Sie sind auf Seite 1von 10

FREDERICK P . R E H M U S AND HARVEY M .

W A G N E R

APPLYING LINEAR P R O G R A M M I N G 89

TO YOUR PAY S T R U C T U R E

ments in order to bring about a consistent


scale of compensation within a firm. It also
ow much should a man be paid?"
H Many a manager has wished that he
aims at making the finn's scale generally
competitive with those in the industry. As a
could use an objective and systematic ap-
consequence, the job evaluation approach
proach to this sensitive question. Job eval-
discovers positions currently being incor-
uation has become accepted as an integral
element of salary administration, since, de- rectly compensated and establishes a means
spite its imperfections, it leads to more fac- whereby salaries for new positions are easily
determined. Thus, when the results of a job
tual and rational compensation decisions.
Specifically, it diagnoses a salary structure evaluation study are fully implemented over
a period of time, the compensation struc-
by determining the relative importance of
the compensable elements of every job and ture adjusts toward a system of adequate re-
assigns monetary equivalents to these ele- wards and incentives for service.
This article suggests how linear program-
ming, a modern technique of management
Mr. Rehmus is Vice-President--Administratlon of science usually applied to the production,
the Midwestern Financial Corporation in Denver,
and Mr. Wagner is Professor of Business Admlnls. distribution, marketing, and planning facets
~ratlon at Stanford University. of a corporate enterprise, can be turned to-

WINTER, 1963
FREDERICK P. REHMUS AND HAnVEY M. WAGNER

ward perfecting a company's wage and sal- with this method every salary analyst who
ary program. Before getting into detail, two faces the same raw data will arrive at the
points must be made clear at the outset. same compensation scale.
First, we do not attempt to justify the job This article tackles the problem of assign-
evaluation approach per se; to do so would ing proper weights to the compensable ele-
require a discourse of at least equal length. 1 ments in a job evaluation system, a facet of
We realize, of course, that in bypassing a the procedure that has often been subject to
discussion of the pros and cons we forgo en- severe criticism. 2 A bias created in the selec-
gagement in a number of pithy controver- tion of weights can invalidate sound ana-
sies. Our only contribution to these polemics lytical work done at earlier stages in the job
is that our approach, insofar as it enhances evaluation process. Moreover, the possibility
the validity of job evaluation, lends tacit of misjudgment at this stage is great because
support to applying the method. an error here is less apparent, intuitively or
This article also does not delve into the empirically, than at the other stages. Hence,
intricacies of obtaining the raw data to be a technique that reduces or removes the ele-
used in a job evaluation program. The analy- ment of judgment from factor weighting can
sis begins just after the basic data have been substantially improve the process.
collected.
Second, we urge that our claim to a means
of making job evaluation more objective and THE PROBLEM OF FACTOR WEIGHTING
90
scientific be understood in the context of
having to select one from among several pos- The difficulties inherent is setting factor
sible approaches to wage and salary admin- weights can be readily understood through
istration. One would be hard pressed to a hypothetical example. In surveying an or-
devise a method for properly setting com- ganization's job demands, a salary analyst
pensation levels that did not in some way has determined that three factors (A, B, and
embrace human judgment. The approach to C) are the compensable elements of jobs in
job evaluation suggested here eliminates the this organization. Salary will be based on the
arbitrary elements usually present, reduces relative demands of individual jobs in terms
discretion to its essential level, and assists of these three factors. Further, the analyst
management to focus sharply on the multi- has observed six identifiable degrees for A,
ple interacting influences that have to be ac- four for B, and five for C. These degrees are
counted for in a compensation scheme. Thus, essentially rankings, arranged from the least
with this technique of management science, to the greatest level of contribution. The
an executive needs to use his own judgment complete factor evaluation structure appears
only where it is appropriate and necessary. in Table 1.
The powerful scope of the method guaran- In Table 2, five of the jobs analyzed are
tees that both factual data and managerial rated in terms of the evaluation factors. An
goals are melded correctly. Perhaps the best examination of the above data indicates that
way to state the contention that the approach the sum of the factor degree ratings does
is objective and scientific is to assert that not correspond closely to the existing sala-
ries for individual jobs. Two reasons may
account for the discrepancies: ( 1 ) the sala-
1 See Edwin F. Beal, "In Praise of Job Evalua-
tion," California Management Review, V (Sum-
mer, ]963), 9-16 for a review of the approach's 2 A lucid critique of some of the factor-weighting
merits. Also see Arch Patton, Men, Money and techniques and a strong argument for an improved,
Motivation (New York: McGraw-Hill Book Co., more valid technique may be found in William M.
Inc., 1961) for a general picture of salary and Fox, "Purpose and Validity in Job Evaluation,"
wage administration practices. Personnel 1ournal (October, 1962), pp. 432-37.

BUSINESS HORIZONS
APPLYING LINEAR PROGRAM2klING TO YOUR PAY STRUGTURE

T3xBLE 1 ual's salary in determining correct factor


weights.
Factor Evaluation Structure: Factor De-
gree Levels On the contrary, management judgment
must be applied not only in establishing job
A1 B1 C1 requirements but also in reviewing whether
A2 B2 C2 the current salaries significantly represent a
A3 B3 C3 historical condition that is no longer con-
A4 B4 C4
sistent with the job evaluation survey. We
A5 C5
A6 firmly believe that management can exercise
this degree of judgment with far more ob-
jectivity than is possible in setting factor
TABLE 2 weights. Various techniques for weighting
Rating of Five Jobs in Terms of Evaluation factors have been developed, and it is pre-
Factors .cisely here that linear programming can
provide improved, objectively based an-
Present
Factor Sum of Faetor Monthly swers in contrast to the prevalent alterna-
Degree Ratings Degree Ratings Salary tive approaches.
Job 1 A2 B4 C2 2+4+2= 8 $ 800
Job 2 A4 B1 C4 4+1+4= 9 1,000
TRADITIONAL TECHNIQUES F O R FACTOR
Job 3 A5 B2 C5 5+2+5=12 1,100
Job 4 A5 B4 C4 5+4+4=13 1,400 WEIGHTING 91
Job 5 A6 B4 C2 6-t-4+2=12 1,500
The most common approach to solving the
weighting problem is through the use of
ries for these jobs may be out of line at "best guess" weights for each factor. Let us
present, or (2) the relative importance of assume that the existing salaries in Table 2
the factors and of the degree levels is not reflect a fair degree of rationality, except
adequately represented by the numerical perhaps for some minor discrepancies that
values given to factor degree ratings. In the need adjusting. Then factor A might be
design and implementation of actual job given twice the weight of factors B and C,
evaluation programs, these discrepancies or some other system of weighting might be
are encountered almost universally. devised to bring the sum of the factor de-
A high degree of management subjectivity grees more into line with salaries. Adding
usually enters into ferreting out the reasons artificial factor degrees as needed in order to
for the discrepancies. Unfortunately, it is not give logical salary relationships within the
possible with raw data such as those in Table total evaluation structure offers further re-
2 to answer separately the questions,"Are the finement. In the illustration in Table 2, a
data correct indicators of the compensable seventh degree level for factor A can be
elements?" and "Which current salaries are created. Then job 5 can be reevaluated to
unfair measures of contribution?" From the have a rating of A7 rather than A6. Con-
very start, one must recognize that the an- sequently, the sum of the factor degrees for
swers are inevitably mutually determined. job 5. more closely con'esponds to the pres-
This is not to say that in performing job eval- ent salary. With this type of manipulation,
uation management must ignore experience factor degree A7 is artificially induced to
and common sense to become a slave to the make the evaluation structure more rational.
data. For example, if a particular individual In the hands of a skilled analyst, this type
is paid an exceptional salary because of eir- of adjustment technique can often produce
eumstanees that are not reflected in the fae- satisfactory results. The obvious disadvan-
tot ratings of his job, it would be patent tage of the approaeh is that it is wholly sub-
nonsense for the analyst to use this individ- jeetive and arbitrary. The appearance of the

WINTER, 1963
FREDERICK 1~. REHMUS AND HARVEY M. WAGNER

final structure might differ if another analyst ratings and existing salary. 5Stated in another
performed the adjustments, and the ap- fashion, it seeks a factor weighting system
proach is thus difficult to defend on any that minimizes the discrepancy between the
scientific basis. salaries for given jobs and the factor ratings
Standard multiple correlation and regres- of these jobs. 6
sion analysis is another technique for de- With this system of weights, it is possible
termining weights? Subjectivity is decreased to examine the relationship of one job to
to some extent with this method, and it is another in quantitative terms and to estab-
necessary only to express a general mathe- lish logical salary differences between jobs
matical relation for the weights; the sub- in the organization. The only data required
sequent mathematical computations give the for analysis are the factor degree ratings and
ultimate determination. A major disadvan- existing salary for each jobd To obtain the
tage of this approach is that in actual prac- weights through linear programming, it is
tice the final result often proves to be necessary to employ an electronic digital
unworkable because the calculated weights computer. Standard computer routines are
do not correlate closely enough with existing readily available for solving linear program-
salaries; some weights are negative, and ming problems. Job evaluation applications
others seem to be disproportionately large or suggested in this article typically would re-
small. In an effort to correct these numerical quire only a few minutes of computer time.
inequities, the analyst once again must fall It is important to note that a weighting
back on subjective adjustments. structure developed through the use of linear
92 Other established techniques for solving programming should not (and ordinarily will
the weighting problem are primarily vari- not) completely "rationalize" the relation-
ants of the two described above. ship between the weighted job ratings and
existing salaries. Discrepancies will remain
after the linear programming weights are
A LINEARPROGRAMMINGAPPROACH
used, and jobs where existing salaries are
out of line may be identified through the
With the development of linear program- presence of these discrepancies.
ming techniques, an opportunity has been
created to solve the weighting problem in
job evaluation with minima/amounts of sub- LINEAR PROGRAMMING APPLIED
jectivity and associated effort.4 The linear
programming approach has as its initial goal
Two sets of data are given below to demon-
the selection of the weights that explain most
strate the effectiveness of the proposed
precisely the relationship between the factor
linear programming method. The first set
pertains to administrative and clerical posi-
s An extensive discussion of this statistical tech-
nique for developing factor weights appears in
Robert W. Gilmour, Industrial Wage and Salary
Control (New York: John Wiley & Sons, Inc., 5 For the mathematical description underlying
1956). this approach see H. M. Wagner, "Non-Linear
4An interesting pioneer application of linear Regression with Minimal Assumptions," Journal
programming was developed in A. Charnes, W. of the American Statistical Association, LVII (Sep-
Cooper, and O. Ferguson, "Optimal Estimation of tember, 1962), 572-78.
Executive Compensation by Linear Programming," 6 As previously, we assume for the discussion to
Management Science, I (January, 1955), 138-51. follow that existing salaries are the most relevant
Our experience with this approach has shown of all available compensation data for determining
that it is not flexible enough in establishing factor factor weights and reiterate that this assumption
weights to produce workable salary structures. must always come under management scrutiny.
The method proposed in the present article is We return to this point at the end of the article.
different in that it provides freedom in selecting 7 In cases where an unusually large number of
factor weights that is essential to the establish- job classifications are being evaluated, a sample of
ment of a rational compensation system. jobs will give valid results.

BUSINESS HORIZONS
APPLYING LINEAR PROGRAM"hIING TO YOUI:( PAY STRUCTURE

TABLE 3 based on thirty-five salaried job categories


ranging from messenger to purchasing rep-
Base Data of Thirty-five Administrative and
resentative and systems analyst. Each job
Clerical Job Classifications
classification was rated in terms of four
Job Factor Degree Ratings factors:
Factor K: Factor D: Factor K: Knowledge and experience; 14
Number of Knowledge, Factor A: Factor J: Job Average
Positions Employees Experience Accuracy Judgment Demands Salary* levels were identified ( K 1 . . . K14)
1 17 1 1 1 6 .535 Factor A: Accuracy; 9 degrees were iden-
2 2 1 1 1 3 .623 tiffed ( A 1 . . . A9)
3 33 3 2 1 4 .612 Factor J: Judgment; 10 degrees were iden-
4 3 3 2 1 4 .635 tiffed ( J 1 . . . J10)
5 6 3 3 2 4 .617 Factor D: Job demands; 6 levels were
6 7 2 3 1 4 .662
identified ( D 1 . . . D6).
7 9 4 2 2 4 .618
8 4 3 3 2 4 .602 Table 3 shows the ratings for each of the
9 32 4 2 2 4 .572 thirty-five job categories, the number of em-
10 1 3 3 1 5 .758 ployees in each job, and the average salary
11 11 3 2 1 5 .622 of employees in that classification, expressed
12 1 4 3 2 4 .660 as a fraction of the highest average salary
13 2 5 3 2 4 .620 (in this example, position 33). The conver-
14 14 6 3 3 2 .667 sion of average salaries to decimal fractions
15 2 7 4 4 1 .869 is numerically convenient, as will be seen
16 2 7 4 3 2 .767 below. Thus, if the highest average salary 93
17 2 7 4 4 1 .792 is $10,000, a decimal fraction of .660 cor-
18 1 7 4 3 2 .833
responds to a salary of $6,600. It is a simple
19 24 7 4 4 3 .765
20 1 7 5 5 1 .925 process to, relate to dollar salaries the re-
21 12 8 6 5 3 .780 sulting factor weights based on these deci-
22 1 9 5 6 1 .860 mal fractions.
23 8 10 6 7 3 .908
24 14 10 7 8 1 .915 Linear Program ]oh Equations The goal
25 12 9 6 7 1 .885 of a weighting system is to equate factor
26 3 10 5 5 6 .822 ratings with compensation. As noted previ-
27 2 9 5 5 1 .882 ously, a weighting system cannot be ex-
28 1 10 7 8 1 .857 pected to explain the salary structure com-
29 7 9 6 5 2 .920 pletely, that is, to result in total evahmtions
30 2 10 7 8 3 .918
in which no deviations exist between the
31 10 10 6 6 1 .940
32 2 12 7 7 3 .980 weights and current salaries. The linear pro-
33 5 11 7 6 2 1.000 gram consequently contains a relation for
34 1 12 8 8 1 .947 each job that exhibits the possibility of a dis-
35 16 14 9 10 3 .947 crepancy occurring; in this example, thirty-
*Each salary is expressed as a decimal fraction of the five job equations were included. Drawing
highest salary a m o n g the thirty-five job classifications.
from the data in Tab/e 3, the equations for
job classifications 6, 16, and 26 are written
tions, the second to executive positions. as follows:
Equally satisfactory results are obtained
from both applications. Job 6 K2q-A3WJlq-D4-d6 =.662;
Job 16 K7-I-A4-t-J3-}-D2-d16=.767;
Job 26 K10q-A5q-JSq-D6 --- d26=.822.
AD~INISTlqATIVEANDCLElqICALPOSITIONS
Notice that a discrepancy (d) between the
Sample Data The data used in this exam- total evaluation and an existing salary may
ple, taken from a major oil company, are be either positive or negative (-+), recog-

~V/NTE~R~ 1963
FREDERICK P. REHMUS AND HARVEY M. WAGNER

nizing that the job currently may be over- TABLE4


paid or underpaid.
Computer Factor Weights for Administra-
tive and Clerical Evaluation Structure
Weighting System Obiective The linear
programming technique finds a weighting Factors
system that minimizes the sum of the dis- Factor Degree K A J D
crepancies for all the jobs being analyzed,s 1 .050 .144 .050 .245
In some instances, it may be more appro- 2 .057 .224 .057 .252
priate to find weights that minimize the
3 .064 .231 .064 .259
sum of the discrepancies among all individ-
4 .071 .308 .071 .266
uals rather than among jobs. This objective
5 .078 .315 .078 .284
is accomplished by weighting the deviations
for individual job classifications by the num- 6 .120 .322 .085 .291
ber of individuals (N) in that job. 9 7 .127 .336 .092
8 .134 .343 .099
Constraints on the Weight Values In the 9 .226 .350 .106
selection of this firm's rating structure, which 10 .235 .113
contains four factors and thirty-nine degrees, 11 .242
it was believed that each of these factors 12 .249
and each of the degrees within these factors 13 .256
94 should have some bearing on compensation. 14 .263
Past experience with workable job evalua-
tion schemes indicated that no one factor
should be permitted to dominate the evalua-
tion structure. The need to give the linear TABLE5
program a frame of reference incorporating Computer Discrepancies Between Salaries
these considerations was satisfied by adding and Evaluation Weights: Administrative
mild constraints on the permissible range and Clerical Positions
of values for the optimal weights. 1° In this
illustration, the following constraints were Positions Discrepancy Positions Discrepancy
deemed appropriate: 1 --- 19 --
1 No factor may contribute more than 2 .098 20 .160
35 per cent to the sum of all factor weights. 3 --- 21 --
This restriction is accomplished by limit-
4 .023 22 (.010)
ing the weight of the top degree of each
5 --4 2~ ----
factor (K14, A9, J10, D6) to a value not
.045 24
exceeding .35. For example, the relation 6
.006 25
K14 _< .35 is added; a total of four such 7
constr~fints are put into the program. 8 (.015) 26 (.081)
2 Each factor must have a positive 9 (.040) 27 .012
weight in the final system. This is aecom- 10 .131 28 (.058)
plished in the illustration by limiting the 11 -- 29 .035
value of the first degree of each factor (K1, 12 .043 30 .003
A1, J1, D1) to a value of at least .05. For 13 ---- 31 .032
14 ---- 32 .033
s T h a t is, m i n i m i z e ( d l -[- d2 -[- •. • q- clzs).
15 .104 33 .053
s T h a t is, m i n i m i z e ( N l d l -[- N2d2 --[- Nad3 -~- 16 .002 34
• . • q- N85das). 17 .027 35
lo The method of formulating appropriate con-
straints is described later. 18 .068

BUSINESS HORIZONS
APPLYING LINEAR PROGBAMIVIING TO YOUR PAY STRUCTURE

TABLE 6 example, the relation K1 ~ .05 is added;


four such constraints are in the program.
Evaluation Structure for Executive
Positions
3 A difference must exist between the
weights of successively higher degrees
Number of within a factor to ensure that each factor
Factor Factor Degrees degree has some distinguishable bearing on
A-Responsibility for Personnel 18 the final results. In this study, the minimum
B-Responsibility for Revenue 14 acceptable difference between weights for
C-Responsibility for External successive factor degrees was judged to be
Relations 15 .007. This is accomplished by adding a
relation such as K2 => K1 q- .007; a total of
D-Responsibility for Physical
Resources 14 thirty-five such constraints are used.
The net result of these forty-three con-
E-Planning, Problem Solving,
straints is to allow a range of possible an-
and Creativeness 10
swers that avoided a swamping effect of
F-Decision Making 11
any given factor.
G-Knowledge, Skills,
and Experience 10
Results of the L~nea~rProgram Table 4
illustrates the results of the program's ap-
plication to the thirty-five clerical positions,
showing weights assigned to each factor 95
TABLE 7
degree within the evaluation structure. The
Base Data on Twenty-one Executive evaluation provides a minimum sum of dis-
Positions crepancies between the salaries paid and
the factor ratings given. The magnitude of
Job Factor Degree Ratings salary discrepancy for individual positions,
Positions A B C D E F G Salary found by applying these weights to the
1 18 14 15 14 10 11 10 1.000 thirty-five positions, is identified in Table
2 17 13 14 13 9 9 9 .583 5. A discrepancy reflects the differences
3 16 11 13 12 9 10 9 .716 between the sum of the factor weights for
4 15 6 14 11 9 8 8 .583 any position (K + A + J ÷ D), as shown in
5 9 10 11 10 7 5 7 .333 Table 4, and the average salary for that
6 12 10 12 6 5 7 6 .450 position, as shown in Table 3. It can be seen
from Table 5 that salary reappraisal is re-
7 14 12 10 9 4 6 4 .375
quired for the positions numbered 2, 10, 15,
8 15 11 10 8 3 6 4 .350
and 20, which are out of line in terms of
9 6 4 12 10 5 6 6 .333 the salaries paid. If the highest existing
10 13 5 5 10 6 6 4 .333 salary is $10,000, the largest discrepancy is
11 10 9 9 6 2 6 4 .375 $1,600 (position 20), indicating the advis-
12 11 9 9 7 2 5 3 .350 ability of reviewing this job and perhaps
13 8 3 7 9 3 4 5 .333 of establishing a range in which the maxi-
14 10 8 8 5 1 5 2 .242 mum is lower than the average salary now
15 5 7 7 4 4 3 4 .308 being paid for it.
16 4 8 6 2 6 3 1 .233
17 3 4 7 1 6 3 2 .225
EXECUTIVE POSITIONS
18 7 2 2 8 2 4 2 .216
19 1 1 3 3 8 1 7 .250
20 5 1 4 7 2 3 1 .235 Sample Data Twenty-one executive posi-
21 2 2 1 2 1 2 2 .180 tions in a public utility company provided

WINTER, 1963
FREDERICKP. REHMUSAND Ha~avEy M. WACNER

the data used in the following example. The ing factor weights, but more than twenty
top position was that of the president and jobs should usually be included if a sample
chief executive officer; the lowest paid po- is used. Obviously, the degree of variance
sition was that of the assistant treasurer. in the raw data will influence sample size
The evaluation structure included seven and will be the major determinant of the
factors with varying numbers of degrees number of jobs to be included. There are
within factors, as shown in Table 6. Table practical limitations on the extension of the
7 shows the ratings for each of the twenty- sample and the number of positions that
one positions. Again, salaries have been ex- might be included. If the evaluation struc-
pressed as a fraction of the top salary-in ture is too complex, the sheer size of the
this case, the president's. problem could defy present computer ca-
pability. In reaching a decision on sample
Program Formulation, Constraints, and size, we believe that an individual with a
Results As in the previous example, a mathematical-statistical orientation should
linear equation exists for each of the twenty- assist salary analysts. The approach re-
one positions, and the program selects a set quires the insights of both points of view.
of weights that minimizes the sum of the
possible discrepancies.
SETTING PROGRA/V[ CONSTRAINTS
The numerical values for the constraints
placed on the executive position program
96 differed from those used in the administra- Once the factors and factor levels to be in-
tive and clerical program: cluded in an evaluation structure have been
1 No factor at its highest degree level selected, each of them should have some
may have a value greater than .25. bearing on the compensation decision ulti-
2 No factor at its lowest degree may mately reached. This requirement is car-
have a value less than .001. ried into the formulation of the linear pro-
3 The minimal difference between two gram by placing constraints on the permis-
successive factor levels must be .001. sible values of any factor degree. In the ex-
Table 8 shows the factor degree weights amples presented above, three kinds of con-
calculated by the program for the executive straints were used: (1) a maximum value
positions; Table 9 illustrates the discrepan- on the weight of the highest rating of each
cies between weights and existing compen- factor; (2) a minimum value on the weight
sation levels, which identify positions in of the lowest rating of each; and (3) a mini-
need of salary review (notably positions 2, mum value on the difference between the
5, and 11). If the highest existing salary is weights of successive factor levels. In com-
$50,000, the largest discrepancy is $1,550 bination, these constraints required all fac-
(position 5), indicating a current low level tors and degrees to, have some distinguish-
of compensation. able significance in the solution. For the
two examples, we found it appropriate to
use identical minimum and maximum con-
PREPARING FOR THE LINEAR PROGRAM straining values for each factor, but in other
applications experience and judgment may
DETERMINING SAMPLE SIZE dictate the use of differing values. For ex-
ample, it may be desirable to let one or two
Satisfactory results were found in the pre- factors account for most of the weight. The
ceding illustrations when only twenty-one maximum value for these factors would be
and thirty-five jobs were to be evaluated by accordingly large.
the program. It is not necessary to include The selection of values to be used in the
every position in the scheme for determin- constraints actually requires only a modi-

BUSINESS H O R I Z O N S
APPLYING LINEAR PItOGtlA~£MING T O YOUR P.4.Y $TRUCTLrltlg

TABLE 8 factor levels in the job evaluation structure


C o m p u t e r Factor Weights for Executive for which weights are being solved. 11 It is
E v a l u a t i o n Structure desirable to k e e p these constraints as mild
as possible; the p r o g r a m is thus given a
Factor Factors
Degree A B C D E F G b r o a d r a n g e in which to seek an optimal
solution.
1 .001 .001 .001 .001 .001 225 .114
2 .002 .002 .002 .012 .014 .046 .115
3 .003 .003 .003 .013 .015 .047 .180 REFINING THE DATA
4 .004 .004 .026 .014 .016 .048 .182
5 .005 .005 .027 .015 .017 .049 .183 Obviously, the results of the suggested job
6 .006 .006 .028 .025 .018 .051 .184 evaluation p r o g r a m cannot b e any better
7 .007 .007 .035 .026 .019 .138 .185 than the r a w data used in the analysis. The
8 .014 .008 .036 .027 .020 .139 .186 linear p r o g r a m m i n g a p p r o a c h does not assist
directly in the selection of factors and factor
9 .015 .025 .037 .028 .021 .140 .187
levels or in the establishment of ratings for
10 .016 .026 .038 .029 .022 .249 .188
particular jobs. But it can diagnose which
11 .017 .027 .039 .030 .250
factors are relatively most important, pro-
12 .018 .028 .040 .031 vided that the constraints on the weights are
13 .019 .029 .041 .032 weak. T h e linear p r o g r a m m i n g approach
14 .020 .212 .042 .033 should therefore b e viewed primarily as a 97
15 .157 .043 means of effectively converting r a w evalua-
16 .158 tion data into c o m m e n s u r a t e compensation.
17 .159 T h r o u g h o u t this article, existing salaries
18 .250 h a v e b e e n used in the job equations of the
linear program. I t is obvious that factor
weights cannot b e evaluated in a vacuum;
TABL~ 9 they m u s t be related to some kind of mean-
C o m p u t e r Discrepancies b e t w e e n Salaries ingful and systematic compensation data.
and E v a l u a t i o n Weights: O u r predilection for using existing salaries
Executive Positions is b a s e d on the premise that they m a y well
give the best indication of the relative value
Positions Discrepancy Positions Discrepancy
1 - - - 12
2 (.029) 13 205
u In the sample of administrative and clerical
3 ---- 14 positions, the four factors seemed to be of equal
4 -- 15 importance, but the weighting system was kept
5 (.031) 16 flexible by permitting the maximum weight of any
6 17 factor to be as large as .35. Similarly, in the
7 -- 18 sample of executive positions, the maximum allow-
able weight was set at .25, the smaller value re-
8 -- 19 suiting from an increase in the number of factors
9 20 to 7. The minimum allowable weight (.05 in the
10 21 first example and .001 in the second example) re-
11 .022 ---- flected a reasonable minimum salary contribution
for each of the factors. The difference between
successive factor levels (.007 in the first example
cum of m a n a g e m e n t discretion. The values and .001 in the second) was determined by taking
should allow sufficient flexibility so that the into account the smallest salary contribution that
would reflect a distinguishable difference between
results are both meaningful a n d acceptable factor levels, the range of salaries (from .535 to
in terms of the general restrictions implicit 1.000 in the first example, from .180 to 1.000 in
in the rating structure selected. T h e values the second) and the number of degree levels for
each of the factors (from 6 to 14 levels in the
are a function of the n u m b e r of factors and first example, from 10 to 18 levels in the second).

WINTER, 1963
FREDERICK P. REttMUS AND HAnV~ M. WAGNER

of jobs to an organization, and, consequent- several ways may nonetheless prove desir-
ly, the weights should be derived in terms able in certain situations. As has been men-
of existing salary relationships. Although tioned, gross inequities that are immediately
this approach would seem to treat some- apparent to the analyst through an examina-
what lightly the inequities that may exist tion of current salaries and position respon-
within an organization or in comparison to sibilities should be excluded from the
external salaries, two points must be made sample. If external relationships are im-
in defense of the method. portant, industry data or salaries developed
First, employees are typically more con- through special surveys can be used in place
cerned with internal than with external of actual salaries. We cannot stress too
salary relationships. The employee is in a strongly, however, the need to ensure com-
better position to evaluate his contribution parable job content if external salary in-
and his salary relative to others in the same formation is used. Should any adjustments
organization than to those with comparable of the foregoing types be made, the linear
positions in other organizations. Thus, in- programming approach suggested herein is
ternal salary relationships are of prime con- applied with the salary information ad-
cern, and these are best expressed in terms judged appropriate? 2 A manager must de-
of existing salaries. cide which data are relevant, no matter
Second, the linear program by its very what approach he adopts. Linear program-
nature has the ability to compensate for ming assists him only in analyzing the full
98 internal inequities that may exist within an implications of these data.
organization. Thus, ff several jobs are out
of line in terms of internal relationships, The desirability of the linear pro-
they will not significantly distort the solu- gramming approach has been argued on the
tion derived through linear programming. basis of its being objective and scientific.
The design of the linear program incorpo- Management judgment and experience have
rates an averaging effect that reduces the been used where they are proper and direct,
impact of individual jobs on the over-all but arbitrary data manipulation has been
solution. discarded, and the results are reproducible
Despite the apparent usefulness of exist- by any analyst employing the same data.
ing salaries, refining salary data in one of Although it is necessary to use a computer to
perform the arithmetic calculations, the
ready availability of linear programming
routines and the negligible amount of com-
puter time required make the approach
practical for any firm desiring to apply it.
As the two actual case examples have illus-
trated, the method of analysis is a powerful
tool for establishing effective salary sched-
ules.

12 In the event that management has, for one rea-


son or another, established bench-mark salaries
for specified jobs, these decisions cart easily be in-
corporated in the approach by eliminating the
possibility of a dlserepancy in the associated po-
sitions. Other management policies can be incorpo-
rated, but doing so, of course, weakens the claim
that the approach is an objective way to arrive at
a rational salary structure.

BUSINESS HORIZONS

Das könnte Ihnen auch gefallen