Jarrod W. Wilcox.

Accounting time series as evidence of university strategy and structure online

. (page 2 of 3)
Online LibraryJarrod W. WilcoxAccounting time series as evidence of university strategy and structure → online text (page 2 of 3)
Font size
QR-code for this ebook

financed in association with revenues from sponsored research which later
influence in a complex way the expansion of academic budgets. An additional
link might be through the financing of advanced graduate students and research-
oriented faculty. That is, increases in sponsored research off-campus
might conceivably lead to increases in enrollment of graduate students or
even the hiring or promotion of faculty and consequent potential increases
in later academic department expenditures. In any case, the evidence suggests
that immediate effects are not as important as sources of organization dis-
turbances as are possible longer-term effects.

Of course, a complementary explanation of the relationship between
Factors III and I which does not involve the propagation of uncertainty puts
both as consequences of a third outside influence, the post-Sputnik federal
government activity in support of technology. This influenced both increased
R&D and increased support to graduate education. The analysis of Registrar's
data regarding student enrollment for this period will enable us to better
judge the strength of this outside, "coincidental" relationship between Factor
III and Factor I.

Another possible outside third influence is the association of stock-
market-driven funds used for new building with sponsored research revenues.
It may be that for various reasons the Institute's portfolio gains are correlated
with increases in off -campus sponsored research.

Part of this lagged relationship was estimated at FI =.004 + (.57)FIII _ ,
R = .21, sig. at the 10 per cent level unadjusted for heteroscedasticity or
any non-normality. Of course, an alternative non-causal explanation is that
changes in both variables are caused by the same exogeneous variable.

- 19 -

Figure VI shows year-to-year changes in the number of graduate
students. For purposes of comparison, these data have been normalized by
subtracting their mean and dividing by their sample standard deviation.
Let us relate Figure IV and Figure VE. The latter shows dramatic fluctuations
of year-to-year changes in graduate students in the period 1953-57.
Figure IV, however, shows very small fluctuations during that period in the
Treasurer's Report Factor I, which was associated with academic departmental
expense. The fluctuations in change rates of graduate students during the
period 1961 through 1965 were of the same order of magnitude as before,
but were accompanied by very much larger changes in academic departmental
expense. This might be thought to reflect more intensive use of graduate
students in teaching assistant positions. However, a check showed that the
ratio of teaching assistants to graduate students has remained roughly
constant throughout the nineteen year observation period. Also, it is
doubtful that the ratio of teaching assistant salaries to faculty salaries
has increased. Thus, one is led toward the conclusion that the correspondence
suggested in Figure IV between large changes in Treasurer's Report Fill and FI
two or three years later is largely unexplainable in terms of coincidental
influxes of graduate students.

Although the evidence is not conclusive, it is hard to escape the
notion that the observed relationship between changes in off-campus
sponsored research activities and lagged changes in academic departmental
expenses is probably endogeneous to the organization and causal in nature.
That is, apparently the organization has responded to fluctuations in off-
campus research revenues or associated portfolio gains in such a way as to
propagate them in the on-campus activities several years later. The exact
mechanisms, however, are still undetermined.








- 21 -
The primary external sources of disturbance revealed by the data

of this section are:

1. variations is sponsored research revenue

2. variations in the stock market

3. variations in enrollment.

The simple correlations observed may be fruitful indicators to the
university administration of the propagation of these disturbances throughout
the structure of the organization.

Section E. Cluster Analysis of Treasurer's Report Data

The procedure used for cluster analysis of year-to-year changes was
McQuitty's "iterative, intercolumnar correlational analysis". This method
builds a hierarchical tree from the top down in which variables on nearby
branches are highly correlated. Such methods have been useful in establishing
initial taxonomies of classification in fields such as anthropology and
comparative biology. The particular procedure used struck the writer as
somewhat analogous to the normative procedure of divisionalization noted

earlier in Section B. At the least, it seemed to merit a trial. Unfortunately,

the statistical sampling properties of this procedure are not yet established.

Unless testable against other data, its result is therefore useful mainly

as a heuristic adjunct to factor analysis and regression. Thus, the following

discussion is presented in a tentative and experiu-.ental spirit.

"""^See McQuitty and Clark [1968]

This does not mean that its output cannot be tested against new data.
For example, one could test for a non-random relationship between two
trees by first rotating their branches for maximum congruence and then
calculating the Kendall tau coefficient of rank correlation. Ideally,
this method could be used to test the null hypothesis of no relationship
between an authority structure and an uncertainty structure.

- 22 -

A matrix of intercorrelations of the year-to-year changes in the
Treasurer's Report variables listed in Figure III was calculated. A new
matrix of the absolute values of these correlation coefficients was formed,
since it was not intended to distinguish between positive and negative rela-
tionships. This second matrix was analyzed for clusters using McQuitty's
iterative, intercolumnar correlational analysis.

This procedure operates approximately as follows. At the top level,
all variables are regarded as part of the same cluster. This cluster is
then fragmented into sub-clusters by recursively correlating the column data
of the original correlation matrix. This process converges to a matrix of
plus and minus ones. All the variables of each sub-cluster are inter-associated
by plus ones in the appropriate off-diagonal entries in the asymptotic matrix.
After the sub-clusters are formed, they are each in turn subdivided through
the same process. The procedure ends when all clusters have been fragmented

to single variables. The hierarchical tree thus obtained from the Treasurer's

Report data is presented in Figure VII.

The results are interesting. In very broad outline, the hierarchical

tree of Figure IX, which is intended to represent the uncertainty structure,

is reminiscent of the formal authority structure of M.I.T. Using the listings

of administrative officers and their titles in the 1962-63 M.I.T. catalogue,

one derives such an authority structure — see Figure VIII.

Without extensive reference to historical records and interviews, it is


A listing of the author's Fortran computer program for this procedure may

be obtained on request.


Parenthetically, of course, a further drawoack of the analysis is the lack

of inclusion of lagged relationships. The limited data did not appear to

justify investigating even the one-year lags, because this would have reduced

the number of observations to 18 and increased the number of variables to 64.

- 23 -

somewhat difficult to map responsibility for each of the Treasurer's Report
variables into the list of officers on Figure VII. Thus, the formal testing
of the hypothesis of a non-random relationship between the authority
structure and the uncertainty structure was left for a further study. However,
the hierarchical tree derived through cluster analysis of our measure of

uncertainty appears intuitively to have some correspondence with the

organizational authority structure. For example, sponsored research

revenues and expenditures are almost all in Cluster C22. Most of the funds

transactions associated with the treasurer are in Cluster C12. Instructional

expenses, student aid, and auxiliary activities. Including dormitory and

dining facility expenses, are together in Cluster Cll, th".s grouping the

academic departments and the so-called academic offices. An interesting

confirming result is the association in C22 of fees, services and other

receipts to funds with various sponsored research activities. Looking at

the organization chart, one finds that in 1963 the Vice President of

Research Administration (on campus) was also charged with the industrial

liaison office and the summer session, which give rise to these receipts.

Speculative Discussion of Cluster Anomalies

On the other hand, there are some curious exceptions to the overall
congruence of the cluster analysis tree with the organizational chart.
What may be the reasons? First, as the reader may note, at the higher levels
the writer vas not sure even of the authority structure. Second, some of
these exceptions no doubt reflect errors of measurement. Some, however, may
reflect genuine incongruity of the true authority structure with the true
uncertainty structure. This would imply potential disequilibria between the


Again, however, this appearance may be deceptive. We must wait for more
evidence before making a final judgment.

ln/«stm«n"fe Inoonoe
Auxiliary Ac+iv{f(«$

R«c«ipt5 for Sicdeni ^I'd
ScV\e)arsK;pi- pelUuJsklp AwApJls

AcdJemJc PepAftmewk
Reseo.rcVi Oret). 4- AolmiM. f xp,

A Actvo.y»ce«l r. Cor rent -P^r
Eel ocat iona I f I CLuf


( Se€ CO n^i rj t/a»f >b M

-Per cz;

Fio^vre VII

Clusfer At^glysi^

- 25

Exp. o? Ser/ice Activi*»es
/OHi«r Charges foFwi^

Gr«/> feral -♦ Mv*»rt>«tfTinttT»'
S. R. Pension^ £«rf ««*»-'

p;eeSy5er>'io«,*a Recei^$

Sp. ReS'Travtel

Qovt. Co»i«+^"f-erdeiOT.
l-abs ^

Seled^J XHfevdl€pAv4»He*\+ft| Leibs.

.. Provost

C, TtfiCMtf^

Figure VilL ^.I.T. Aclwi^^is^^r^x^|•|•


Online LibraryJarrod W. WilcoxAccounting time series as evidence of university strategy and structure → online text (page 2 of 3)