Jarrod W. Wilcox.

Accounting time series as evidence of university strategy and structure online

. (page 1 of 3)
Online LibraryJarrod W. WilcoxAccounting time series as evidence of university strategy and structure → online text (page 1 of 3)
Font size
QR-code for this ebook





Jarrod W. Wilcox
October, 1971


Jarrod W. Wilcox
October, 1971


The strategic management of large organizations is impeded by the
massive complexities to be analyzed. New approaches are needed to make
possible a sharper focus on the key variables. The Treasurer's Report
of the Massachusetts Institute of Technology offers a wealth of data
amenable to statistical analysis. This paper illustrates multivariate
methods for using these data to obtain evidence as to an organization's
structure and modes of response to environmental disturbances.

Because of the ready availability of financial data the study
emphasizes problems largely financial in character.


This paper was presented in the College of Measurements at the
1971 national T.I. M.S. Meeting held in Detroit. It is based on research
on the management of education underway within the Management Information
for Planning and Control Group at the Sloan School of Management, M.I.T.
Financial support was provided by the Ford Foundation.

- 1 -

Section A. Introduction

The purpose of this paper is to illustrate the application of explicit
statistical analysis to problems of analyzing organizational structure
and strategic response to environmental uncertainties. After a brief
discussion of relevant theory, an examination is made of a specific case,
the Massachusetts Institute of Technology during the period 1950-1970.
The results of factor analysis and cluster analysis of year-to-year changes
in M.I.T. financial figures are reported. Some implications of interest
to M.I.T. 's policy-makers are noted. Finally, the strengths and weaknesses
of the methods employed are considered in the light of the results of the

Administrative Use of Statistical Methods for Planning and Control

The increased feasibility made possible by computers for extensive
statistical analysis of data from the management Information systems of
large organizations has not yet been much exploited. Many routine intelli-
gence activities for more effective managerial planning and control could
be carried out using existing data. There exists the popular view that
statistical methods, indeed quantitative methods In general, are relatively
inapplicable to problems of designing changes in major policy and
organizational structure. Therefore, a major motivation of this study is
to find out more about to what extent these problems can usefully be attacked
through statistical analysis.

Section B. Background Theory

One of the most common drawbacks to the use of statistics in modeling
causal systems is the lack of any guiding framework to narrow the focus of

"""See Zannetos [1968] .

- 2 -
facts to be considered. In this section, the reader is informally introduced
to the theoretical viewpoint of the study. This introduction is necessarily
brief. However, it will serve to provide some delimiters on the direction
of the statistical analysis presented later.

Policies Underlying Partly Repetitive Decision-Making

The assumptions underlying managerial decisions are always based
on someone's observation of past experience. In some sense, therefore,
decisions are almost always repetitive in many aspects. Thus, it is useful
to analyze a series or stream of decisions based on the same underlying
assumptions or policy. When analyzing organizational decisions, an
additional complication must be kept in mind. Organizational decisions very
often represent a sequential aggregation of individual decisions.

In order to improve the effectiveness of these decisions, it is often

critical to determine the assumptions and aggregating mechanisms which

underly them. Unfortunately, in large organizations, no one person

may be aware of more than a few of the causal links involved. In such

cases, we may say the policies which underly the decisions are implicit,

or hidden. In general, the less explicit the policy, the more unrealized

potential there is for its improvement.

This potential is especially striking when decisions are embedded

in hidden feedback loops.

The value of statistical analysis of data from management information

systems in such cases is to direct managerial attention to as many such

causal links as possible. Thereby, implicit policies will be made more

explicit and thus susceptible to re-design.

^See Wilcox [1972]

See Forrester [1961]

- 3 -
The Costs of Uncertainty

We will take it as axiomatic that additional vincertainty induces
additional costs into the performance of any task.

In addition, one may observe empirically that the longer a situation
behaves as a stable process, the more opportunity one has to learn more
about it, and thus, on average, the less uncertainty. On average, though
not in each individual event, increased instability of the process implies
increased uncertainty for the manager.

Now some instability sometimes promotes the allocation of more resources
toward learning, and thus future lessening of uncertainty. The immediate
effect, however, is to increase uncertainty and thus costs.

It is also apparent that quantitative changes In the environment are
often not perceivable by managers as arising from a stable process. Thus
such changes are typically associated with increases in uncertainty. In
consequence, external changes may be taken as an indicator of increased
uncertainty and thus of potentially increased costs, at least in the
short run.

We can analyze changes in externally-related data recorded by the
management information system as indicators of uncertainty stresses
facing the organization. It may be extremely useful to locate in particular
cases the major sources of this uncertainty, so that it can be reduced in
the future.


One can view a smoothly-running organization as a large system of
local control loops dealing with localized problems. For the purposes of
this paper, we will ignore the problem of motivating people to assume roles
within this system. Coordination among these controls is achieved by
higher-order feedback control which monitors job completion.

- 4 -
resource allocation, destructive internal conflict, and so on. This control
authority structure is often represented as a hierarchical tree. We will
assert that in a mature, effective organization, the authority structure will
be well matched with the uncertainty structure imposed by the task environ-
ment. Let us see what is meant by this assertion and the assumptions which
underly it.

In a mature organization, the further near-term gains from learning
will be small compared to the gains from effective utilization of the
knowledge already gained. Thus, the dominant influence will be toward
efficiency rather than innovation. In such circumstances, it will be to
the organization's advantage to keep the aggregate organizational uncertainty


One vay to achieve lower aggregate uncertainty is to prevent the
propagation of local uncertainties to other areas of the organization. This
is done through divisionalization.

The propagation of local uncertainty to other divisions can be
reduced through buffering. It can often also be thwarted through coordinating
communication and control. Empirically, we know that usually some combina-
tion of these two approaches is used. In either case, there is an argument
for divisions with potentially strong interactions of local uncertainties to
be placed directly under a common supervisory authority. First, this will reduce
coordination costs. Second, such a superior is in the best position to
make iecisions on the appropriate buffers to reduce propagation of

See Emery [1969], Lawrence and Lorsch [1967], and Zannetos [1965a].
Remember that in order to keep our theory simple we are explicitly
disregarding the complications induced by motivational requirements.

Divisionalization thus promotes specialization, further local learning,
and thus further local efficiencies. See Thompson [1961].

uncertainties arising in one of the two divisions.

The extent of propagation of uncertainties among organizational
divisions may be partly measured by the covariances in changes in
divisional performance data from the management information
system. In this paper, we will not attempt to appraise the relative
quantitative importance of such covariances; thus we will analyze correlation
coefficients rather than covariances.

The pattern of covariance or correlation of uncertainties we will
term the organizational uncertainty structure. Our measure of the
uncertainty structure will be the pattern of correlation of year-to-year
changes in accounting data. These patterns of correlation will be analyzed
further through the multivariate statistical techniques of factor analysis


and cluster analysis.

In empirical work with this approach, it is important to remember,
however, that change is an imperfect measure of uncertainty. Clearly some
changes are both planned and highly predictable. Clearly, also, the
costs of uncertainty are not linear functions of the quantity of change,

even when change is unexpected. We use it as a measure only where no

better seems to be available.

In some cases this authority could be exercised through an interdivisional

■^See Zannetos [1965b]. He first proposed a similar measure for the effective-
ness of divisionalization used here. He suggested accounting variances,
rather than change, as the indicator of uncertainty.

^See Harman [1967] for a description of factor analysis. See Tryon [1970]
for a description of cluster analysis. Also, ordinary multivariate
regression or covariance analysis will be useful when dealing with only
a few variables .

Accounting variances from standard may be superior measures of uncertainty
in organizations with well-developed standard cost accounting systems.
Again, see Zannetos [1965b, 1968].

As previously stated, one would expect that in hierarchical organizations
good divisionalization would place highly correlated disturbances under the
immediate control of the same administrator. If not, there is the indication
of a possibility of the fruitfulness of organizational redesign.

Section C. Description of M.I.T. and the Available Data

The Massachusetts Institute of Technology is a private, non-sectarian
institution of higher education. Its undergraduate and graduate student
bodies each number approximately four thousand students. It has an inter-
national reputation for research, especially applied research. In the past
it has specialized in physical science and engineering. Its programs,
however, have been expanding in the humanities and social sciences. In
the period since World War II, it has managed several large off-campus
laboratories. During the last few years it has been characterized broadly
by development of new fields and departments, by a fair share of the
recent student unrest, and most recently by a financial strain brought on
by a decision to spin off one of the large research laboratories and by
cessation of rapid growth in government-funded research.

Its total annual budget, including that for the research laboratories,
exceeds two hundred million dollars.

The publicly available financial data in the M.I.T. Treasurer's
report for the year 1969 is approximately represented in Figure I . The
only major inaccuracies are the absences in the figure of changes in
unrealized profits and losses in asset holdings and of certain pension
plan operations.

Eg u i ^ i e s

fl AW).fVtU»loli»tq»>^C^

a net F*;wdi&
for Curr«ni


~%f>»yi%.»roJl Acs.

S fi€>v\%c^edl ;^«s.'


er*.i»^* tf

ft bxtle.

6riR-s ^5-



1 3

Online LibraryJarrod W. WilcoxAccounting time series as evidence of university strategy and structure → online text (page 1 of 3)