.0539
0.0184
0.0263
0.0995
0.1253
0.3174
0.3091
0.1644
0.0709
0.0359
0.0571
0.1649
0.6605
88
Figure 1
The computed recruitment age distribution for grade one (see online version for
colours)
0.35
0.3
Proportions
0.25
0.2
0.15
NldPr1
0.1
0.05
0
-0.05 0
20
40
60
80
-0.1
Rectt Age
Figure 2
The computed recruitment age distribution for grade two (see online version for
colours)
0.25
Proportions
0.2
0.15
NldPr2
0.1
0.05
0
-0.05
20
40
60
80
Rectt Age
Figure 3
The actual recruitment age distribution for grade one (see online version for colours)
0.3
Proportions
0.25
0.2
0.15
ActDstr1
0.1
0.05
0
-0.05
20
40
Rectt Age
60
80
89
The actual recruitment age distribution for grade two (see online version for colours)
0.3
0.25
Proportions
0.2
0.15
ActDstr2
0.1
0.05
0
-0.05
20
40
60
80
Rectt Age
ActDstr1
NldPr1
CuActDst1
CumPr1
KSD1
22
24
0.091
0.10438
0.091
0.10438
0.01338
26
0.161867
0.254047
0.252867
0.358427
0.10556
28
0.093025
0.069626
0.345892
0.428053
0.08216
30
0.034222
0.00809
0.380113
0.41996
0.03985
32
0.01259
0.01274
0.392703
0.407224
0.01452
34
0.004631
0.01738
0.397334
0.389848
0.007487
36
0.001704
0.03362
0.399038
0.356225
0.042813
38
0.000627
0.01625
0.399665
0.339979
0.059686
40
0.000231
0.003452
0.399895
0.34343
0.056465
42
0.105
0.153118
0.504895
0.496548
0.008347
44
0.274332
0.294632
0.779227
0.791181
0.01195
46
0.139548
0.124137
0.918775
0.915318
0.003458
48
0.051337
0.084682
0.970112
0.02989
50
0.018886
0.988998
0.011
52
0.006948
0.995946
0.00405
54
0.002556
0.998502
-0.0015
56
0.00094
0.999442
-0.00056
58
0.000346
0.999788
-0.00021
60
0.000127
0.999915
-8.5E-05
Max
0.059686
Min
0.10556
90
ActDstr2
NldPr2
CuActDst2
CumPr2
KSD2
20
22
24
26
0.008
0.035791
0.008
0.035791
-0.02779
28
0.092
0.088225
0.1
0.124016
-0.02402
30
0.279323
0.195549
0.379323
0.319565
0.059758
32
0.139545
0.094892
0.518868
0.414457
0.104412
34
0.051336
0.018247
0.570204
0.432703
0.137501
36
0.018885
0.027019
0.58909
0.459722
0.129367
38
0.006948
0.00922
0.596037
0.450499
0.145538
40
0.002556
0.01318
0.598593
0.437315
0.161278
42
0.00094
0.049877
0.599533
0.487192
0.112341
44
0.092
0.06281
0.691533
0.550003
0.141531
46
0.161009
0.159106
0.852543
0.709108
0.143435
48
0.093077
0.154945
0.94562
0.864053
0.081566
50
0.034241
0.08241
0.979861
0.946463
0.033397
52
0.012597
0.035541
0.992457
0.982004
0.010453
54
0.004634
0.017996
0.997091
-0.00291
56
0.001705
0.998796
-0.0012
58
0.000627
0.999423
-0.00058
60
0.000231
0.999654
-0.00035
Max
0.161278
Min
-0.02779
To check the validity of the results, they were compared with the actual recruitment age
distributions which were as under.
The recruitment to grade one had two streams: young engineers from external sources
as well as of those taken from the non-supervisory cadres below grade one. The grade
was therefore, a mix of people who had risen up from the ranks and were hence of higher
age (of age 42 and above) and young engineers recruited to the grade from external
sources who belonged to a lower age group (of age 24 and above), with both groups
skewed towards the lower ages. The recruitment ages in grade two were again found to
fall into two groups, one between 28 and 34, and one between 44 and 48. These are given
in Table 5, and shown in Figure 3 and Figure 4 for grades one and two respectively.
A goodness of fit test was performed using the Kolmogorov-Smirnov D statistic and
is shown in Table 4(a) and Table 4(b). We can note a good degree of conformance of the
calculated recruitment age distributions with the actual data.
ActDstr2
0
0
0
0.008
0.092
0.279323
0.139545
0.051336
0.018885
0.006948
0.002556
0.00094
0.092
0.161009
0.093077
0.034241
0.012597
0.004634
0.001705
0.000627
0.000231
Notes: S is the state space of the discrete-time Markov chain, and consists of the n-tuples
(n1, n2, n3, , nk) , where each of the nj can take on non-negative values.
The state space of the chain is thus the k-fold Cartesian product of the set of
non-negative real numbers, R+, i.e., S = R+ R+ R+ R+ = {R+}k.
91
92
Conclusions
This paper has studied the determination of the recruitment ages which would meet two
prime objectives of the manpower planner, viz. to ensure that the grades are manned by
staff with adequate experience, and to ensure smooth flow of personnel up the grades of
the hierarchy.
The first two methods proposed treat recruitment ages as deterministic, namely, the
one based on MeanPA, and the method of restricted tails. Both methods are quick, robust,
distribution-free, and easy to put into practice, and would be of good value to planners.
Under stochastic recruitment ages, firstly, conditions for maintainability of a collection of
PA distributions across the grades have been derived, and then the recruitment age
distributions for the grades which would maintain them. The results have been illustrated
with computations, and validated with real-world data.
The methods and results extend in an identical manner to aggregate tenure
distributions also.
A detailed study of the maintainable set offers scope for further research.
Acknowledgements
The authors would like to thank the three anonymous referees for their very detailed
comments, the incorporation of which has substantially enhanced the quality of the paper.
This research was fully funded and supported by the Department of Management Studies,
Indian Institute of Science, Bangalore-560012, India.
References
Aronson, J.E. and Thompson, G.L. (1985) The solution of multi-period personnel planning
problems by the forward simplex method, Large Scale Systems, Vol. 9, pp.129139.
Arthur, J.B. (1994) Effects of human resource systems on manufacturing performance and
turnover, Academy of Management Journal, Vol. 37, pp.670687.
Bartholomew, D.J. (1982) Stochastic Models for Social Processes, 3rd ed., John Wiley & Sons,
New York.
Bateman, T.S. and Strausser, S. (1984) A longitudinal analysis of the antecedents of organisational
commitment, Academy of Management Journal, Vol. 27, pp.95112.
Bazargan Lari, M., Gupta, P. and Young, S. (2003) A simulation approach to manpower planning,
Proceedings of the 2003 Winter Conference of ACM, Vol. 2, pp.16771885.
Blythe, J., Baumann, A., Zeytinoglu, Isik, U., Denton, M., Akhtar-Danesh, N., Davies, S.,
Kolotylo, C. (2008) Nursing generations in the contemporary workplace, Public Personnel
Management, Vol. 37, No. 2, pp.137159.
Carette, P. (1999) Modelling hierarchical systems by a continuous-time homogeneous Markov
chain using two-wave panel data, Journal of Applied Probability, Vol. 36, No. 3, pp.644653.
Cassidy, D. and Sutherland, J. (2008) Going absent then just going? A case study examination
of absence and quitting, Economic Issues, Vol. 13, No. 2, pp.119.
Cennamo, L., and Gardner, D. (2008) Generational differences in work values, outcomes, and
person-organization values fit, Journal of Managerial Psychology, Vol. 23, No. 8,
pp.891906.
93
Chandra, S. (1988) Two characteristic Markov type manpower models, Opsearch, Vol. 19, No. 2,
pp.1830.
Chandra, S. (1989) Three characteristic Markov type manpower flow models, Opsearch, Vol. 20,
No. 1, pp.2835.
Chattopadhyay, A.K. and Gupta, A. (2007) A stochastic manpower planning model under varying
class sizes, Annals of Operations Research, Vol. 155, No. 1, pp.4149.
Chattopadhyay, P. (1999) Beyond direct and symmetrical effects: the influence of demographic
dissimilarity on organizational citizenship behaviour, Academy of Management Journal,
Vol. 42, No. 3, pp.273287.
Chew, L.W. and Khoong, C.M. (1994) Optimization in staff movement planning, TIMS, XXXII
International Meeting, Anchorage, Alaska.
Cho, J.E. and Hu, H. (2009) The effect of service quality on trust and commitment varying across
generations, International Journal of Consumer Studies, Vol. 33, No. 4, pp.468476.
Cohen, A. (1993) Age and tenure in relation to organizational commitment: a meta-analysis,
Basic and Applied Social Psychology, Vol. 14, No. 2, pp.143159.
Daily, C.M. (1994) In Japan, birds of a feather do flock together, Academy of Management
Executive, Vol. 8, No. 3, pp.103104.
De Feyter, T. and Guerry, M-A. (2009) Markov models in manpower planning: a review, HUB
Research Centre, Centre for Manpower Planning, Articles in Economics and Management,
No. 156971, Nova Science Publishers, Brussels.
Donatienne, D. and Mathieu, G. (2008) When a worker becomes an older worker: the effects of
age-related social identity on attitudes towards retirement and work, Career Development
International, Vol. 13, No. 2, pp.168185.
Finegold, D., Mohrman, S. and Spreitzer, G.M. (2002) Age effects on the predictors of technical
workers commitment and willingness to turnover, Journal of Organizational Behaviour,
Vol. 23, No. 5, pp.655674.
Fry, M.J., Magazine, M. and Rao, U.S. (2003) Firefighter staffing including temporary absences
and wastage, Technical paper, College of Business, University of Cincinnati, July 2003.
Gaimon, C. and Thompson, G.L. (1984) A distributed parameter cohort personnel planning model
that uses cross-sectional data, Management Science, Vol. 30, No. 6, pp.750764.
Georgiou, A.C. and Vassiliou, P.C.G. (1997) Cost models in non-homogenous Markov systems,
European Journal of Operations Research, Vol. 100, No. 1 pp.8196.
Gerontidis, I.I. (1995) Periodicity of the profile process in Markov manpower systems, European
Journal of Operations Research, Vol. 85, No. 3, pp.650659.
Glen, J.J. (1977) Length of service distributions in Markov manpower models, Operations
Research Quarterly, Vol. 28, No. 4, pp.975982.
Glen, J.J. and Yang, C.L. (1996) A model for promotion rate control in hierarchical manpower
systems, The IMA Journal of Management Mathematics, Vol. 7, No. 3, pp.197206.
Grinold, R.C. (1978) Optimal design of a manpower system, Mathematical Programming,
Vol. 15, No. 1, pp.2635.
Guerry, M.A. (1999) Using fuzzy sets in manpower planning, Journal of Applied Probability,
Vol. 36, No. 1, pp.155162.
Hallier, J. (2001) Greenfield recruitment and selection, Personnel Review, Vol. 30, No. 3,
pp.331350.
Hartline, M.D. and De Witt, T. (2004) Individual differences among service employees: the
conundrum of employee recruitment, selection and retention, Journal of Relationship
Marketing, Vol. 3, Nos. 2/3, pp.2532.
Hayne, W.J. and Marshall, K.T. (1977) Two-characteristic Markov type manpower flow models,
Naval Logistics Research Quarterly, Vol. 24, No. 2, pp.235255.
Hewitt, S.D. (2009) The secrets of successful succession planning in the new age wave, Industrial
and Commercial Training, Vol. 41, No. 4, pp.181186.
94
Hult, C. and Edlund, J. (2008) Age and labour market commitment in Germany, Denmark,
Norway and Sweden, Work, Employment and Society, Vol. 22, No. 1, pp.109128.
Iverson, R.D. (1999) An event history analysis of employee turnover: the case of hospital
employees in Australia, Human Resource Management Review, Vol. 9, No. 4, pp.397418.
Jamal, M. and Baba, V.V. (1991) Type A behaviour, its prevalence and consequences among
women nurses: an empirical examination, Human Relations, Vol. 44, No. 11, pp.12131228.
Kalamatianiou, A.G. (1987) Attainable and maintainable structures in Markov manpower systems
with pressure in the grades, Journal of Operational Research Society, Vol. 38, pp.183190.
Khoong,
C.M.
(1999)
Some
optimization
models
in
manpower
planning,
Information-Knowledge-Systems Management, Vol. 1, No. 2, pp.159171.
Kirchmeyer, C. (1995) Demographic similarity to the work group: a longitudinal study of
managers at the early career stage, Journal of Organizational Behaviour, Vol. 16, No. 1,
pp.6783.
Klingman, D. and Phillips, N. (1984) Topological and computational aspects of pre-emptive
multi-criteria military personnel assignment problems, Management Science, Vol. 30,
pp.13621374.
Kumar, B.P. and Giri, V.N. (2009) Effect of age and experience on job satisfaction and
organizational commitment, ICFAI Journal of Organizational Behaviour, Vol. 8, No. 1,
pp.2836.
Liao, H., Joshi, A. and Chuang, A. (2004) Sticking out like a sore thumb: employee dissimilarity
and deviance at work, Personnel Psychology, Vol. 57, No. 4, pp.9691000.
Lord, R.L. and Farrington, P.A. (2006) Age-related differences in the motivation of knowledge
workers, Engineering Management Journal, Vol. 18, No. 3, pp.2026.
Lorence, J. (1987) Age differences in work involvement, Work and Occupations, Vol. 14, No. 4,
pp.533557.
Loscocco, K.A. and Kalleberg, A.L. (1988) Age and meaning of work in the United States and
Japan, Social Forces, Vol. 67, No. 2, pp.337356.
Meyer, J.P., Stanley, D.J., Herscovitch, L. and Topolnytsky, L. (2002) Affective, continuance, and
normative commitment to the organization: a meta-analysis of antecedents, correlates and
consequences, Journal of Vocational Behaviour, Vol. 61, pp.2052.
Mukherjee, S.P. and Chattopadhyay, A.K. (1990) A stochastic analysis of a staffing problem,
Journal of Operational Research Society, Vol. 40, No. 5, pp.489494.
Nasrudin, A.M. (2005) The role of noninstrumental justice and age in predicting organizational
commitment, Gadjah Mada International Journal of Business, Vol. 7, No. 1, pp.114.
Nehra, V. and Khurana, D.K. (1990) A computerized manpower planning model for hierarchical
organizations, Opsearch, Vol. 27, No. 10, pp.1932.
Nilakantan, K. and Raghavendra, B.G. (2008) Length of service and age characteristics in
proportionality Markov manpower systems, The IMA Journal of Management Mathematics,
Vol. 19, No. 1, pp.245268.
Onyemah, V. (2009) The effects of coaching on salespeoples attitudes and behaviours: a
contingency approach, European Journal of Marketing, Vol. 43, Nos. 7/8, pp.938960.
Ornstein, S., Cron, W.L. and Slocum, J.W. Jr. (1989) Life stage versus career stage: a comparative
test of the theories of Levinson and Super, Journal of Organizational Behaviour, Vol. 10,
No. 2, pp.117133.
Papadopoulou, A.A. and Tsaklidas, G. (2004) Non-homogeneous semi-Markov systems in discrete
time and related renewal paths, Technical paper, Dept. of Mathematics, University of
Thessaloniki, Greece.
Poornachandra Rao, (1991) A dynamic programming approach to determine optimal manpower
recruitment policies, Journal of Operational Research Society, Vol. 41, No. 10, pp.983988.
95
Popoola, S.O. and Oluwole, D.A. (2007) Career commitment among records management
personnel in a State Civil Service in Nigeria, Records Management Journal, Vol. 17, No. 2,
pp.107116.
Popova, E. and Morton, D. (1998) Adaptive manpower scheduling, in Medeiros, D.J.,
Watson, E.F., Carson, J.S. and Manivannan, M.S. (Eds.): Proceedings of the Winter
Simulation Conference, University of Texas at Austin, pp.661668.
Raghavendra, B.G. (1991) A bivariate model for Markov manpower planning systems, Journal of
the Operational Research Society, Vol. 42, No. 7, pp.565570.
Ragu-Nathan, T.S., Tarafdar, M., Ragu-Nathan, B.S. and Quiang, T. (2008) The consequences of
technostress for end users in organizations: conceptual development and empirical validation,
Information Systems Research, Vol. 19, No. 4, pp.417433.
Redman, T. and Snape, E. (2006) The consequences of perceived age discrimination amongst
older police officers: is social support a buffer?, British Journal of Management, Vol. 17,
No. 2, pp.167175.
Riordan, C.M., Griffith, R.W. and Weatherly, E.W. (2003) Age and work-related outcomes: the
moderating effects of status characteristics, Journal of Applied Social Psychology, Vol. 33,
No. 1, pp.3757.
Sager, J.K. (1991) Recruiting and retaining committed salespeople, Industrial Marketing
Management, Vol. 20, No. 2, pp.99103.
Shadur, M.A., Rodwell, J.J. and Bamber, G.J. (1995) Factors predicting employees approval of
lean production, Human Relations, Vol. 48, No. 12, pp.14031426.
Sheard, M. (2009) Hardiness commitment, gender, and age differentiate university academic
performance, British Journal of Educational Psychology, Vol. 79, No. 1, pp.189205.
Silverman, J., Steuer, R.E. and Whisman, A.W. (1988) A multi-period, multi-criteria optimization
system for manpower planning, European Journal of Operations Research, Vol. 34,
pp.160170.
Sinha, J.B.P. (1990) Work Culture in the Indian Context, Sage Publications, New Delhi.
Sirvanci, M. (1984) Forecasting manpower losses by the use of renewal models, European
Journal of Operations Research, Vol. 16, pp.1318.
Skulj, D., Vehovar, V. and Stamfelj, D. (2008) The modelling of manpower by Markov chains a
case study of the Slovenian Armed Forces, Informatica, Vol. 32, pp.289297.
Snape, E. and Redman, T. (2003) Too old or too young? The impact of perceived age
discrimination, Human Resource Management Journal, Vol. 13, No. 1, pp.7889.
Standridge, C.R. (1979) Using simulation in health manpower planning, ACM SIGSIM Digest,
Vol. 10, No. 4, pp.6062, ACM New York.
Sun, L-Y. and Pan, W. (2008) HR practice perceptions, emotional exhaustion, and
work-outcomes: a conservation of resources theory in the Chinese context, Human Resource
Development Quarterly, Vol. 19, No. 1, pp.5574.
Tabbodi, M.L. (2008) Effects of leadership behaviour on the faculty commitment of humanities
departments in the university of Mysore, India: regarding factors of age group, educational
qualifications and gender, Educational Studies, Vol. 35, No. 1, pp.2126.
Tsantas, N. (2001) Ergodic behaviour of a Markov chain model in a stochastic environment,
Mathematical Methods of Operations Research, Vol. 54, No. 1, pp.101117.
Tyagi, P.K. and Wotruba, T.R. (1998) Do gender and age really matter in direct selling? An
exploratory study, Journal of Marketing Management, Vol. 8, No. 2, pp.2233.
Ugwuowo, F.I. and McClean, S.I. (2000) Modelling heterogeneity in a manpower system: a
review, Applied Stochastic Models in Business and Industry, Vol. 16, No. 2, pp.99110.
Vajda, S. (1978) Mathematics of Manpower Planning, John Wiley & Sons, Chichester.
Vassiliou, P.C.G. (1986) Asymptotic variability of nonhomogeneous Markov systems under cyclic
behaviour, European Journal of Operations Research, Vol. 27, pp.215228.
96
Vassiliou, P.C.G. and Gerontidas, I. (1985) Variances and covariances of the grade sizes in
manpower systems, Journal of Applied Probability, Vol. 22, pp.583597.
Wagner, S.L. and Rush, M. (2000) Altruistic organizational citizenship behaviour: context,
disposition and age, Journal of Social Psychology, Vol. 140, No. 3, pp.379391.
Wegge, J., Roth, C., Kanfer, R., Neubach, B. and Schmidt, K-H. (2008) Age and gender diversity
as determinants of performance and health in a public organization, Journal of Applied
Psychology, Vol. 93, No. 6, pp.13011313.
Whitener, E.M. (2001) Do high-commitment human resource practices affect employee
commitment? A cross level analysis using hierarchical linear modeling, Journal of
Management, Vol. 27, pp.515535.
Witt, L.A., Treadway, D.C. and Ferris, G.R. (2004) The role of age in reactions to organizational
politics perceptions, Organizational Analysis, Vol. 12, No. 1, pp.3952.
Wood, S. and De Menezes, L. (1998) High commitment management in the UK: evidence from
the workplace, industrial relations survey and employers manpower and skill practices
survey, Human Relations, Vol. 51, pp.485515.
Zanakis, S.H. and Maret, M.W. (1981) A Markov chain application to manpower supply
planning, Journal of Operational Research Society, Vol. 31, pp.10951102.