Peter Hammann.

Choice of the organization structure: a framework for quantitative analysis of industrial centralization/decentralization issues online

. (page 1 of 2)
Online LibraryPeter HammannChoice of the organization structure: a framework for quantitative analysis of industrial centralization/decentralization issues → online text (page 1 of 2)
Font size
QR-code for this ebook







Peter Hammann


December 1972



FEB 10 1973



Peter Hammann


December 1972

Thanks are due to Jean Rutherford for kindly and meticulously typing the manuscript.

MAR 7 1973



Quantitative analysis of centralization/decentralization problems has
so far mainly centered around the question of how decentralization might be
achieved without departing from the system optimum, involving decomposition
techniques of linear and nonlinear models. Positive effects, besides
division of labour, of an application of the decentralization principle to
a firm's organization were not taken into account. In order to do so, a
decision model for a hypothetical industrial organization has been developed,
leading to a comparison of the effectiveness of various more or less decen-
tralized organization structures. The problem appears to be one in capital
budgeting, where management is required to generate information of sub-unit
performance through a heuristic routine. System alternatives may be ranked
by one or more evaluation criteria, the one with highest rank, compared to
present achievement, indicating steps toward desirable organizational change
of the system. For simplicity's sake, the model, in a first stage, has been
conceived as static and deterministic.

This paper is partly based on the author's habilitation theses (14), submitted
to and accepted by the Department of Economics, University of Munich, July
1969 and February 1970, respectively.

Present research has been supported jointly by Technical University of Berlin
and the Ford Foundation.

I. What is the issue?

II. An overview of related work in Economics, Operations Research
and Management Science

III. Some definitions and statements

IV. The model

V. Generation of return functions

VI. Conclusions and outlook

The principle of task delegation or the division of labour is practically
as old as mankind. What value can one, then, contribute to a quantitative
analysis conducted on the question "is delegation really worthwhile"?, if
generally an affirmative answer has been supposed to be unavoidable? A
closer look at the problem, however, will reveal several ambiguous under-
lying assumptions and hazardous generalizations due to some rather one-sided
considerations of basic facts. Let us examine briefly the main pitfalls,
which led to conclusions that seem to have lived too long, considering the
lack of empirical evidence:

1. No clear distinction of tasks by their qualitative characteristics
has been made in this context, thus dealing in an identical fashion with
such different tasks as filling out an order form and the decision on the
order quantity. From the point of view of delegating task both these show
quite different implications due to quantitative dissimilarities and, there-
fore, different effects with regard to responsibility for their completion.

2. Delegation of task requires the evaluation of the various qualities
and capabilities of those person(s) with whom tasks will be shared. If
these persons are incompetent, task delegation might well prove to be
disastrous. Likewise, their ability to cooperate will enhance or reduce

the efficiency of the work of all concerned.

3. Not an isolated analysis of the necessity to delegate task at some
particular hierarchical level is called for. Problems of task delegation
can only be dealt with from the system's point of view in order to take
interdependencies of communication and information into account. The decision
to delegate task in some specific way depends on the ioint effort of all


individuals in the system. The quality of a single individual's performance
is irrelevant.

4. The various implications of delegation of management task, especially
decision-making, have been studied from the system's point of view. Howevei^,
analysis was conducted under the limiting assumption that the system optimum
under centralization may not be violated. Thus, positive effects of task
delegation, possibly leading to a higher system optimum were intentionally
neglected. Moreover, managpment - in this context - is assumed to have perfect
knowledge of the system's structure. Hence, delegation of decision-making
would seem to be quite unnecessary. This sadly limiting view may have its
fatal origins in the desire to give some economic interpretation to a well-
known algorithm for decomposing large optimization models into a set of
sub-models so as to solve them more swiftly and elegantly.

5. In many analyses, control over units to which tasks have been
delegated is supposed to be exerted by prices, and/or goals and quota setting.
Control by prices alone has been shown to be ineffective (7), not least be-
cause markets for commodities and services do not necessarily exist within a
firm. Whether this is so or not depends largely on the adopted organizational
structure itself. Especially, if decision-making is delegated, interdependen-
cies of sub-unit decision processes may generate "markets" in which outside
control (by central management) is required to arrive at decisions at all.
However, the question remains, whether direct control media such as prices,
goals or quotas will be accepted by the new decision-makers. Therefore, some
subtler indirect control methods or combinations of direct and indirect methods
seem to be called for which, alas, have not been investigated so far.

6. Motivation of individuals to whom tasks have been delegated is,
typically, taken to be granted or irrelevant, when rigorous control methods
are applied centrally. The latter may be true in some cases, but the case

itself, unfortunately, appears to be quite untypical. Motivational methods,
therefore, will have to be taken into consideration, though it should be
realized that - from the point of quantitative analysis - information on the
effects of variations in these methods are extremely scarce.

7. Centralization of task, in particular decision-making, is assumed
to generate economies of scale for a firm. This need not be true. It may
hold in cases of simultaneous central planning. In fact, it depends on the
completeness and accuracy of information at the central decision-maker's
hands. The second point is that some sort of centralized decision-making
is no issue at all for a particular firm. Therefore, economies of scale -

if any - will have to be achieved in other, more viable or appropriate organ-
ization structures.

8. The conviction that economies of scale are generally at hand in
centralized systems has led to another conviction: control must be exerted
in such a way that these same economies are guaranteed, even when some
(necessary) delegation of task takes place. Management, therefore, could be
tempted to fake decentralized decision procedures that, eventually, have no
value at all (see also (4) above). The procedures are time consuming and
costly, possibly offsetting other benefits available. Moreover, they are
too flimsy not to be understood by recipients of delegation who, in turn,
could try to cheat the central decision-maker. Therefore, it seems to be
quite improbable that any real-life organization would work along these lines,

Let us summarize. Traditional analysis has not centered around the
problem whether decentralization in some particular form, if ever, should
take place and if so, in which way. Attention has been focussed instead on
the narrow and irrelevant issue, how some quas i-decentralization could be
effected without violating the system optimum. As this is always known to
management, decentralization appears to be initiated only when some "demo-

cratic soothing" of staff is appropriate. The influences of information,
motivation and possible initiative of sub-unit decision-makers have been
completely ignored. The task, therefore, seems to be clear. The problem
should be restated as follows: Given that viable alternatives exist, which
organization structure (and hence the degree of decentralization) should
management choose if it wants to realize organizational change (see Bennis'
concept (6), pp. 59ff.) in a firm? Finding an efficient, by no means supposedly
optimal, organizational structure for a new firm may be a second task. However,
it should be tackled, once the somewhat less bothersome first problem has been
solved satisfactorily. We shall try to improve the mentioned defects of earlier
research by incorporating the notions of information, motivation and initiative
into the decision model. However, this model should only act as a framework
for more detailed research related to practical applications. At this stage,
we do not know enough about this many-sided problem so as to offer a thorough
solution .


For completeness' sake let us review very briefly relevant research don«
in recent years from which one should start in dealing with problems of this

The study that came nearest to the present one and (14) may well be the
one done by Thomas Marschak (28) . Tn this he tried to evaluate alternative
organizational structures by a time criterion for decision and communication
processes. Marschak' s view, thus, is a partial one. It is of considerable
interest as time considerations, in most real cases, will play an important
role when analyzing problems of this kind, typically involving more than one
decision criterion.

Decomposition theory is marked by such important papers as those of
Dantzig (10), Dantzig-Wolfe (11), Baumol-Fabian (5), Arrow (1), Arrow-
Hurwicz (2), Charnes-Cooper (8), Charnes-S tedry (9) and Koopmans (20), (21).
Extensions and discussions on various limitations of the Dantzig-Wolfe model
are given in Hass (16), Charnes-Clower-Kortanek (7), Geoffrion (13), Balas (4)
who developed a different algorithm, Kronsio (24) and Ruefli (35). In the
German literature H. Hax (17) has dealt extensively with the same problem.
Ruefli's paper is of considerable value as it offers a goal programming formu-
lation of the decomposition problem without recurring to a global objective
set for the system (which is the starting point in all other papers on
decomposition cited above). Moreover, his analysis emphasizes the hitherto
ignored fact that the solution from the model, reached after a Sequence of
setting goals and prices, depends on the structure of the organization.
Though the well-known implications of goal programming (especially the setting
of goal weights) limit Ruefli's approach, his results point into a direction
that further research might follow.

Kornai and Liptak (22), (23) have tackled decentralized decision-making
by the assumption that management or a central decision unit exert control
over sub-units in response to resource prices generated by the sub-units, as
might be the case in centrally planned economies. The solution is found by
reformulating the problem as a game.

The problem of optimizing a given organizational structure has been
Investigated by Hanssmann (15) as well as, more recently and in a more general
fashion, by Mueller-Hagedorn(31) , this being the lower-stage problem to the
one under review.

The interested reader is also referred to the ideas and suggestions of
J. Marschak(27), Radner (33) and McGuire (30) on the theory of teams for the
sake of considering cooperative behavior of individuals in an organization,
if required.

In this context, the system approach, at which this paper aims, has
been most forcefully stressed by Sengupta and Ackoff (36), revealing a
useful proposition on how to control a decentralized system.

Non-quantitative research has been conducted most extensively and the
literature on it is far too vast to be listed or commented on here. Some
particularly relevant findings will be cited later at appropriate instances,

Centralization and decentralization of task are not absolute, but relative
categories. If we are talking of "decentralization", we should bear in mind
that it is the degree of decentralization (or centralization") which an organi-
zation has attained or is supposed to arrive at after organizational change
has taken place, that matters. (See Albach (1), p. 112). Decentralization,
at least in this context, will be defined as the simultaneous delegation of
task and rights to decide on the task's best ways of accomplishment. (See
Koontz-O'Donnell (19), pp. 336 ff.). The primary question will not be how
decentralization may be effected but whether at all and to what degree it
should be realized. It is clear that there cannot be a normative prescription
of particular organizational structures for firms in general. The following
methodology aims at providing a framework for individual evaluations.

We shall not deal in great detail with the in terdependencies of organization
structures and information system as has been done by Johnson, Kast, and
Rosenzweig; (18? pp. 318ff.X Putnam, Barlow and Stilian (32, pp. llff.)or
Richards and Greenlaw (34, pp. 26 7ff-). However, there seems to be some
empirical evidence for the argument that an increase in quality and quantity
of information might entail a trend towards centralization of management task.


Let us review more closely the reasons claimed for some sort of decentral-
ization in industrial organizations, as defined above. It is because people
on lower hierarchical levels often stay in a closer contact to concrete decision
problems than management mostly does, that they might be given the rights to
decide on the task delegated to them in order to create:

- a greater sense of responsibility

- better motivation

- a more intense commitment to their work.

This implies a division of labor and responsibility and should result in

- hopefully, some initiative on behalf of the "new" decision-makers
(e.g., new production, marketing and investment alternatives)

- possibly, internal economies of scale.

They are facilitated through an increase in quantity and quality of in-
formation which is at the "new" decision-maker's disposal (viz. a new or -
more often - improved way of planning) and reflects the closer contact he has
to a particular task.

The main reasoning for an increase in centralization (or, equivalently ,
a decrease in decentralization), on the other hand, is the belief that the
quality of planning, coordination and control within the organization might
be brought about much more easily when a "central decision unit", whatever
that may be, takes over or remains in charge. Apart from that, one hopes
for economies of scale through more efficient resource allocation. These,
then, are all hypotheses, in both cases. What one should not do, however,
is to claim for an a-priori empirical and/or theoretical evidence for preferring
one of these (in fact, many) alternatives (see also, e.g., Ruefli (35), pp.

If one sets out to "compare" different organization structures and
their related decision processes, human behavior emerges as the crucial

and, possibly, most sensitive parameter of all. All human behavior, con-
sciously or subconsciously, appears to be motivated by the kind or degree
of identification of human beings with their environment. These facts are
far from new. (See Simon (37), p. XVI). However, they will have to be
considered very carefully in this context. In an industrial organization,
to give just one relevant example, members of the central decision unit
will try to motivate subordinate decision-makers quite deliberately. This
is done with respect to management goals and/or decision alternatives by
means of a variety of compensation schemes bound to a certain degree of
accomplishment. This sort of motivation is certainly perceived as such
by its recipients. (See March-Simon (26), pp. 34f. and pp. 52f). On the
other hand, a central managment or decision unit should not feel too confi-
dent about the outcome of this. Monetary and other incentives may be, but
need not be, taken up by the recipients or they iust work once and fail the
second time. There exists no such thing in an organization as built-in
conformity or joint motivation of its members that might be exploited to
the full when the need arises. (See, e.g., McGregor's theory Y (28), pp. 45f).
Besides these very deliberate forms of exerted and perceived motivation there
exists a variety of motivational influences which, typically, will elude a body
such as a central decision unit which might be called for to decide on viable
organizational alternatives. They are of a speculative nature and might best
be described by such homonyms as aspiration and ambition. E.g., the fact
that some particular task and rights to decide on it are delegated to a
person on a lower hierarchical level may constitute (and there is some
empirical evidence for that) some considerable motivational force though its
impact may, likewise, diminish again with time. Analyzing the question
whether and to what degree decentralization should take place in an organi-
zation will have to take such effects into consideration even if their


causes may remain unknown, nebulous or vague and even if there may be no
way to influence or control them.

Control, in this study, will be effected by budgetary measures. The
central management unit allocates a fixed budget to decentralized decision
units. Budgeting may not be the only way to control an organization. How-
ever, it is a typical one, which may be encountered in many industrial firms.
If we seek to compare structural alternatives we will have to ensure that a
comparison will be a reasonably "fair" one. Introducing a fixed budget for
all the firm's operations in any organizational form may be regarded as
sufficient for the moment. Thus, the problem of evaluating more or less
decentralized organization systems appears to be a problem in capital budget-
ing, the system alternatives being investment "portfolios" with uncertain
earnings streams. In these "portfolios", the decentralized decision-making
units, which themselves control a certain sub-system, represent the various
investment alternatives. From these remarks, then, it will be absolutely
clear that the problem at hand can only be solved from the system's point
of view.

Let us illustrate these ideas with the help of a simple model which
has been constructed to serve the purpose of demonstration of a methodology.

Let us assume that the firm manufactures m products in n different
production departments (mtn) . The products can be marketed and sold without
difficulties. Product markets are tangent, but not overlapping. The pre-
sent organization is such that separate departments for different product
groups have been formed, where all work is done. Thus, no competition arises


for jointly needed capacities. A total budget limits all managerial actions
within and outside the firm. Acquisition of raw materials and other produc-
tion factors shall be assumed as unconstrained. Its production will be sold
immediately, inventories will not build up. This assumption is a basic one,
reducing the problem to a static level. An entirely different model would
have to be developed in order to take dynamic aspects into account.

We are going to compare the following three organizational structures
for the decision process (see the discussion in (38), pp. 130f.):

1. Centralized decison-making by a central decision unit (simultaneous
planning; see Fig. 1).

2. Decentralized decision-making in functional decision units such as
marketing and production (successive planning, as production decisions
are assumed to depend on prior marketing decisions; see Fig. 2).

3. Decentralized decision-making in product divisions (simultaneous
planning; see Fig. 3).

A fourth alternative might be considered, too: decentralized decision-
making in product divisions with successive planning of marketing and produc-
tion decisions. We have left out this alternative somewhat arbitrarily
because coordination and control in organizations such as this seem to be
rather difficult to manage. However, the list of alternatives could be
augmented without violating the methodological principles which we set out
to demonstrate.

Alternatives 2 and 3 require a-priori management decisions on who is
supposed to serve as decision-maker in the decentralized units. That is
to say, the pertinent personnel assignments have to be effected - at least
on paper - before an analysis can take place. However, it would be possible
to check the problem under consideration with alternative assignment
schedules. As will be seen, this might prove to be very costly because





Fig. 1 : Centralized Organization Structure







Online LibraryPeter HammannChoice of the organization structure: a framework for quantitative analysis of industrial centralization/decentralization issues → online text (page 1 of 2)