Content uploaded by Joseph Galaskiewicz
Author content
All content in this area was uploaded by Joseph Galaskiewicz on May 09, 2016
Content may be subject to copyright.
Interorganization Contagion in Corporate Philanthropy
Author(s): Joseph Galaskiewicz and Ronald S. Burt
Source:
Administrative Science Quarterly,
Vol. 36, No. 1 (Mar., 1991), pp. 88-105
Published by: Sage Publications, Inc. on behalf of the Johnson Graduate School of Management,
Cornell University
Stable URL: http://www.jstor.org/stable/2393431 .
Accessed: 13/01/2014 20:15
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
.
Sage Publications, Inc. and Johnson Graduate School of Management, Cornell University are collaborating
with JSTOR to digitize, preserve and extend access to Administrative Science Quarterly.
http://www.jstor.org
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
Interorganization
Contagion
in Corporate
Philanthropy
Joseph Galaskiewicz
University
of Minnesota
Ronald S. Burt
Columbia
University
? 1991 by Cornell University.
000 1-8392/91/3601-0088/$1 .00.
0
Funding for this research was provided to
Professor Galaskiewicz by the Program for
Nonprofit Organizations at Yale University
and the National Science Foundation
(SES-8008570) and to Professor Burt by
Columbia University's Faculty Develop-
ment Program, the National Science
Foundation (SES-8208203), and consulting
revenues to the Research Program in
Structural Analysis. The discussion has
been improved in response to comments
from Roberto Fernandez, Mark Mizruchi,
Gerald Salancik, Thomas Schqtt, and the
ASO referees.
Two network contagion models are used to describe cor-
porate contributions officers' evaluations of nonprofit or-
ganizations seeking philanthropic donations. Contagion
by cohesion predicts that behavioral communication be-
tween contributions officers results in their sharing the
same evaluation. Contagion by structural equivalence
predicts that symbolic communication via role playing
between officers similarly positioned in the interorganiza-
tion network of contributions officers results in similar
evaluations. We find strong evidence of contagion, robust
over differences in the evaluated nonprofit organizations
and differences between officers. The evidence is over-
whelmingly of contagion by structural equivalence.'
Highlighting
the importance of institutionalizing
processes,
DiMaggio
and Powell (1983) brought
together several
streams of work to account for the similarities
found among
formal
organizations.
They offered the unifying
concept of an
organizational
field of firms.
They argued
that firms come to
resemble one another
and have their resemblance enhanced,
even enforced, by institutionalizing
processes within
and be-
tween organizations in the field. More specifically,
they dis-
cuss coercive processes that stem from political
influence
and the problem of legitimacy, mimetic processes that in-
volve standard
responses to uncertainty, and normative
pro-
cesses associated with professionalization
(DiMaggio
and
Powell, 1983: 150ff).
An important
component of their
argu-
ment is that the network of contacts among organizations
or
their
agents within
a field drives organizations
toward isomor-
phism (Moch
and Seashore, 1981; Zucker,
1987; Mizruchi,
1990, review related
work). The appeal of DiMaggio and
Powell's thesis lies in its focus on taken-for-granted
social
processes that explain
how and why organizations,
and the
texture of organizational
fields, change over time.
In
an effort to verify
empirically
DiMaggio
and Powell's thesis
of normative
control,
Galaskiewicz (1
985b) studied how di-
rect contact between corporate
philanthropic
officers in large
business firms had an impact
on how they viewed clientele
(prospective
donees). He found that the nonprofits
that were
recognized and evaluated
positively
by contributions
officers
in Minneapolis-St. Paul
received more money from corporate
donors, irrespective
of their size, activities,
and fund-raising
capacity
(Galaskiewicz,
1985a). It
was important
to under-
stand how these opinions
arose and survived.
For
reasons that will become obvious from our results, Gala-
skiewicz (1985b) reported
limited
support
for his hypothesis.
The limited
support resulted from the kind of mechanism
presumed to generate contagion. Debate in network
theory
has crystallized
around
two contagion mechanisms: cohesion
and structural
equivalence.
Adapted
in their
simplest form to
interorganization
contagion,
the concepts involve
an officer,
ego, expressing an evaluation in the context of evaluations
by
other officers as alters. Contagion
involves alter evaluations
influencing
ego's.
Contagion
by cohesion focuses on behavioral
communication
between ego and alter. In
the case of Galaskiewicz's contri-
butions officers, the more frequent
and empathic
the com-
munication
between ego and alter, the more likely
that ego
88/Administrative
Science Quarterly,
36 (1991): 88-105
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
Interorganization Contagion
and alter will share the same evaluations
of individual non-
profit
organizations (for
an extended review of the literature
on cohesion effects, see Burt, 1987: 1289-1290). Discussing
the activities of a nonprofit organization with other contribu-
tions officers, ego comes to a normative
understanding of the
organization-an understanding
that includes recognition of
the nonprofit and an evaluation of its contribution to the com-
munity, an understanding charged
with the interests of the
people with whom the nonprofit
has been discussed. Cohe-
sion is the mechanism traditionally
presumed to generate
contagion between people. Extending
this to contagion be-
tween people in separate organizations,
it is natural to find
cohesion implicit
in DiMaggio
and Powell's (1983) discussion
and explicit in Galaskiewicz's
(1985b) empirical
analysis.
The received tradition was challenged in the mid-1970s with
the rapid diffusion of structural
equivalence models opera-
tionalizing status and role sets in social structure as positions
in networks. Theoretical models of interpersonal
influence
grounded in structural
equivalence rather
than cohesion
quickly
followed to guide empirical
research on the manner
in
which the structural
equivalence concept of group defines
social pressures on subjective evaluations
(e.g., Burt,
1982:
chap. 5). Evidence
on interpersonal
contagion
favors struc-
tural
equivalence over cohesion (Burt
and Doreian,
1982;
Burt,
1982: chap. 6; 1987; Burt
and Uchiyama,
1989), as
does the emerging evidence on interorganizational
contagion
(e.g., Schmtt
and Morrissey,
1988; Mizruchi,
1989, 1990,
1991).
Contagion
by structural
equivalence
focuses on symbolic
rather than behavioral
communication.
Contagion arises from
role playing
among people who perform
similar
occupational
roles, as indicated
by their similar
positions in the network of
contributions officers. This includes, in the extreme, people
competing for the same job but applies more generally
to
people putting
themselves in one another's position
to eval-
uate the relative
adequacy
with which they are performing
their current job. The more similar
ego's and alter's work re-
lations
with others in their occupational
community-that is,
the more that one officer could substitute for another in the
network
of contributions officers-the more likely
that ego
will adopt beliefs or evaluations
perceived
to make alter more
attractive
as the object or source of professional
opinion,
in-
formation,
and relationships.
In the case of Galaskiewicz's
contributions
officers, coming to know how other officers
perceive the activities of a nonprofit
organization,
ego comes
to a normative
understanding
of the organization-an under-
standing
of the propriety
for a person in ego's position recog-
nizing
and endorsing
the organization,
an understanding ego
perceives to be shared by others in the same position,
an un-
derstanding
colored by ego's interest in advantage accruing
to
anyone performing ego's roles within the network of contri-
butions officers. To affirm
in-group
membership
and distance
themselves from the out-group,
individuals in structurally
similar
positions are expected to express similar
perceptions
and attitudes.
The frame of reference shifts from dyad to social system.
Regardless
of the frequency
or empathy of communication
between individual pairs
of officers, ego can be indifferent
to
89/ASQ, March 1991
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
the professional
evaluations
of officers in positions above,
below, or apart
from ego's own position in the social struc-
ture of his or her profession. Their evaluations
might make
ego nervous
about his or her own evaluation
in the sense that
they indicate
to ego that he or she will soon have to resolve
his or her own evaluation
of a nonprofit being discussed
among contributions professionals, but the primal force
shaping ego's evaluation
is the opinion of officers with whom
he or she is structurally equivalent.
These are the people who
indicate for ego the evaluation
proper for his or her position
in
the profession. As an opinion
comes to be shared within
ego's position,
ego is expected to follow rapidly to avoid the
embarrassment
of being the last to espouse a belief that has
become a recognized
feature of occupying
his or her position
in the contributions community
(cf., Burt,
1982: chap. 5;
1987: 1323).
Elements of this argument
can be seen in DiMaggio
and
Powell's discussion. They argued
that professional
networks
that span organizations
and diffuse information and attitudes
"create a pool of almost interchangeable
individuals who oc-
cupy similar
positions across a range of organizations
and
possess a similarity
of orientation
and disposition
that may
override variations
in tradition and control
that might other-
wise shape organizational
behavior"
(DiMaggio
and Powell,
1983: 152). These positions are defined by the "exchange of
information
among professionals [which]
helps contribute to
a commonly recognized
hierarchy
of status, of center and pe-
riphery,
that becomes a matrix for information
flows and per-
sonnel movement across organizations" (DiMaggio
and
Powell, 1983: 153). A professional
community
is thus a social
structure
of multiple positions, each with its own, internally
reproducing
beliefs and attitudes about professional
work.
The idea that there exist different
normative
systems or sub-
cultures within an organizational
field adds a new dimension
to institutional
analysis. Most theorists within this framework
recognize
that organizational
action is influenced
by the
norms or standards
within the organizational
environment or
standards
set by a professional
community (Zucker, 1987).
We extend this line of thought
and argue that different norms
and standards develop among clusters of agents within an
organizational
field and that these norms are co-variant
with
the informal stratification among organization
representatives.
Furthermore,
these norms can influence
how individual
agents come to view their environment.
While contagion
by
cohesion focuses on dyadic
influence
processes that operate
on a one-to-one basis; contagion by structural
equivalence
fo-
cuses on the informal
stratification
of agents in an organiza-
tional
field, the identification of strata within
this field, the
normative
standards
that mark these different
strata,
and the
influence of these strata standards
on the perceptions
and at-
titudes of agents.
METHOD
Our data describe the contact network
among corporate
con-
tributions
officers in the Minneapolis
and St. Paul
metropol-
itan area and officers' evaluations
of local nonprofit
organizations
eligible
to receive contributions.
The study pop-
ulation is described in detail by Galaskiewicz
(1985a, 1985b).
90/ASQ, March 1991
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
Interorganization
Contagion
In brief,
the population
is made up of publicly
held firms in the
metropolitan
area. The 67 firms contacted for the study were
publicly
owned, headquartered
in the Twin
Cities,
with 200 or
more employees, and made corporate
contributions
in 1979,
1980, or 1981 (see Galaskiewicz,
1985a: Appendix
A, for de-
tails).
The person most responsible
for each firm's charitable
activities
was interviewed.
We refer to this person, our re-
spondent, as the firm's
contributions
officer.
Respondents' job responsibilities
vary
between firms. Contri-
butions
are disbursed more often than not through
a com-
mittee composed of employees, with the contributions
officer
screening requests, executing grant
approval,
and handling
correspondence, payment procedures,
and record
keeping. In
larger
firms with more substantial
budgets, contributions
of-
ficers typically
work on long-range
planning,
formalize
proce-
dures, and initiate projects in the community (Troy,
1982: 3).
Galaskiewicz
(1985b: 642) argued
that contributions
officers
are especially important
as boundary spanners who reduce
the uncertainty
surrounding
contributions
by supplying
infor-
mation
on community
needs and nonprofit
organizations
to
the committee that ultimately
allocates a firm's charitable do-
nations.
The interviewed
officer in each of the 67 firms was presented
with a list of all publicly
held companies in the Twin Cities
area and asked to "check off those firms where you know
personally
individuals
involved in corporate
contributions,
i.e.,
on a first-name
basis, would feel comfortable calling
for lunch
or drinks after work, etc." Sixty-one
of the interviewed
of-
ficers completed the question. Their
responses define a (61,
61) sociometric choice matrix
among the largest firms. Four-
teen officers had no personal
contacts in other firms.
A few
were in touch with as many as half
of the other firms. The
average officer had personal
contact with contributions
per-
sonnel in 9.6 other firms.
Each contributions
officer was also presented with a list of
326 nonprofit
organizations
in the Twin Cities area. The 326
nonprofits
are a stratified
sample of nonprofit
organizations
in
the area, amounting
to about one in five of all eligible
organi-
zations (Galaskiewicz,
1985a: Appendix
C). In
response to the
list, officers put a check by the nonprofits
that they recog-
nized and a second check next to those they felt had
achieved extraordinary
accomplishments. Sixty of the 61 of-
ficers in the network
provided
evaluations
of the organiza-
tions, from which we have a three-category
response
variable: For
each nonprofit
organization,
each officer did not
recognize the organization
(0), or recognized
it, but not for
outstanding
projects (1), or recognized
it for outstanding
projects (2).
The evaluated nonprofit
organizations
varied over the whole
range of services provided
by such organizations,
but we
were most interested in highly
visible organizations,
given ex-
tremely different
evaluations
by the officers, that would be
difficult
to evaluate by concrete criteria.
The first two points
are important,
because we are trying
to explain
differences in
the officers' evaluations.
The third
point
is important
because
it is in these cases of having
to make an evaluation
in the ab-
sence of concrete criteria that social contagion is expected to
91/ASQ, March 1991
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
affect officers' evaluations. There was substantial variation in
evaluations
of the nonprofit organizations, but those over
which the officers most disagreed
all provided cultural, health,
or welfare services. These were also organizations widely
seen as providing outstanding
services. Of the fifteen organi-
zations over which officers most disagreed, four provided cul-
tural
services and the remainder all provided health and
welfare services.
Rather
than focus on the two or three organizations most
variably evaluated by the officers-and so have our results
potentially depend on the idiosyncratic qualities of recent
projects by individual nonprofit organizations-we estimated
contagion effects in the evaluations of the ten nonprofits
given the highest and most variable evaluations. The ten in-
clude all four of the cultural
nonprofits among the most vari-
ably
evaluated
organizations.
The remaining
six are health and
welfare nonprofits.
As a group,
the ten selected nonprofit or-
ganizations include,
from all 326, the organizations
with the
highest average officer evaluations and the organizations re-
ceiving the most variable evaluations. Across the ten organi-
zations, mean level of evaluation is uncorrelated with the
standard
deviation
of evaluations. The four nonprofit organiza-
tions providing cultural services are Twin Cities Public Televi-
sion, Minnesota Public
Radio,
the Minnesota Orchestral
Association,
and Film in the Cities. The six health and welfare
organizations vary in the kinds of services they provide:
the
Fairview
Community Hospitals,
the Harriet Tubman
Women's
Shelter, the Wilder
Foundation,
the Opportunity Workshop,
the Sabathani
Community Center,
and St. Mary's Rehabilita-
tion Center.
Detecting Contagion
We used network autocorrelations to detect contagion.
Brought
into sociology by network
analysts (e.g., Doreian,
1981; Dow, 1984), autocorrelation models have become a
standard vehicle for expressing contagion
effects (see Burt,
1987: Appendix,
for a quick review and leads into the litera-
ture).
The models describe interpersonal
influence in terms of
correlated
responses:
yJ= a + by,* + e,
where yj in our data is an evaluation
of a nonprofit organiza-
tion by officer
j and 6* is the typical
evaluation of the same
nonprofit by other officers who contribute to the social conta-
gion effect on officerj. Under
contagion by cohesion, the
other officers are those with whom officer
j has personal
contact. Under
contagion by equivalence, they are officer
j's
peers in the social structure
of contributions officers. The
typical response of these other officers as a generalized
alter
presumed to influence officer
j as ego is the weighted
average of their individual
evaluations:
y*= Wilyl + Wj2Y2 +
where the weights are fractions
indicating
the extent to
which each other officer plays a role in the social contagion
pressure on officer
j and summation is across all other of-
ficers in the network (0 < W.- < 1, Wjj = 0, and >2kw, = 1).
92/ASQ, March 1991
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
Interorganization Contagion
To estimate contagion by equivalence,
the w,, are measured
in terms of the extent to which officers i and
J are similarly
positioned in the social structure
of contributions officers. Be-
ginning with the raw sociometric choice data, each officer's
position in the social structure is defined by the pattern
of di-
rect and indirect relations
linking
the officer with each other
officer. Two officers are structurally equivalent
to the extent
that they have identical
relation patterns-they are directly
connected to the same other contributions
personnel;
through these connections they reach the same other firms
with which they do not have direct
contact; and they are
themselves directly
and indirectly
cited by contributions of-
ficers in the same other firms. The network weight wj, mea-
suring equivalence
varies from 0 to 1 with the extent to which
officers i and
j have identical
patterns
of direct
and indirect
relations with contributions personnel in the 61 firms under
study.
The network weight is defined as follows: The Euclidean
dis-
tance, dii, measuring
the structural
equivalence
of officers i
and
j was computed in the usual way from path distances
derived from the raw sociometric choice data (e.g., see Burt,
1988, for review). The network
weight wj.
was then com-
puted as the relative
strength of officer
j s felt equivalence
to
officer i:
wij
= (proximityj
to )v/Ik(proximity
j to k)v,
k =, j,
where j's proximity to i reverses Euclidean distance to mea-
sure structural
equivalence rather than distance (maximum
distance from
J minus dj and self-relations are set to 0 so that
the w--
weights measure the relative
contribution of each
other officer i/to the contagion
effect on officerj (Burt,
1982:
176-177; 1987: 1296-1297). A search through alternative
values of the power exponent v with a tolerance of .01 in se-
quential contagion
correlations settled on a value of 2. The
contagion
correlation between officer and alter evaluations
(yj
and y.*) across officers and the ten nonprofit organizations
is
.527 with v set to 1, .535 with v set to 2, and .534 with v set
to 3. We also reanalyzed
the data for contagion
effects using
popular
variations
on the raw Euclidean
distance measure of
structural
equivalence.
We computed Euclidean distances di-
rectly
from the raw binary
sociometric choice data, which ig-
nores the indirect relations in terms of which officers might
be nonequivalent.
This reduces the aggregate contagion
cor-
relation
only slightly,
from .54 to .48. We computed Euclidean
distances from z-score measures of relations,
as in CONCOR,
ignoring
mean and variation differences between officer rela-
tion patterns.
This reduces the aggregate contagion
correla-
tion from .535 to .345. These results with alternative
measures are not intended
to illustrate a general preference
for the one selected here, but only to show that the measure
selected is well suited to this study population.
Given the network
weights defining
each officer's structurally
equivalent peers, the alter evaluation
y1*
can be computed
and correlated
with the corresponding
observed evaluation,
yi, to test for equivalence contagion.
Correlations were com-
puted across 60 officers evaluating
the ten nonprofit organi-
zations, and jackknife-test
statistics were used to evaluate
the
significance of effects.1
93/ASQ, March 1991
I
For reasons of inefficiency and inconsis-
tency, ordinary-least-squares estimates of
a network autocorrelation are not max-
imum likelihood. Inconsistency is not as
great a problem here as it is in general,
because our data describe the population
of firms in a case study. To get around the
inefficiency of ordinary-least-squares stan-
dard errors, we used jackknife statistics
based on Fisher z-transformations of the
network autocorrelations to provide a
sense of the variability
of contagion ef-
fects within the study population.
Mosteller and Tukey (1968: 133-160)
provided an introductory discussion, Efron
(1982) provided a more theoretical discus-
sion, and Hinkley (1978) and Miller (1974)
discussed jackknife statistics for correla-
tions in particular.
The gist of the infer-
ence is to create a sampling distribution
around an estimated contagion effect by
re-estimating the effect from subsamples
of the study population. Let b be the ordi-
nary-least-squares estimate across all 60
officers of the contagion effect in the
equation displayed in the text predicting
officer j's evaluation, yp,
from alter evalua-
tion yj*. Let bj be the same effect esti-
mated from the 59 officers excluding
officer j-that is to say, excluding j as a
respondent on the left side of the equa-
tion and excluding j as an alter in the net-
works of all other officers on the right side
of the equation. The b, estimate is the
contagion effect that would have been
observed in the study population if officer
j had not been interviewed, an estimate
independent of j's evaluations. By re-
peating this for each officer and weighting
differences between the by
and b (in a
",:pseudovalue"),
a sampling distribution of
estimates for the contagion effect can be
constructed where the mean and standard
deviation of the distribution can be used
to make statistical inferences about the
magnitude of the effect. The standard
error for an effect is small to the extent
that very similar estimates are obtained
across subsamples of the study popula-
tion.
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
RESULTS
The results show weak evidence of contagion
by cohesion
and strong evidence of contagion
by structural
equivalence.
For equivalence contagion,
the correlation
was .535, indi-
cating that 28.6 percent of the variation
in evaluations
across
officers and nonprofits
can be described
with the evaluations
of their
peers in the social structure
of contributions
officers.
To hold constant differences between the nonprofits,
we
computed partial correlations
between y and y*, holding
con-
stant nine dummy
variables
distinguishing
the ten nonprofits.
Net of differences between the evaluated
organizations,
the
correlation was still
a high .398, and the jackknife
t-test sta-
tistic showed a very low probability
of getting a contagion
ef-
fect this strong by chance (8.16 t-test with 59 d.f., giving
the
null
hypothesis a probability
well below .001).
The same is not true of contagion by cohesion. The correla-
tion for cohesion contagion
was .290, indicating
that only 8.4
percent of the observed variation
in evaluations
across of-
ficers and nonprofits
can be described
with the evaluations
of
contacts in other firms. Net of differences between the eval-
uated nonprofit
organizations,
the correlation
for cohesion
was .120 and the jackknife
t-test fails to reject
the null hy-
pothesis (1.89 t-test with 38 d.f., p = .07).
We searched through
alternative
cohesion measures for
stronger
evidence To estimate contagion
by cohesion, the wj,
are measured in terms of officer
j's personal
contact with
other officers. We began with the usual practice
of using raw
sociometric choices as a measure of cohesion (e.g., see
Coleman, Katz,
and Menzel, 1966; Burt,
1987). The network
weight wj, is one over the number
of cited contacts for each
other firm
i in which officerj has personal
contacts. The alter
evaluation
y.* is the average evaluation
of officers in the firms
that
j cites. l[he correlation
between this and the observed
officers' evaluations
is .218-not much worse, but certainly
no better, than the .290 correlation
for cohesion. Extending
the measure to include
indirect
connections between firms
did not yield stronger
evidence of contagion.
We then re-
stricted the model to only the strongest of relationships.
Net-
work autocorrelation
evidence of contagion
by cohesion can
be expected to increase with the strength of the cohesive re-
lationship (e.g., Burt and Uchiyama,
1989). We obtained
the
results for contagion,
above, with cohesion restricted to recip-
rocated
choices, following
Galaskiewicz
(1985b).
The network
weight w--
measuring
cohesion varies from 0 to 1, with non-
zero values indicating
that officers
j and i acknowledge per-
sonal contacts with contributions
officers in one another's
firms. Here, too, we considered indirect
contacts through
in-
termediary
organizations
(path
distances varied from 1 to 5
steps in length)
but again obtained
weaker evidence of conta-
gion.2
We were suspicious of the sample size change between co-
hesion and structural
equivalence. Structural
equivalence
doesn't depend on contacts between officers influencing
one
another. Alter
evaluations
can be computed for all officers.
The contagion by equivalence effect is estimated across 60
officers evaluating
10 nonprofits,
a total of 600 evaluations.
Cohesion depends on contact between officers influencing
94/ASQ, March 1991
2
The network
weights for cohesion are de-
fined
by the same perceptual
model
de-
scribed above to define equivalence
weights, but "proximity
j to i" is mea-
sured by the strength
of officer
j's per-
sonal relation
to the contributions
personnel
in officer
i's firm.
In the conta-
gion
analysis,
proximity
is 1 if
j and
i cite
one another,
0 otherwise.
The v exponent
does not change
the strength
of the con-
tagion
effects and is left at a default
value
of 1.
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
Interorganization
Contagion
one another.
Alter evaluations
cannot be computed for of-
ficers who have no contacts. Fourteen
officers had no per-
sonal contact with contributions officers in other firms and
another
seven had no reciprocal contact with other firms.
Therefore,
the usual operationalization of cohesion using raw
sociometric choices yields alter evaluations for only 46 of-
ficers evaluating 10 nonprofits (a total of 460 evaluations),
and
our reciprocated-choice operationalization
yields alter evalua-
tions for only 39 officers evaluating
10 nonprofits. A possible
explanation
for the weaker evidence of contagion by cohesion
is the fact that it is estimated from a subsample of officers.
Further
analysis shows that this doesn't account for the
stronger
evidence of contagion by structural
equivalence, but
it is a point
to be noted. The correlation for contagion by
structural
equivalence is .433 if limited to the 39 officers in-
volved in reciprocated
contacts and .455 if limited
to the 46
officers who have personal
contact with contributions
per-
sonnel in at least one other firm. These effects are weaker
than the effect given above, estimated across all officers, but
still quite significant
(with respective jackknife
t-tests of 3.24
and 5.12, holding
constant differences between the evaluated
nonprofit
organizations).
Differences between Evaluated Nonprofits
Evaluations
vary significantly
between nonprofit
organizations,
which provide different kinds
of services. For
this, or their ef-
fectiveness in delivering
whatever services they provide,
they
vary
in appearance
to the officers. Some of the organizations
are widely recognized
for outstanding projects.
Others are the
object of more narrowly
distributed acclaim.
Without
defining
specific qualities
of the nonprofits
that might
affect evalua-
tions, we can get a sense of how much differences between
them matter
by regressing the evaluations across dummy
variables that distinguish every nonprofit
from every other.
With nine dummy variables
distinguishing
the 10 evaluated
nonprofit
organizations,
we can describe 15.6 percent of the
variation
in the 600 evaluations
(multiple
correlation of .394).
This strong association raises the possibility
that contagion
might vary
significantly, depending on the specific nonprofit
organization
under evaluation.
Differences between the evaluated
organizations
are quickly
apparent
from the results in Table 1. Contagion by structural
equivalence is not uniformly
successful. It
generates nega-
tive, although
negligible
network
autocorrelations in the eval-
uations of two organizations:
the Minnesota Orchestral
Association and the Fairview
Community Hospitals. Contagion
by cohesion is not uniformly poor. It
generates statistically
significant
network autocorrelations
in the evaluations of two
organizations:
Twin Cities Public Television and Fairview
Community
Hospitals.
We know from the aggregate analysis that there is evidence
of contagion
even after holding
constant these differences
between the evaluated organizations.
The question remains
of accounting
for the different effects in Table 1. The vari-
ables we have to work with are the kind of services provided
by an evaluated nonprofit,
the extent to which officers gener-
ally share a positive opinion about the nonprofit, and the ex-
tent to which officers differ in their evaluations. We tested
95/ASQ, March
1991
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
Table 1
Contagion in Evaluating Each Nonprofit*
Structural
Evaluated nonprofit Mean S.D. equivalence Cohesion
Cultural
Twin Cities Public Television 1.41 .75 .106 .502
(1.3) (2.6)
Minnesota
Orchestral
Association 1
.28 .68 - .287 - .183
(-0.3) (-0.5)
Minnesota
Public Radio 1.57 .66 .477 .169
(3.4) (1.1)
Film in the Cities .61 .63 .654 .351
(6.0) (1.8)
Health
and Welfare
St. Mary's
Rehabilitation
Center .84 .83 .312 .035
(2.6) (1.0)
Harriet
Tubman Women's Shelter .74 .81 .669 .264
(4.8) (1.8)
Opportunity Workshop 1.23 .78 .151 -.108
(1.5) (-0.2)
Wilder Foundation .92 .75 .402 -.070
(2.7) (-0.6)
Sabathani
Community
Center .75 .71 .630 .175
(6.1) (1.6)
Fairview Community Hospitals 1.10 .67 -.256 .256
(- 0.3) (2.5)
* The mean and S.D. describe the average and standard deviation
of evalua-
tions given by the 60 contributions officers to each nonprofit organization.
Ordinary-least-squares
estimates of the network autocorrelations are pre-
sented for contagion by structural
equivalence
and cohesion. Jackknife
t-
tests are given in parentheses and have 59 degrees of freedom for conta-
gion by equivalence,
38 for contagion by cohesion.
these variables
for slope adjustments
to the aggregate conta-
gion effects reported above.
There are no significant adjustments
to the effects of conta-
gion by structural
equivalence.
The strongest is a slight ten-
dency for contagion
to weaken as a nonprofit
becomes more
widely admired
(- 1.95 t-test for the slope adjustment
with
mean level of evaluation). Similarly, slope adjustments
for the
mean and standard
deviation of evaluations do not reveal
stronger
evidence of contagion by cohesion. However,
the
evidence of contagion by cohesion is clearly stronger
for non-
profits providing
cultural services. In Table 1, the network
autocorrelations
and test statistics for cohesion are higher
among the four culture
nonprofits
than among the six health
and welfare nonprofits.
The slope adjustment
for the higher
contagion
effect among culture
nonprofits generates a routine
t-test of 3.5. The jackknife
t-test for cohesion contagion
in the
evaluations of the culture
nonprofits
is 2.5 (with
three dummy
variables, holding
constant differences between the four cul-
ture nonprofits;
38 d.f., p = .01) versus a negligible
1.2 for
cohesion contagion
in the evaluations of the health and wel-
fare nonprofits
(with
five dummy variables, holding
constant
differences between the six health
and welfare nonprofits).
It is tempting to speculate about why cohesion should gen-
erate contagion in officers' evaluations
of culture
nonprofits
rather than health
and welfare nonprofits.
This could turn
out
to be fruitful,
but we caution
against expecting too much. The
stronger evidence of contagion
by cohesion is still much
96/ASQ, March 1991
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
Interorganization Contagion
weaker than the corresponding evidence of contagion by
structural equivalence. If an officer's evaluation of a nonprofit
providing cultural services is regressed simultaneously over
the evaluation predicted by structural equivalence and the
evaluation predicted by cohesion, the routine t-test for struc-
tural equivalence is 4.0 versus 1.9 for cohesion. Holding con-
stant three dummy variables distinguishing the four culture
nonprofits to get a more precise estimate of contagion, the
partial
correlation
providing
evidence of contagion by struc-
tural equivalence generates a 5.0 jackknife t-test (59 d.f., p <
.001), versus the above-mentioned
2.5 t-test for cohesion.
The evidence of contagion in officers' evaluations is thus not
uniformly distributed across what might seem to be substitu-
table objects of evaluation. Future research on the mecha-
nisms of interorganization contagion
should include
multiple
criterion variables to compensate for idiosyncratic
factors in
the evaluations of any one stimulus in an organization's envi-
ronment. Focusing on the characteristic rather than the idio-
syncratic, our more central
conclusion concerns the relative
stability of contagion effects predicted by structural
equiva-
lence versus cohesion. Across the evaluated nonprofit organi-
zations showing variable evidence of contagion,
structural
equivalence effects are stronger in the aggregate and signifi-
cant in more evaluations of individual
nonprofits.
Differences between the 'Officers
Evaluations
vary
significantly
between officers making
the
evaluations.
The officers come from different
industries,
dif-
ferent firms, and have different roles in their firms and dif-
ferent personal backgrounds.
Without
defining specific
qualities of the officers that might
affect evaluations,
we can
get a sense of how much differences between them matter
by regressing the evaluations
across dummy variables
that
distinguish
each officer from every other. With 59 dummy
variables
distinguishing
the 60 officers, we can describe 28.9
percent of the variation
in evaluations
(multiple
correlation of
.537), just about the same amount of variation
described by
the contagion effects in the first analysis. Again,
this strong
association raises the possibility
that the contagion
effects
may vary
significantly depending
on the kind of officer making
the evaluation.
We focused on personal
and professional
differences be-
tween the officers as sources of evaluations. With respect to
personal differences, we studied gender, education,
and
birthplace (Twin Cities, north central
region
of the country,
or
elsewhere). With respect to professional differences, we
studied human service experience, membership
in local pro-
fessional associations (officers in one or more professional
associations were members in at least one of the following:
Minnesota
Council
on Foundations,
the Business Action Re-
source Council,
and Women in Foundations/Corporate
Philan-
thropy),
whether their
job was as a semi-professional
or
professional contributions officer, and their prominence
in the
network of Twin Cities contributions
officers. Low-promi-
nence officers were employed by a firm
in which fewer than
four of the respondents cited personal
contact with contribu-
tions personnel. High-prominence
officers were employed by
a firm in which 14 or more of the respondents cited personal
97/ASQ, March 1991
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
contact with contributions
personnel.
Average-prominence
officers are everyone between these extremes. These factors
define 12 variables
that together describe 18.5 percent of the
variance
in officers' evaluations.
Three
alone describe 17.3
percent of the variance:
gender, membership
in at least one
of the local professional
associations, and network
promi-
nence. These three officer variables
capture
only a portion of
the description
provided
by the 59 dummy
variables distin-
guishing each officer from every other, but it is a large por-
tion, captured
with few variables.
The first point
to establish is that the differences between of-
ficers do not explain
away the observed contagion effect in
their evaluations.
The bottom two rows of Table
2 show con-
tinuing
evidence of contagion, at least by structural
equiva-
lence: The contagion
effect remains
significant
after holding
constant differences between the evaluated
nonprofits and
gender, membership,
and network-prominence
differences
between the officers (4.42 jackknife
t-test, 59 d.f., p < .001).
A more stringent
test is to test the partial
autocorrelation
among officers' evaluations,
holding constant nine dummy
variables
distinguishing each evaluated nonprofit
organization
and 59 dummy variables
distinguishing
each officer. Even
with this exaggerated control,
in which all covariation
with
differences between officers and evaluated nonprofits
is re-
moved from the contagion
effect, there is significant evi-
dence of contagion by structural
equivalence (3.52 jackknife
t-test, 59 d.f., p < .001).
That
point established, it is clear that much of the zero-order
contagion
effect can be attributed
to differences between of-
ficers. The contagion variance that remains
after our controls,
although
statistically
significant,
is much reduced-from the
Table 2
Contagion and Officer Differences*
Structural
Officer characteristics Mean S.D. equivalence Cohesion
Men (50 officers) 1.01 .79 .543 .291
Women (0
lofficers) 1.34 .73 .547 .283
No professional associations
(36 officers) .82 .80 .525 .293
One or more associations
(24 officers) 1.43 .61 .267 .254
Low prominence (24 officers) .73 .81 .570 .395
Average prominence (16 officers) 1.14 .74 .380 .223
High prominence (20 officers) 1.40 .63 .244 .211
Correlation net of above
differences and differences
between evaluated nonprofits .184 .073
Jackknife-test statistic 4.42 1.13
* The mean and S.D. describe the average and standard deviation of evalua-
tions made by the officers in each row category (each officer evaluates 10
nonprofit organizations). Ordinary-least-squares estimates of the network
autocorrelations are presented for contagion by structural equivalence and
cohesion. The two partial correlations are the correlations between officer
evaluations removing any variance attributable to sex, membership in pro-
fessional associations, network prominence, and nine dummy variables dis-
tinguishing the 10 evaluated nonprofit organizations. The jackknife t-tests
have 59 degrees of freedom for structural equivalence, 38 for cohesion.
98/ASQ, March 1991
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
Interorganization Contagion
.535 correlation in the first analysis to the partial correlation of
.184 at the bottom of Table 2.
The first two rows in Table 2 show evaluation and contagion
differences between men and women. Evaluations by
women were more positive than those provided by men, but
otherwise there were no gender differences. Variation in
evaluations by men and women were about the same and,
most importantly, contagion had almost identical effects on
the evaluations of men and women. Gender
was not a factor
distorting contagion.
The other two variables in Table 2 are more interesting.
Of-
ficers socially integrated into the society of other officers
gave higher evaluations to the nonprofits and showed less
evidence of contagion in their evaluations. These findings
bring three new factors into the analysis: social integration,
a
positive correlation between integration
and evaluation,
and a
negative correlation between integration and contagion.
Social Integration
A structural equivalence analysis of the sociometric choice
data shows that Twin Cities contributions officers live in a
center-periphery social structure. There is a single position at
the center of the system occupied by prestigious officers
employed by the largest firms, a satellite position
of officers
employed by small firms who claim personal
relations with
officers in the center position,
and a hierarchy
of four
other
positions variably
interconnected
in descending prominence
under the center position. In
such a system, row and column
marginals of the network describe much of each person's po-
sition. In the Twin Cities, the number of sociometric choices
an officer "sends" is a good indicator of the number he "re-
ceives" (.73 correlation),
and eigenvector measures of officer
prestige and centrality
are highly
correlated
(.889; see Knoke
and Burt, 1983, for review).
An officer has prestige if widely
cited by officers who are in turn
cited by prestigious
officers.
An officer is central to the extent that he or she has recipro-
cated contact with many officers who are similarly
connected
to other central officers. Our
distinction
in Table 2 between
low-, average-,
and high-prominence
officers is robust
largely
because the study population
has a simple center-periphery
structure. We made the distinctions based on graphs
of con-
tagion effects at each level of alternative
prominence
mea-
sures. The simple three-category prominence
distinction
is
correlated .91 with the eigenvector measure of officer pres-
tige and .83 with the eigenvector measure of officer
centrality.
The three prominence categories are linked with member-
ships in professional
associations. Prominence
and member-
ships each have a direct positive effect on officers'
evaluations
(respective t-tests of 2.6 and 3.3 with 57 d.f.), and
there is no interaction effect above and beyond the direct ef-
fects (-0.6 t-test); however, the two variables both tap into
integration
differences between the officers. Of the 24 low-
prominence officers, no one was a member in any one of the
three area professional
associations. Of the 20 high-promi-
nence officers, 17 are members in at least one of the associ-
ations, and seven are members in two. The
average-prominence
officers were about evenly divided
be-
tween members and nonmembers
of the associations.
99/ASQ, March 1991
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
The officers thus vary in social integration
between two ex-
tremes: At the center of the system are officers who partici-
pate in one or more of the area professional
associations and
work for a large
firm with which many local contributions of-
ficers have personal contact. On the periphery of the system
are officers who have no affiliation
with the area professional
associations and little or no personal contact with other con-
tributions
officers.
Integration and Evaluation
Turning to the positive and statistically significant
correlation
between integration
and evaluation,
the means in Table 2
show that higher
evaluations were made by officers more in-
tegrated into the society of other officers. More graphically,
the average evaluation doubled between the periphery and
center of the system. The 24 low-prominence
officers outside
the area professional
associations gave an average .73 evalu-
ation to the 10 evaluated nonprofit organizations.
This in-
creases to 1.43 for the 17 high-prominence
officers who were
members in one or more of the area professional
associations.
The-substance of the association remains hidden
in these re-
sults. Revealing
it will inform our description
of contagion's
decline with increasing
integration.
The three-category
evalu-
ation of a nonprofit distinguishes recognized
nonprofits
from
the invisible,
and outstanding nonprofits
from the merely rec-
ognized. Table
3 contains the frequencies with which officers
at each level of prominence
gave each kind
of evaluation to
one of the 10 nonprofit organizations.
The numbers in paren-
theses are z-score test statistics from a loglinear
model of the
frequencies. A positive
z-score indicates a frequency larger
than would be expected if evaluations
were independent
of
prominence.
We are not concerned with the absolute magni-
tudes of these statistics so much as with their relative
mag-
nitude.
The point
we wish to highlight
is how the effects change
across the rows and columns of the table. Reading
down the
columns, there is a linear
transition from low evaluations
by
low-prominence
officers to high evaluations
by high-promi-
nence officers. For
example, the third column shows that
low-prominence
officers are unlikely
to give an outstanding
rating,
officers of average prominence give outstanding
ratings
about as often as would be expected in the absence
of an association between prominence
and evaluation,
and
Table 3
Prominence and Evaluation*
Evaluation of nonprofit organization
Officer
prominence Unrecognized Recognized Outstanding
Low 119 67 54
(8.97) (-4.33) (-5.32)
Average 34 69 57
(-0.36) (0.64) (-0.23)
High 16 89 95
(- 6.24) (3.39) (5.32)
* Ratings given by all 60 officers to the 10 evaluated nonprofit organizations
are tabulated, with loglinear
z-score tests for interactions given in paren-
theses.
1 00/ASO, March 1991
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
Interorganization Contagion
high-prominence officers are likely to rate nonprofits as out-
standing.
The rows of Table
3 show a more discontinuous
pattern, switching between response modes at the point of
recognizing a nonprofit
organization. In the first row, for ex-
ample, low-prominence officers are likely
to say they do not
recognize a nonprofit. They are about equally unlikely to rate
a recognized nonprofit
as not outstanding
or outstanding. In
the third row, high-prominence
officers are unlikely to say that
they do not recognize a nonprofit. Once they recognize a
nonprofit, they are likely to rate it as outstanding only slightly
more than they are likely to rate it as not outstanding. The
same discontinuity appears in the association between evalu-
ations and memberships (not presented in Table
3). Officers
who are members in one or more of the area professional as-
sociations are unlikely to say that they do not recognize
a
nonprofit
(- 8.44 z-score). They are more likely
to rate recog-
nized nonprofits as outstanding rather
than not outstanding,
but both tendencies are high (respective
z-scores of 7.17 and
4.78).
These results show that the positive association between in-
tegration and evaluation is driven by variation in recognizing
nonprofits
rather than variation
in the tendency to view them
as outstanding.
Officers more integrated
into the society of
other officers have more information
on nonprofits
or a lower
response threshold for saying that they recognize a nonprofit,
or some combination of both. The membership
and promi-
nence variables
in Table 2 both have positive effects on the
tendency to recognize nonprofits
(3.56 and 3.58 t-tests, 57
d.f., p < .001). Neither is associated with an officer's ten-
dency to rate recognized
nonprofits
as outstanding
(0.88 and
-0.02 t-tests).
Integration and Contagion
The detailed
explanation
of integration's
link with evaluation is
consequential,
because it is clear from Table
2 that contagion
plays a smaller role in the evaluations of more integrated
of-
ficers. For both structural equivalence
and cohesion, conta-
gion is strongest among the low-prominence
officers and
weakest among the most prominent.
We know from the dis-
cussion of the partial
correlations from Table
2 that strong
evidence of contagion remains after stringent
controls
for dif-
ferences between the officers. The issue at hand is to under-
stand the systematic decline in that effect with increasing
social integration.
In
particular,
is the eroding contagion
effect
true of officers' evaluations
generally
or only of the recogni-
tion aspect of their
evaluations,
which is known
to change
with social integration?
The results in Table
4 show that the latter is true. We focus
on structural
equivalence because it provides more explana-
tory variance. The same pattern
can be seen in the evidence
of contagion by cohesion, but much of the pattern
involves
negligible effects, as might be expected from the negligible
aggregate effect at the bottom of Table 2. The first column in
Table
4, repeating
the results in Table 2 but now with jack-
knife t-tests, shows how contagion by structural
equivalence
decreases as integration
increases. This is a balance of the
two trends in the second and third columns of Table 4. In
the-
101tASQ,
March 1991
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
Table 4
Prominence and Structural Equivalence Contagion*
Evaluation of nonprofit organization
Officer
prominence Evaluation Recognition Outstanding
All officers .535 .543 .295
(14.42) (10.51) (6.50)
Low prominence .570 .571 .258
(11.03) (10.77) (3.02)
Average prominence .380 .265 .325
(5.19) (2.64) (4.06)
High prominence .244 .025 .281
(4.11) (0.41) (3.33)
* Correlations
between officer and structural equivalence alter evaluations are
presented for categories of officers. The criterion variable in the first column
is the three-category evaluation variable used in all prior tables. The criterion
variable
in the second column is a dummy variable distinguishing recognized
from unknown nonprofit organizations.
The criterion
variable
in the third
column is a dummy variable
distinguishing outstanding recognized non-
profits from the merely recognized (unrecognized coded as missing data).
Jackknife t-tests are in parentheses.
Across the four rows of the table, the
t-tests have 59, 23, 15 and 19 degrees of freedom.
second column, contagion in officers' recognition of a non-
profit is strong among the peripheral officers and completely
absent among the most prominent officers. In the third
column, contagion in the tendency to rate a recognized non-
profit as outstanding is not as strong as contagion across all
three evaluation categories, but it is consistent and significant
across all levels of prominence
(all
t-tests in column three are
significant beyond a .01 level of confidence).
The same pat-
tern occurs among officers who are members of professional
associations (not presented in Table 4). There is no evidence
of structural equivalence contagion
in their recognition of
nonprofits
(.080 correlation;
1.21 jackknife t-test, 23 d.f., p =
.12) and strong evidence of contagion
in their evaluations of
certain
recognized nonprofits
as outstanding
(.302 correlation;
4.60 t-test, 23 d.f., p < .001).
There are two ways to interpret
the declining contagion in
recognizing nonprofits.
One is to argue that the socially pe-
ripheral
officers are more insecure about their social legiti-
macy and so are more affected by the opinions
of their
peers.
If
this is responsible
for the declining contagion
with in-
creasing integration,
then variation
in officers' evaluations
should increase with social integration, reflecting
the freedom
that the most prominent
officers have to focus their attention
on whatever nonprofit
activities interest them. The second is
to argue that contagion has reached saturation
among the
most prominent
officers. If
this is responsible
for the de-
clining contagion
with increasing integration,
then variation
in
officers' evaluations
should decrease with social integration
as the most prominent
officers come to share the same
opinions, opinions characteristic
of the social center of officer
society.
The second interpretation
better describes the Twin
Cities
contributions officers. In Table
2, the standard
deviation
of
evaluations decreases as social integration
increases. With
respect to recognition alone, the mean and standard
deviation
of the recognition dummy variable are .504 and .501 for low-
102/ASQ, March 1991
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
Interorganization
Contagion
prominence
officers, .788 and .410 for average-prominence
officers, and .920 and .272 for the most prominent
officers. In
other words, the most prominent
officers have come to rec-
ognize all of the 10 nonprofit organizations.
In
contrast, the
dummy variable distinguishing
outstanding
nonprofits from
the merely recognized has a relatively
consistent mean and
standard deviation across prominence levels: .446 and .499
for low-prominence
officers, .452 and .500 for average-promi-
nence officers, and .516 and .501 for the most prominent
of-
ficers. Our evidence of contagion
depends on variation
in
officer evaluations.
Where officers all agree, there is no evi-
dence of contagion-as there would be no evidence if their
opinions
varied independent
of one another. Either condition
creates a zero correlation.
The fact that their
agreement is
correlated
with social integration
means that contagion
has
reached saturation
with respect to recognizing
the selected
10 nonprofit
organizations.
In this sense, and in the context of
strong evidence of contagion
elsewhere in the system, the
dissolving
evidence of contagion
in recognition
is yet another
indicator
of contagion's
effect on the Twin
City officers' eval-
uations.
CONCLUSIONS
The purpose of this article
was to see if, and how, normative
processes were at work in shaping
the perceptions
and eval-
uations of corporate
officers in a professional
community.
We
draw three conclusions from the analysis. First,
there is
strong evidence of contagion in the officers' evaluations of
nonprofit
organizations:
The way evaluations
vary
between
organizations
can be predicted
from the network of personal
contacts between people in the separate organizations.
Second, replicating
results in other studies, the observed
contagion
operates through
structural
equivalence rather than
cohesion. Direct
and indirect contacts had little
or no impact
on the officers' perceptions
and evaluations.
Officers
close to
one another in the contact network
did not recognize
the
same organizations,
nor did they think that the same non-
profits
had achieved extraordinary accomplishments.
This
leads us to believe that knowledge of and opinions
about
donees does not spread through
direct
contacts among giving
officers within
this occupational
subculture.
The contributions
officers are guided in their evaluations
by the opinions
of their
peers in the social structure
of other officers rather than the
opinions
of officers with
-whom they have personal
contact.
Third,
there are important
differences between evaluations,
depending on the object, criterion,
and source of the evalua-
tion. The evidence of contagion
persists with these factors
held constant, but the factors explain
substantial
variation in
the officers' evaluations
and highlight
interesting
leads for fu-
ture research. For
example, there is the question of why con-
tagion should be stronger
for nonprofit organizations
providing
cultural services. One explanation
is that culture is more diffi-
cult to evaluate than health and welfare. The less tied an
evaluation
is to empirical
fact, the more it will be affected by
contagion. If officers' opinions
of cultural
nonprofits
are more
contagious for this reason, then a variable could be con-
structed to measure the ambiguity of the service provided
by
103/ASQ, March 1991
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
a nonprofit.
Contagion effects should be stronger in the eval-
uations of any nonprofit higher on the variable. Our results
also demonstrate the importance of the evaluation criterion.
Quality is distinct from recognition.
There is strong evidence
of contagion in both aspects, but the evidence on recognition
varies with social integration-from a high level of contagion
among peripheral officers to a low level among the most
prominent
officers, who have come to share a recognition of
all the nonprofits selected for study. The evidence of conta-
gion in evaluating certain recognized
nonprofits as out-
standing is consistent and significant
across levels of
prominence.
The moral is that variables selected for studying
contagion have to be selected with attention to the evaluation
they request of respondents. With respect to research de-
sign, having data on multiple
evaluations is a good idea to
avoid idiosyncratic components in any one, and distribution is
important. Little will be learned from looking for (cross-sec-
tional)
contagion effects in issues on which members of the
study population have come to share a single opinion. With
respect to social contagion theory, kinds
of evaluations can be
distinguished
for their ambiguity, just as kinds
of evaluated
objects can be distinguished
for their
ambiguity. Recognizing
the name of a nonprofit organization is a low-ambiguity evalu-
ation; the name rings a bell or it doesn't. This is the evalua-
tion that shows less evidence of contagion
as officers
become more integrated
into the professional
community
and
thus exposed to more information about the nonprofit organi-
zations supported
by the community.
For officers on the pe-
riphery
of the community,
information is more often
rumor-and it is among them that recognition
evaluations
show evidence of social contagion. In contrast to merely rec-
ognizing
a nonprofit organization,
claiming
that it provides
ex-
traordinary
services is much more a matter of subjective
judgment. It is this more ambiguous
evaluation that shows
consistent evidence of contagion
for central as well as pe-
ripheral
officers.
The results are important
to organization
research
in particular
and network
theory more generally.
For network
theory,
these results add evidence from another
context, the context
of interorganization
networks, showing that contagion ap-
pears to operate by structural
equivalence rather than the tra-
ditionally presumed cohesion. To organization
research,
these
results are important
because they show how social net-
works contribute
to the isomorphism
found in organizational
fields. The common knowledge base and value consensus
among contributions officers in the Twin Cities can be traced
to the personal
contact network
among them. Beyond rein-
forcing
the idea of institutionalized
isomorphism,
our results
add clarity
to the picture
of how beliefs are transmitted be-
tween organizations.
Contagion
is typically presumed to
operate through socializing
communication,
officers in contact
with one another
developing
a shared view of the corporate
environment
by their
frequent
and empathic
communication.
We found little
support
for this thesis in our
analysis. Rather,
we found contagion
between officers structurally equivalent
in the contact network. Actors in a professional
community
do not always play
on a level playing
field. In the Twin Cities,
contributions
officers were stratified across statuses in the
hierarchy of their profession, and contagion took place among
104/ASQ, March 1991
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
Interorganization Contagion
those similarly
positioned in the hierarchy.
Rather than being
a function of behavioral
communication,
officers' views of the
corporate environment
are driven by what is expected of a
person occupying
their position in the social hierarchy of the
profession.
REFERENCES
Burt, Ronald S.
1982 Toward
a Structural
Theory
of
Action.
New York:
Academic
Press.
1987 "Social
contagion
and innova-
tion, cohesion versus structural
equivalence."
American
Journal
of Sociology,
92:
1287-1335.
1988 "Some properties
of structural
equivalence
measures derived
from sociometric
choice data."
Social
Networks,
10: 1-28.
Burt,
Ronald S., and Patrick
Doreian
1982 "Testing
a structural
model of
perception:
Conformity
and
deviance
with respect to
journal
norms in elite sociolog-
ical methodology."
Quality
and
Quantity,
16: 109-150.
Burt,
Ronald S., and Tetsuji
Uchiyama
1989 "The conditional
significance
of communication
for interper-
sonal influence."
In
Manfred
Kochen
(ed.),
The Small
World:
67-87. Norwood,
NJ: Ablex.
Coleman, James S., Elihu Katz,
and Herbert
Menzel
1966 Medical
Innovation.
New York:
Bobbs-Merrill.
DiMaggio, Paul J., and Walter
W.
Powell
1983 "The iron cage revisited:
Insti-
tutional
isomorphism
and col-
lective rationality
in
organizational
fields."
Amer-
ican Sociological
Review,
48:
147-160.
Doreian, Patrick
1981 "Estimating
linear
models with
spatially
distributed
data."
In
Samuel Leinhardt
(ed.), Socio-
logical
Methodology
1981:
359-388. San Francisco:
Jossey-Bass.
Dow, Malcolm
1984 "A biparametric approach to
network autocorrelation." So-
ciological Methods and Re-
search, 13: 201 -216.
Efron, Bradley
1982 The Jackknife, the Bootstrap
and Other Resampling Plans.
Philadelphia: SIAM.
Galaskiewicz, Joseph
1
985a Social Organization of an
Urban Grants Economy. Or-
lando, FL: Academic Press.
1985b "Professional networks and
the institutionalization of a
single mind set." American
Sociological Review, 50:
639-658.
Hinkley, David V.
1978 "Improving the jackknife with
special reference to correlation
estimation." Biometrika, 65:
13-21.
Knoke, David, and Ronald S. Burt
1983 "Centrality." In R. S. Burt and
M. J. Minor (eds.), Applied
Network Analysis: 195-222.
Beverly Hills, CA: Sage.
Miller, Rupert G.
1974 "The jackknife-A review."
Biometrika, 61: 1-15.
Mizruchi, Mark S.
1989 "Similarity
of political behavior
among large American corpo-
rations." American Journal of
Sociology, 95: 401-424.
1990 "Cohesion, structural equiva-
lence, and similarity of be-
havior: An approach to the
study of corporate political
power." Sociological Theory,
8: 16-32.
1991 The Structure of Corporate
Po-
litical Action: Interfirm
Rela-
tions and Their Consequences.
Cambridge, MA: Harvard
Uni-
versity Press (forthcoming).
Moch, Michael K., and Stanley E.
Seashore
1981 "How norms affect behaviors
in and of corporations."
In
P. C. Nystrom and W. H.
Starbuck (eds.), Handbook
of
Organizational
Design, 1:
210-237. New York:
Oxford
University Press.
Mosteller, Frederick, and John W.
Tukey
1968 "Data
analysis, including
sta-
tistics." In
Gardner
Lindzey
and Elliot
Aronson (eds.),
Handbook of Social
Psy-
chology, 2: 80-203. Reading,
MA: Addison-Wesley.
Sch9tt, Thomas, and Joseph P.
Morrissey
1988 "Organizational
behavior
and
normative environments:
Competitors
and collabo-
rators." Paper presented
at the
Eighth
Annual Sunbelt
Social
Networks
Conference.
Troy, Kathryn
1982 The Corporate
Contributions
Function. New York: The Con-
ference Board.
Zucker,
Lynne G.
1987 "Institutional
theories of orga-
nization."
Annual Review
of
Sociology,
13: 443-464.
105/ASQ, March 1991
This content downloaded from 150.135.24.59 on Mon, 13 Jan 2014 20:15:24 PM
All use subject to JSTOR Terms and Conditions
