Structural Inertia and Organizational Change

Author(s): Michael T. Hannan and John Freeman

Reviewed work(s):
Source: American Sociological Review, Vol. 49, No. 2 (Apr., 1984), pp. 149-164
Published by: American Sociological Association
Stable URL: http://www.jstor.org/stable/2095567 .
Accessed: 26/06/2012 17:51
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 [email protected].

American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to
American Sociological Review.

http://www.jstor.org

STRUCTURAL INERTIA AND ORGANIZATIONAL CHANGE*

MICHAEL

T.

JOHN FREEMAN

HANNAN

Universityof California,Berkeley

StanfordUniversity

Theory and research on organization-environment relations from a population

ecology perspective have been based on the assumption that inertial pressures on
structure are strong. This paper attempts to clarify the meaning of structural inertia
and to derive propositions about structural inertia from an explicit evolutionary
model. The proposed theory treats high levels of structural inertia as a consequence
of a selection process rather than as a precondition for selection. It also considers
how the strength of inertial forces varies with age, size, and complexity.

Most prominent organization theories explain variability in organizational characteristics, that is, diversity, throughreferenceto
the history of adaptationsby individualorganizations, Earlier(Hannanand Freeman, 1977),
we challengedthis view and arguedthat adaptation of organizational structures to environmentsoccurs principallyat the population
level, with forms of organization replacing
each other as conditions change. This initial
statementof populationecology theory rested
on a number of simplifying assumptions. A
majorone was the premise that individualorganizationsare subjectto stronginertialforces,
that is, that they seldom succeeded in making
radicalchanges in strategyand structurein the
face of environmentalthreats.
How strong are inertial forces on organizational structure?This question is substantively interesting in its own right. It is also
strategicallyimportant,because the claim that
adaptation theories of organizationalchange
shouldbe supplementedby populationecology
theoriesdependspartlyon these inertialforces
being strong.
Many popularizeddiscussions of evolution
suggest that selection processes invariably
favor adaptableforms of life. In fact the theory
of evolution makes no such claim, as we made
clear earlier (Hannan and Freeman, 1977;
Freemanand Hannan, 1983).This paper goes
beyond our earlier theory in acknowledging
that organizational changes of some kinds
occur frequentlyand that organizationssometimes even manageto make radicalchanges in
strategies and structures. Nevertheless, we

argue that selection processes tend to favor

organizationswhose structuresare difficult to
change. That is, we claim that high levels of
structuralinertia in organizationalpopulations
can be explained as an outcome of an
ecological-evolutionaryprocess.
In additionto derivingstructuralinertiaas a
consequence of a selection process, this paper
explores some of the details of inertialforces
on organizationalstructure. It considers how
inertial forces vary over the life cycle, with
organizationalsize, and with complexity, and
suggests some specific models for these dependencies.
BACKGROUND
Ourearlierformulationof an ecological theory
of organizationalchange pointedto a varietyof
constraints on structural change in organizations:
...

for wide classes of organizations there

are very strong inertial pressures on

structure arising from both internal arrangements (for example, internal politics)
and from the environment (for example,
public legitimation of organizational activity). To claim otherwise is to ignore the
most obvious feature of organizationallife.
(Hannanand Freeman, 1977:957)
Some of the factors that generate structural
inertia are internalto organizations:these include sunk costs in plant, equipment,and personnel, the dynamics of political coalitions,
and the tendency for precedents to become
normative standards. Others are external.
* Direct all correspondenceto: Michael T. Han- There are legal and other barriersto entry and
nan, Departmentof Sociology, StanfordUniversity, exit from realms of activity. Exchange relaStanford,CA 94305.
tions with other organizationsconstitutean inThe work reported here was supported by Na- vestment that is not writtenoff lightly. Finally,
tional Science Foundationgrants SES-8109382 and
ISI-8218013. We would like to acknowledge the attempting radical structural change often
helpfulcommentsof TerryAmburgey,GaryBecker, threatens legitimacy; the loss of institutional
Jack Brittain,GlennCarroll,James Coleman,Susan support may be devastating.
We continue to believe that inertial presOlzak, Jeffrey Pfeffer, Arthur Stinchcombe, and
sures on most features of organizational
three referees on earlierdrafts.
AmericanSociological Review 1984, Vol. 49 (April:149-164)

149

AMERICAN SOCIOLOGICALREVIEW

150
structureare quite strong-much strongerthan
most theorists acknowledge. Moreover, the
assumption that organizations rarely make
fundamentalchanges successfully has proven
to be a useful strategic simplification.It has
allowed a rich and evocative set of ecological
theories and models to be appliedto the probformover time
lem of changesin organizational
(see, e.g., Brittainand Freeman, 1980;Carroll,
1983; Carroll and Delacroix, 1982; Freeman,
1982;Freemanand Hannan, 1983;Freemanet
al., 1983).
However, the claim that organizational
structuresrarely change is the subject of dispute. March(1981:563)summarizeshis review
of research on organizationalchange by asserting:
Organizations are continually changing,
routinely, easily, and responsively, but
change within organizationscannot be arbitrarily controlled . . . What most reports on
implementation indicate .

. is not that or-

ganizationsare rigid and inflexible, but that

they are impressively imaginative.
The contemporaryliteraturecontainsat least
three broad points of view on organizational
change. Populationecology theory holds that
most of the variability in organizational
structurescomes aboutthroughthe creationof
new organizations and organizationalforms
and the replacementof old ones (Hannanand
Freeman, 1977; Freeman and Hannan, 1983;
McKelvey, 1982).A second view, which might
be called rationaladaptationtheory, proposes
that organizationalvariabilityreflects designed
changes in strategyand structureof individual
organizations in response to environmental
changes, threats, and opportunities.There are
numerous variants of this perspective which
differ widely on other dimensions. Contingency theories emphasize structuralchanges
that match organizational structures to
technology-environment pairs (Thompson,
1967;Lawrence and Lorsch, 1967).Resourcedependence theories emphasize structural
changes that neutralize sources of environmental uncertainty (Pfeffer and Salancik,
1978). An institutionally oriented version of
this perspective holds that organizational
structuresare rationallyadaptedto prevailing,
normatively endorsed modes of organizing
(Meyer and Rowan, 1977;DiMaggioand Powell, 1983). Marxist theories of organization
typically assert that organizationalstructures
are rational solutions for capitalist owners to
the problemof maintainingcontrol over labor
(Edwards, 1979; Burawoy, 1979). The third
broad perspective, which might be called random transformationtheory, claims that organizations change their structures mainly in re-

sponse to endogenousprocesses, but that such

changes are only loosely coupled with the desires of organizationalleaders and with the
demands and threats of environments(March
and Olsen, 1976; March, 1982;Weick, 1976).
Progress in explainingorganizationaldiversity and change requires understandingboth
the nature of organizationalchange and the
degree to which it can be planned and controlled. Here we concentrate mainly on the
first issue: does most of the observed variability in organizationalfeaturesreflect changes in
existingorganizations,whetherplannedor not,
or does it reflect changes in populationswith
relatively inert organizations replacing each
other? In other words, does change in major
features of organizations over time reflect
mainly adaptation or selection and replacement?
The selection and adaptation perspectives
are so different that it is hard to believe that
they are talking about the same things. Scott
(1981:204)claims that they are not:
...

the natural selection perspective seems

to us to be particularlyuseful in focusing
attention on the core features of organizations, explaining the life chances of
smaller and more numerous organizations,
and accountingfor changes in organizational
forms over the long run. By contrast the
rational selection or resource dependency
approach emphasizes the more peripheral
featuresof organizations,is betterappliedto
largerand more powerfulorganizations,and
stresses changes occurringover shorter periods of time.
This contrast provides a useful point of departurefor an attemptto clarifythe conditions
under which the two perspectives apply.
TRANSFORMATIONAND
REPLACEMENT
All accepted theories of biotic evolution share
the assumptionthat innovation,the creationof
new strategies and structures,is randomwith
respect to adaptivevalue. Innovationsare not
producedbecause they are useful;they arejust
produced. If an innovation turns out to enhance life chances, it will be retained and
spreadthroughthe populationwith high probability. In this sense, evolution is blind. How
can this view be reconciled with the fact that
humanactors devote so much attentionto predicting the future and to developing strategies
for coping with expected events? Can social
change, like biotic evolution, be blind?
Almost all evolutionary theories in social
science claim that social evolution has
foresight, that it is Lamarckianrather than

STRUCTURALINERTIA AND ORGANIZATIONALCHANGE

Darwinianin the sense that humanactors learn
by experience and incorporate learning into
their behavioralrepertoires(see, e.g., Nelson
and Winter, 1982).To the extent that learning
about the past helps future adaptation,social
change is indeed Lamarckian-it transforms
rather than selects. In other words, major
change processes occur within behavioral
units.
Even when actors strive to cope with their
environments,action may be randomwith respect to adaptationas long as the environments
are highly uncertain or the connections between meansand ends are not well understood.
It is the match between action and environmentaloutcomes that must be randomon
the averagefor selection models to apply. In a
world of high uncertainty,adaptive efforts by
individualsmay turn out to be essentially random with respect to future value.
The realismof Darwinianmechanismsin organizationalpopulationsalso turns on the degree to which change in organizational
structurescan be controlled by those ostensibly in command.Supposethat individualslearn
to anticipate the future and adapt strategies
accordinglyand that organizationssimply mirror the intentionsof rationalleaders. Then organizationaladaptationswould be largelynonrandom with respect to future states of the
environment.On the other hand, if Marchand
others are right, organizational change is
largely uncontrolled. Then organizations
staffedby highlyrationalplannersmay behave
essentially randomly with respect to adaptation. In other words, organizationaloutcomes
may be decoupled from individualintentions;
organizationsmay have lives of their own. In
this case it is not enough to ask whether individual humanslearn and plan rationallyfor an
uncertainfuture. One must ask whether organizationsas collective actors display the same
capacities.
The applicabilityof Darwinianargumentsto
changes in organizationalpopulationsthus depends partly on the tightness of coupling between individualintentionsand organizational
outcomes. At least two well-knownsituations
generate loose coupling: diversity of interest
among members and uncertainty about
means-endsconnections. Whenmembersof an
organization have diverse interests, organizational outcomes depend heavily on internal
politics, on the balance of power among the
constituencies. In such situations outcomes
cannot easily be matched rationallyto chang-

151

consequences may often differ greatly from

long-runconsequences. In such cases, it does
not seem realistic to assume a high degree of
congruence between designs and outcomes.
STRUCTURALINERTIA
To this point we have adopted the frame of
referenceof the existing literature,which asks
whether organizationslearn and adapt to uncertain, changing environments;but we think
this emphasis is misplaced. The most important issues about the applicability of
evolutionary-ecological theories to organizations concern the timing of changes.
Learning and adjusting structure enhances
the chance of survival only if the speed of
response is commensuratewith the temporal
patternsof relevantenvironments.Indeed, the
worst of all possible worlds is to change
structure continually only to find each time
upon reorganizationthat the environmenthas
alreadyshifted to some new configurationthat
demands yet a different structure. Learning
and structuralinertia must be considered in a
dynamic context. Can organizations learn
about their environments and change
strategies and structures as quickly as their
environmentschange? If the answer is negative, replacement or selection argumentsare
potentiallyapplicable.
Three things must be known in order to answer questions about the applicabilityof selection theories to populationsof organizations.
The first issue is the temporal pattern of
changes in key environments. Are typical

changes small or large, regular or irregular,

rapidor slow? The second issue is the speed of
learning mechanisms. How long does it take to

obtain, process, and evaluate informationon

key environments?The third issue is the responsiveness

of the structure to designed

changes. How quickly can an organizationbe

reorganized?
To claim that organizationalstructuresare
subjectto stronginertialforces is not the same
as claiming that organizationsnever change.
Rather, it means that organizations respond
relatively slowly to the occurrence of threats
and opportunities in their environments.
Therefore,structuralinertiamust be definedin
relative and dynamic terms. It refers to comparisons of the typical rates of change of the
processes identified above. In particular,
structures of organizationshave high inertia
when the speed of reorganizationis much
ing environments.
lower than the rate at which environmental
When the connections between means and conditionschange. Thus the concept of inertia,
ends are obscure or uncertain, carefully de- like fitness, refers to a correspondence besigned adaptationsmay have completelyunex- tween the behavioralcapabilitiesof a class of
pected consequences. Moreover, short-run organizationsand their environments.

152
Our definition of structuralinertia implies
that a particularclass of organizationsmight
have high inertia in the context of one environmentbut not in another. For example, the
speed of technicalchangein the semiconductor
industry has been very high over the past
twenty years. Firms that would be considered
remarkablyflexible in otherindustrieshave not
been able to reorganizequicklyenoughto keep
up with changing technologies.
One of the most importantkindsof threatsto
the success of extant organizationsis the creation of new organizations designed specifically to take advantage of some new set of
opportunities. When the costs of building a
new organization are low and the expected
time from initiationto full productionis short,
this kind of threat is intense (unless there are
legal barriers to the entry of new organizations). If the existing organizationscannot
change their strategies and structures more
quickly than entrepreneurscan begin new organizations, new competitors will have a
chance to establish footholds. Other things
being equal, the faster the speed with which
new organizationscan be built, the greater is
the (relative) inertia of a set of existing
structures.
Even such a successful and well-managed
firm as IBM moves ponderously to take advantage of new opportunities.Granted, IBM
eventually moved into the market for
minicomputersand microcomputersand appears poised to dominatethem. Still, the protracted period of assessing these markets,
waitingfor technologies to stabilize, and reorganizingproductionand marketingoperations
created the opportunityfor new firms to become established. As a consequence, the
structure of the computer industry is almost
certainlydifferentthan it would have been had
IBM been willing and able to move quickly.
The point is that IBM did change its strategy
somewhat, but this change took long enough
that new firms using different strategies and
structureswere able to flourish.
REPRODUCIBILITY,INERTIA,
AND SELECTION
As we have emphasized elsewhere, organizations are specialcorporateactors. Like other
corporate actors, they are structures for accomplishingcollective action as well as repositories of corporate resources. Unlike other
collective actors, organizationsreceive public
legitimationand social support as agents for
accomplishingspecific and limited goals. Although individual members often manipulate
organizationsto serve private goals and organizationspursueother goals in additionto their

AMERICAN SOCIOLOGICALREVIEW
public goals, the basis on which organizations
mobilize resources initially and gain support
from society is theirclaim to accomplishsome
specific set of ends (e.g., making a profit,
treatingthe sick, producingbasic scientific research).
Creating an organizationmeans mobilizing
several kindsof scarce resources.Organization
buildersmust accumulatecapital, commitment
of potential members, entrepreneurialskills,
and legitimacy(see Stinchcombe, 1965).Once
such resources have been invested in building
an organizationalstructure,they are difficultto
recover. Although one can sell the physical
assets of a disbandedorganizationand sometimes its name, most resources used to build it
are lost when it is dissolved. Not only are the
costs of startingan organizationnontrivial,but
organizationscontinually use substantialportions of their resources in maintainingand reproducingtheir structures ratherthan in performing collective action. Just as in the case
of biotic creatures, there is a substantial
metabolic overhead relative to the amount of
work performed.Thus the creationof a permanent organizationas a solution to a problemof
collective action is costly compared to other
alternatives.
Why do individualsand other social actors
agree to commit scarce resources to such expensive solutions to problemsof collective action? A number of answers to this question
have been put forth (see Scott, 1981:135-63,
for an insightfulreview). The new institutional
economics argues that organizationsarise to
fill the gaps created by marketfailure(Arrow,
1974). Williamson's(1975) influentialanalysis
proposes that organizationsare more efficient
than markets in situations in which economic
transactionsmust be completed in the face of
opportunism,uncertainty,and small-numbers
bargaining.Althoughsociologists tend to deny
that organizationsarise mainly in response to
marketfailures, they tend to agree that organizations have special efficiency properties,but
emphasize their efficiency and effectiveness
for coordinating complex tasks (Blau and
Scott, 1962;Thompson, 1967).
Although these efficiency arguments are
plausible, it is not obvious that they are correct. Many detailed accounts of organizational
processes raise serious doubts that organizations minimizethe costs of completingmany
kinds of transactions.Indeed, there appearsto
be a strong tendency for organizationsto become ends in themselves and to accumulate
personnel and an elaborate structure far beyond the technical demands of work.
Moreover, many organizationsperform very
simple tasks that involve low levels of coordination. In contrast, collections of skilledwork-

STRUCTURAL INERTIA AND ORGANIZATIONAL CHANGE

ers collaboratingin ad hoc groups can often

complete quite complex tasks. From the perspective of the performanceof a single, complex collective action, it is not obvious that a
permanentorganizationhas any technical advantage.
We emphasize differentkinds of competencies. The first of these is reliability. Organizations have unusual capacities to produce
collective products of a given quality repeatedly. In a world of uncertainty,potential
members, investors, and clients may value reliability of performancemore than efficiency.
That is, rationalactors may be willingto pay a
highprice for the certaintythata given product
or service of a certainminimumqualitywill be
availablewhen it is needed. Reliabilitydepends
on the variance of performance(includingits
timeliness) ratherthan its average level.
Organizationshave higherlevels of reliability than ad hoc collectives in two senses: one
cross-sectionaland the other temporal.Crosssectional reliability means that an outcome
chosen at randomfrom a populationof organizations will have a lower variance than one
chosen at randomfrom a populationof other
kindsof producers.Temporalreliabilitymeans
the variabilityover time in the quality (including timingof delivery) of an outcome is lower
for those produced by organizationsthan for
those producedby ad hoc groups. Overall, we
argue that the distinctive competence of organizations is the capacity to generatecollective
actions with relatively small variance in quality.
Organizationshave a second property that
gives them an advantagein the modernworld:
accountability.The spreadof generalnormsof
rationalityin the modernworld (Weber, 1968)
and a variety of internal and external contingenciesdemandthatorganizationsbe able to
accountrationallyfor theiractions. This means
both that they must be able to document how
resources have been used and to reconstruct
the sequences of organizational decisions,
rules, and actions that producedparticularoutcomes. It does not necessarilymean that organizations must tell the truth to their members
and to the public about how resources were
used or how some debacle came about. What
matters is that organizationscan make internally consistent arguments that appropriate
rules and proceduresexisted to reproducerational allocationsof resourcesand appropriate
organizationalactions.
Norms of proceduralrationalityare pervasive in the modern world. Organizational
legitimacy,in the sense of highprobabilitythat
powerful collective actors will endorse an organization'sactions (Stinchcombe, 1968), depends on ostensibleconformityto these norms.

153

Coleman (1974) has arguedthat corporateactors favor other corporateactors over individuals. We add that corporateactors especially
favorother corporateactors thatgive signalsof
proceduralrationalityand accountability.
Testing for accountability is especially intense duringorganizationbuilding,the process
of initial resource mobilization. Potential
members want assurance that their investments of time and commitment will not be
wasted. When membership involves an employment relation, potential members often
want guaranteesthat careerswithinthe organization are managed in some rational way.
Potentialinvestors (or supporters)also assess
accountability.In fact, the professionof public
accountancy arose in the United States in response to the desires of British investors in
American railroads for assurances that their
investments were being managedin appropriate ways (Chandler, 1977). Demands for accounting rationalityin this narrow sense are
both widespread and intense in modern
societies. For example, the federalgovernment
will not allocate researchgrantsand contracts
to organizationsthat have not passed a federal
audit, meaningthat they have given evidence
of possessing the appropriaterules and procedures for accounting for the use of federal
funds.
Accountability testing is also severe when
resources contract. Membersand clients who
would otherwise be willing to overlook waste
typicallychange their views when budgetsand
services are being cut.
In our judgment, pressures for accountability are especially intense when (1) organizations produce symbolic or informationloaded products (e.g., education, branded
products versus bulk goods)-see DiMaggio
and Powell (1983); (2) when substantial risK
exists (e.g., medical care); (3) when long-term
relationsbetween the organizationand its employees or clients are typical;and (4) when the
organization's purposes are highly political
(Weber, 1968). Our arguments presumably
apply with special force to organizations in
these categories. Still, we thinkthat pressures
towards accountability are generally strong
and getting stronger.The trend toward litigating disputes and pressures for formal equality
in modern polities intensifiesdemandsfor accountability. All organizations seem to be
subject to at least moderatelevels of accountability testing.
We argue that the modernworld favors collective actors that can demonstrateor at least
reasonably claim a capacity for reliable performanceand can account rationallyfor their
actions. These forces favor organizationsover
other kinds of collectives and they favor cer-

154
tain kinds of organizationsover others, since
not all organizationshave these properties in
equal measure. Selection withinorganizational
populations tends to eliminate organizations
with low reliability and accountability. The
selection processes work in several ways.
Partly they reflect testing by key actors and
environments in the organization-building
stage. Potentialmembers,investors, and other
interestedparties apply tests of reliabilityand
accountabilityto proposednew ventures. Such
testing continues after founding. Unreliability
and failuresof accountabilityat any stage in a
subsequent lifetime threatens an organization's ability to maintain commitment of
membersand clients and its ability to acquire
additionalresources.
Assumption 1. Selection in populations of organizations in modern societies favors forms
with high reliability of performance and high
levels of accountability.

When does an organization have the capacity to produce collective outcomes of a

certainminimumqualityrepeatedly?The most
importantprerequisiteis so commonplacethat
we take it for granted. Reliable performance
requires that an organizationcontinually reproduceits structure-it must have very nearly
the same structuretoday that it had yesterday.
Amongother things, this means that structures
of roles, authorityand communicationmust be
reproduciblefrom day to day.
Assumption 2. Reliability and accountability
require that organizational structures be highly
reproducible.

A structurecan conceivably be reproduced

repeatedlyby negotiationand conscious decision making.The membersof an organization
with such practices might happen to decide
each day to re-createthe structurethat existed
the previous day. But this seems unlikely. Reproducibilityis far more likely underdifferent
conditions. In general, organizationsattainreproducibilityof structurethroughprocesses of
institutionalization and by creating highly
standardizedroutines.
The first solution, institutionalization,is a
two-edged sword. It greatly lowers the cost of
collective action by giving an organizationa
taken-for-grantedcharactersuch that members
do not continuallyquestionorganizationalpurposes, authority relations, etc. Reproduction
of structureoccurs without apparenteffort in
highly institutionalizedstructures. The other
edge of the sword is inertia. The very factors
that make a system reproduciblemake it resistant to change. In particular,to the extent
that an organizationcomes to be valued for
itself, changes in structuralarrangementsbe-

AMERICAN SOCIOLOGICALREVIEW
come moral and political ratherthan technical
issues. Attempts at redesigning structures in
organizationsbuilt on moral commitmentare
likely to spark bursts of collective opposition
premisedon moralclaims in favor of the status
quo. Even if such oppositiondoes not prevail,
it delays change considerably.
As a brake on structural change, institutionalizationapplies both to the organization as a whole and to its subunits. But what
aboutthe diversityamongsets of differentiated
activities within the organization?Some kinds
of organizationsperformdiverse sets of activities, sometimes in paralleland sometimes sequentially. Military organizations provide a
striking example; they maintain"peacetime"
and "wartime" structures.' Similarly, labor
unions gear up for organizing drives or for
waves of strikesand then returnto more placid
bread-and-buttercollective bargaining. Manufacturingfirms sometimes concentrateon redesigningproductsand at other times concentrate on marketingan extant set of products.
Each phase of organizationalactivity involves
mobilizing different kinds of structures of
communication and coordination. In a real
sense these kinds of organizationscan be said
to use differentstructuresin differentphases.
Does this meanthatthese organizationshave
somehow escaped inertial tendencies? We
think not, at least from the perspective of attempts at building theories of organizational
change. These organizations have multiple
routines; they shift from one routine(or set of
routines) to another in a fairly mechanical
fashion. We thinkthat organizationshave high
inertia both in the sets of routines employed
and in the set of rules used to switch between
routines.
According to Nelson and Winter (1982:96)
routines are the "source of continuity in the
behavioral patterns of organizations." They
are patterns of activity that can be invoked
repeatedlyby membersand subunits.One way
of conceiving of routines is as organizational
memory-an organization's repertoire of
routines is the set of collective actions that it
can do from memory. Nelson and Winteremphasize that organizationsrememberby doing.
Like knowledgeof elementaryalgebraor high
school Latin, collective knowledgeis the basis
of organizationalroutines and decays rapidly
with disuse. Even occasional use reveals some
decay in recall and demonstratesthe need to
reinvest in learning to keep skills at their
I Janowitz (1960) discusses various conflicting
demands of organizing military activities in
peacetime and war. Etzioni (1975) discusses the
shifts in control problems that arise in armies and
labor unions as a result of such changes.

STRUCTURAL INERTIA AND ORGANIZATIONAL CHANGE

former levels. Organizations that have

the capacity to use a broad repertoire of
routinesdo so by virtueof largeinvestmentsin
keeping their routines sharp. For example,
peacetime armies devote a great deal of their
resources to simulatingwartimesituationsand
training.Armies that fail to make such an investment experience great difficultyin making
the transitionto battle readiness.
The fact that organizationalroutines decay
with disuse implies that organizationsface the
classic specialism-generalismdilemma in deciding how many routines to maintainat any
fixed level of resources. Generalists (those
with many routines)are no less inert than specialists in the manner in which they adapt to
environmentalchange in the sense that they
still use a limited numberof routines. As Nelson and Winter (1982:134)put it:

155

tions in changing environments. But, as we

argued earlier (Freeman and Hannan, 1983),
organizationsthat frequentlytry to reorganize
may producevery little andhave slightchances
of survival. Here the issue is the cause of
structuralinertiaratherthan its consequences.
Our argumentis that resistance to structural
change is a likely by-productof the ability to
reproducea structurewith high fidelity:
Assumption 3. High levels of reproducibility of
structure generate strong inertial pressures.

The three assumptionsform the core of our

first argument.Taken together they imply:
Theorem 1. Selection within populations of organizations in modern societies favors organizations whose structures have high inertia.

This theorem states that structuralinertiacan

be a consequence of selection rather than a
... it is quite inappropriate to conceive of
precondition.All that is requiredis that some
firm behavior in terms of deliberatechoice organizationsin an initialpopulationhave high
from a broadmenu of alternativesthat some levels of reproducibility(hence high levels of
outside observorconsidersto be "available" inertia) and that selection pressures be reato the organization.The menuis not broad,it sonably strong. Under such conditions, selecis narrow and idiosyncratic . . . Efforts to
tion pressures in modem societies favor orgaunderstandthe functioningof industriesand nizations whose structures are resistant to
largersystems shouldcome to gripswith the change, which makes selection argumentsall
fact that highlyflexible adaptationto change the more applicable.
is not likely to characterizethe behaviorof
individualfirms.
A HIERARCHYOF INERTIAL FORCES
We think that it is a reasonablefirst approximation to think of organizationsas possessing So far we have considered organizations as
relatively fixed repertoires of highly repro- unitary actors, either adapting to their enviducible routines. Then the present argument ronmentsor remaininginert. This is simplistic
can be applied either to the organizationas a in that it ignores the obvious fact that some
whole, where the issue is the diversity of the parts of organizations change more quickly
than others and that adaptive changes are
repertoire,or to the individualroutine.
Thus we argue that the propertiesthat give sometimes not difficult to discern or implesome organizationsreproducibilityalso make ment. Universities,for example,are constantly
them highly resistant to structural change, changing the textbooks used for instruction.
whether designed or not. As we noted above, They do so in an adaptiveway, keepingup with
this means that some aspects of structurecan the constantly evolving knowledge bases of
be changed only slowly and at considerable their various fields. Persuading a university
cost (many resources must be applied to pro- faculty to abandonliberal arts for the sake of
duce structuralchange). Such structureshave vocational trainingis somethingelse again.
a dead-weightquality; there are large lags in
Why would the university'scurriculumbe so
response to environmentalchanges and to at- difficultto change?A numberof answerscome
tempts by decision makers to implement quickly to mind. The curriculumembodies the
change. Since lags in response can be longer university'sidentitywith referenceboth to the
than typical environmental fluctuations and broader society and to its participants(i.e.,
longer than the attention spans of decision faculty, students, staff, administration,
makers and outside authorities, inertia often alumni).The kinds of courses offered and the
blocks structuralchange completely.
frequencywith which they are offered serve as
The inertia of reproducibleorganizationsis a statement of purpose which is articulated
usually viewed as a pathology.A classic state- with society's value system. The curriculum
ment of this position is Merton's(1957) essay also represents one of the bases on which reon the "dysfunctions of bureaucracy."High sources are distributed. A change toward a
levels of inertia may produce serious mis- more vocationally oriented set of courses
matchesbetween organizationaloutcomes and threatens entrenched interests. Professors of
the intentionsof membersand clients in situa- classics and other humanistic fields which

156
would have a lesser role in such an institution
can be expected to resist such a change. The
curriculumis difficultto change, then, because
it representsthe core of the university'sorganizational identity and underlies the distribution of resources across the organization. In
these ways, it can be said to lie at the university's "core."
This view of organizationsas having a core
which is more difficult to modify than more
peripheralparts of its structureis not new. As
Parsons (1960:59-69) pointed out, organizational authorityhierarchiesare not continuous; qualitative breaks occur between the
technical, managerial,and institutionallevels.
The technicalsystem is that partof the organization that directly processes the "materials"
used by the organization.The resources used
by the technical system to do the organization's basic work are allocated by a broader
organizational apparatus, the managerial or
administrativesystem, which also relatesthose
technicalactivities to the public served. While
each depends on the other, the managerial
level standsin a superordinateposition. It both
controls and services the technical level's operations, while the reverse is less often the
case.
The third part, the institutionalsystem, articulates the whole organization with the
broadersociety. Parsonsemphasizedits role in
legitimatingthe organization.Boards of trustees and directorsare responsiblefor long-run
policy and for the conduct of the organization
with regardto its reputed goals. Because the
institutionaland manageriallevels of the organization stand prior to the technical level in
controllingthe flow of resources, any importantchange in theiroperationsleads to changes
in the details of the operationsof the technical
system, while the reverse is less often true.
Thompson(1967)adopted these distinctions
in arguingthat organizationsare builtin such a
way as to protect structuralunits carryingout
the primary technology from uncertainties
emanatingfrom the environment.Thompson,
however, drew core-periphery distinctions
with reference to the organization'soperating
technology. Since we thinkthat the importance
of technology in determiningstructurevaries
greatly across kinds of organizations,we emphasize institutionalcharacteristicsmore than
technical ones. In this way our approach is
closer to Parsons than to Thompson.
An argument similar to ours has been advanced by Downs (1967:167-68)in his use of
the metaphorof organizationaldepth:

AMERICAN SOCIOLOGICALREVIEW
of the specific actions taken by the bureau,
the second of the decision-makingrules it
uses, the thirdof the institutionalstructureit
uses to make those rules, and the deepest of
its general purposes.
The layers supposedly differ in characteristic
speeds of response.
We conceptualizeorganizationalstructureas
composed of hierarchicallayers of structural
and strategicfeatures that vary systematically
in flexibility and responsiveness. Our theory
emphasizes the claims used to mobilize resources for beginningan organizationand the
strategies and structures used to maintain
flows of scarce resources. Thus we classify
items of structureaccordingto theirbearingon
resourcemobilization.Fromthe perspectiveof
resource mobilization,the core aspects of organizationare (1) its statedgoals-the bases on
which legitimacy and other resources are
mobilized; (2) forms of authority within the

organization and the basis of exchange between membersand the organization;(3) core
technology, especially as encoded in capital
investment, infrastructure,and the skills of
members; and (4) marketing strategy in a broad

sense-the kinds of clients (or customers) to

which the organizationorients its production
and the ways in which it attracts resources
fromthe environment.The four characteristics
standin a roughhierarchy,with publiclystated
goals subject to the strongest constraintsand
marketingstrategy the weakest. Thus we expect the likelihoodof changeby transformation
to decline as one proceeds up the hierarchy.
These four properties provide a possible
basis on which to classify organizationsinto
forms for ecological analysis. An organization's initial configurationon these four dimensions commits it to a certainform of environmental dependence and to a long-term
strategy. Once an organization has made a
public claim to mobilize resources, has induced individualsto cede some control in return for specific inducements,has invested in
physical and human capital of specific types,
and has designeda productor service to appeal
to a certain audience, it has greatly limited its
range of feasible transformations.
Although organizations sometimes manage
to change positions on these dimensions, such
changes are both rare and costly and seem to
subject an organization to greatly increased
risks of death. Thus these characteristicsserve
as a possible basis for selection and replacement within populationsof organizations.
Although the four properties listed above
. . . organizations have different structural encompass much of organizational strategy
depths. Our analysis recognizes four "orga- and structure, they do not come close to
nizational layers." The shallowest consists exhausting the dimensions of structure that

STRUCTURALINERTIA AND ORGANIZATIONALCHANGE

interest social scientists. In particular,the list
does not include structurein the narrowsense
of numbers and sizes of subunits, numberof
levels in authoritystructures,span of control,
patternsof communication,and so forth. Nor
does it contain what Scott (1981) calls peripheral structures, the detailed arrangementsby
which an organizationmakes links with its environmentand tries to bufferits technicalcore
(for example, interlocking directorates and
joint ventures).
We think that properties of organization
charts and patternsof specific exchanges with
actors in the environmentare more plasticthan
the core set. They tend to change as organizations grow and decline in size, as
technologies change, and as competitive and
institutionalenvironments change. They can
be transformedbecause attempts at changing
them involve relativelylittle moraland political
opposition within the organizationand in the
environment and do not raise fundamental
questions about the natureof the organization.
In short, inertial forces on these aspects of
structureand on peripheralor bufferingactivities tend to be weaker than those on core features.
Most organizationtheories assume that peripheral structures are premised on and
adapted tQ a core structure. Changes in core
structuresusuallyrequireadjustmentin the peripheral structures. However, the reverse is
not true.2If a core structureis subjectto strong
selection pressure, peripheral structures will
also be subject to at least weak (indirect)
selection. In such cases, ecological theory
applies at least indirectlyto changes in peripheral structures. The tighter the coupling between the core and peripheralstructures, the
more direct is the applicabilityof our theory.
Overall we are inclined to agree with Scott
that evolutionary-ecological theories apply
more appropriatelyto core propertiesthan to
others. But we think that is because the
strengthof inertialpressuresdiffer ratherthan
because selection pressures on core and peripheralstructuresdiffer.
In addition to varying by aspects of
structure, the strength of inertial forces may
also vary with life-cycle phase, size, and
complexity. The remainderof the discussion
considers these issues.
LIFE-CYCLEVARIATIONSIN INERTIA
Newly created organizationsapparentlyhave
lower levels of reproducibilitythanolder ones.

157

As Stinchcombe(1965)pointed out, new organizations typically have to rely on the cooperation of strangers. Development of trust and
smoothly working relationshipstake time. It
also takes time to work out routines. Initially
there is much learningby doing and comparing
alternatives. Existing organizations have an
advantageover new ones in that it is easier to
continue existing routines than to create new
ones or borrowold ones (see the discussion in
Nelson and Winter, 1982:99-107). Such
arguments underlie the commonly observed
monotonicallydecliningcost curve at the firm
level-the so-called learningcurve.
In addition,the reliabilityand accountability
of organizationalaction depend on members
having acquired a range of organizationspecific skills (such as knowledge of specialized rules and tacit understandings).Because such skills have no value outside the
organization,membersmay be reluctantto invest heavily in acquiringthem until an organization has proven itself (see Becker, 1975).
Once an organizationsurvivesthe initialperiod
of testing by the environment,it becomes less
costly for members to make investments in
organization-specificlearning-early success
breeds the conditions for later success. Thus
collective action may become more reliable
and accountablewith age simply because of a
temporalpatternof investments by members.
Moreover, the collective returns to investments in organization-specificlearning may
take time to be realized,just like the case for
other formsof humancapital.For bothof these
reasons, the levels of reliabilityand accountabilityof organizationalaction should increase
with age, at least initially.
Once membershave made extensive investments in acquiringorganization-specificskills,
the costs of switching to other organizations
rise. Consequently the stake of members in
keeping the organizationgoing tends to rise as
it ages.
Finally, processes of institutionalizationalso
take time. In particular,it takes time for an
organizationto acquire institutionalreality to
its membersand to become valued in its own
right.
Assumption 4. Reproducibility of structure increases monotonically with age.
Theorem 2. Structural inertia increases
monotonically with age. (From Assumptions 2

and 4)
Theorem 3. Organizational death rates decrease with age. (From Assumption 4 and

2 Hawley's(1%8)principleof isomorphismmakes Theorem 1)

a similarargumentconcerningthe relationshipbeTheorem 3, often called the "liability of
tween "key functions" and other organizational
structures.
newness" hypothesis(Stinchcombe,1965),has

158
been well documented empirically (see
Freeman et al., 1983). Death rates appear to
decline approximatelyexponentiallyas organizations age. One explanationfor this findingis
that reproducibilityrises roughlyexponentially
with age over the early years in an organization's life.
Processes of external legitimationalso take
time. Although an organization must have
some minimal level of public legitimacy in
order to mobilize sufficientresources to begin
operations, new organizations(and especially
new organizationalforms) have rather weak
claims on public and official support.Nothing
legitimates both individual organizationsand
forms more than longevity. Old organizations
tend to develop dense webs of exchange, to
affiliatewith centers of power, and to acquire
an aura of inevitability. External actors may
also wait for an initial period of testing to be
passed before makinginvestmentsin exchange
relations with new organizations. Thus processes of institutionalizationin the environment and exchange relationshipswith relevant
sectors of the environmentmay accountfor the
relationshipsstated in Theorems 2 and 3. The
argumentto this point cannot distinguishbetween the internaland external sources of the
relationships.
SIZE AND INERTIA
We argued above that dampenedresponse to
environmentalthreats and opportunitiesis the
price paid for reliableand accountablecollective action. If this argumentis correct, organizations respond more slowly than individuals
on average to environmentalchanges. However, some organizationsare little more than
extensions of the wills of dominantcoalitions
or individuals;they have no lives of their own.
Such organizationsmay change strategy and
structurein responseto environmentalchanges
almost as quickly as the individualswho control them. Changein populationsof such organizationsmay operate as much by transformation as selection.
Except in exceptional cases, only relatively
small organizationsfit this description.An organizationcan be a simple tool of a dominant
leader only when the leader does not delegate
authority and power down long chains of
command. Failure to delegate usually causes
problems in large organizations. Indeed, the
failureof moderate-sizedorganizationsis often
explained as resulting from the unwillingness
of a founder-leaderto delegateresponsibilityas
the organizationgrew.
One way to conceptualize the issues involved is to assume that there is a criticalsize,
which may vary by form of organization(and

AMERICAN SOCIOLOGICALREVIEW
also, perhaps,by age), at which failureto delegate power sharply limits viability. In such a
threshold model, organizationsmay be quite
responsive below the threshold level of size.
Above the threshold, organizations tend to
have higher inertia. Or the relationship between size and inertiamay be roughlycontinuous. Downs (1967:60)argues that for the case
of public bureaus: ". . . the increasing size of

the bureauleads to a gradualossificationof its

action . . . the spread and flexibility of its op-

eration steadily diminish."Whetherthere is a

thresholdas we have suggestedor a continuous
relationship as Downs suggested, it seems
clear to us that size does affect inertia.
Assumption 5. The level of structural inertia
increases with size for each class of organization.

Assumption5 seems to suggestthat selection

argumentsare more appropriatefor largeorganizationsthanfor smallones, contraryto widespread opinion (Aldrich, 1979; Perrow, 1979;
Scott, 1981; Astley and Van de Ven, 1983).
However, the situation is more complex than
this. The likelihood that an organizationadjusts structureto changingenvironmentalcircumstancesdependson two factors:the rate of
undertakingstructuralchange and the probability of succeeding in implementingchange,
given an attempt. Assumption5 suggests that
the first quantity, the rate of attempting
change, is higher for small organizations.But
what about the second quantity?
It is helpful in answering this question to
complicate the model slightly. Fundamental
change-change in core aspects of structurerarely occurs over night. More commonly, an
organizationspends some period of time reorganizing, either by design or happenstance.
Usually there is a period of time duringwhich
existing rules and structuresare being dismantled (or successfully challenged)and new ones
are being created to replace them. Similarly,
existing links with the environmentare cut and
new links forged. Duringsuch periods, organizations have elements of both old and new
structures.The presence of multiplerules and
structures greatly complicates organizational
action; so too does a shifting set of environmentalrelations.Such changesincreasethe
likelihoodof conflict within an organizationas
contendingparties seek to shape rules to benefit their self-interests.
Fundamentalreorganizationmay sometimes
occur graduallyand imperceptibly,but sometimes sharp breaks with the past can be discerned, and one can identify the approximate
time of onset of the reorganization.One clear
exampleis a declarationof bankruptcyin order
to obtain relief from creditorsduringa period

STRUCTURAL INERTIA AND ORGANIZATIONAL CHANGE

of attempted reorganization.In many other

circumstances, organizational leaders announce plannedshifts in strategyand structure
such as entries into new marketsand internal
restructuring.In such cases it may be helpfulto
introducea new state into the model: the state
of attemptingfundamentalreorganization.Figure 1 depicts the possible transitions in this
expanded state space. The parametersassociated with each transition, the r's, are instantaneous transitionrates. In terms of this representation, Assumption5 states that the rate of
movingto the state of reorganizationdecreases
with size. But it says nothing about the other
rates.
The processes of dismantlingone structure
and buildinganother make organizationalaction unstable. Consequently, the variance of
quality and timeliness of collective action decline duringreorganization.
Assumption 6. The process of attempting reorganization lowers reliability of performance.

Assumptions 1 and 6 together imply:

Theorem 4. Attempts at reorganization increase death rates.

Organizations undergoing structural

transformationare highly vulnerableto environmentalshocks. Large size presumablyenhances the capacity to withstandsuch shocks.
Small organizations have small margins for
error because they cannot easily reduce the
scope of their operationsmuch in response to
temporary setbacks. Indeed, the claim that
death rates decrease with size is nothingmore
thana restatementof the idea advancedearlier
(Hannanand Freeman, 1977)that longer time
spans must be used to study replacementin
populationsof large organizations.
Assumption 7. Organizational death rates decrease with size.

We assume that size has qualitativelysimilar

effects on all three death rates in Figure 1: rd,
re, and rf. Thus small organizations are assumed to be more likely than large ones to
ra

ORIGINAL

NEW

REORGANIZATION

STRUCTURE

STRUCTURE

rb

re

DEATH

Figure 1. State Space for the Process of Fundamental Change in Organizational

Structure(The r 's are instantaneoustransition rates)

159

enter the state of reorganization,but are also

more likely to exit this state by death.
Finally, there is the issue of success at implementing change (the rate of moving from
"reorganization"to "new structure").An organizationundertakingreorganizationcan successfully make the transitionto the new state
or it can drift back to its original structure,
assuming that it does not die. The model in
Figure 1 contains two rates that pertain to
these processes: rc, the rate of moving to the
new structure,and rb, the rate of returningto
the old one. The effect of size on these rates is
unclear.On the one hand, the greaterinertiaof
largeorganizationsmightlower the rate of successes at reorganization,On the other hand,
successes at reorganizationmight depend on
the magnitudeof resourcesappliedto the task.
Since large organizationstypically have more
resources than small ones, this line of reasoning suggests that the rate of achieving
structuralchange increases with size.
The relationshipbetween size and the rate of
structural change is indeterminate in our
theory for two reasons. The first is ignorance
about the effects of size on rates of completing
structuralreorganization,conditional on having attempted it. The second source of indeterminacyis the implicationthat small organizations are more likely to attempt structural
change but are also more likely to die in the
attempt. Although our analysis does not offer
an answer to the main question about size and
inertia, it does not support the widespread
view that ecological argumentsare particularly
appropriatefor the study of change in populations of small organizations.
The model in Figure 1 may be substantively
interestingin its own right, assumingthat approximate information on dates of leaving
states of reorganizationcan be obtained. it
provides a frameworkfor addressinga variety
of questions about inertia and change. It has
the advantageof transformingwhat have been
mainly rhetoricalquestions about the applicability of the ecological perspectiveinto specific
research questions.
Consider again the question of life-cycle
variations discussed in the previous section.
Recall that we assume that reproducibilityincreases with age (Assumption 4) because
routines become worked out, role relations
stabilize,and so forth. Whateffect, if any, does
structuralreorganizationhave on these processes? We think that reorganizationis sometimes tantamountto creating a new organization (with a given level of resources). When
reorganization is that fundamental, work
groups are reshuffled, bringingstrangersinto
contact, routines are revised, and lines of
communicationare reshaped. In this situation

160

AMERICAN SOCIOLOGICALREVIEW

reorganizationrobs an organization'shistoryof
survivalvalue. That is, reorganizationreduces
the reliabilityof performanceto that of a new
organization. The stability of the previous
structuredoes not contributeto reducingvariability with new sets of procedures, role relations, etc.
If internal processes are solely responsible
for the tendency of organizationaldeath rates
to decline with age (Theorem3), the deathrate
for an organizationthat has just entered the
state "new structure"should be no lower than
the deathrate of a completelynew organization
with that structure(andlevels of resources).In
this sense, reorganizationsets the "liabilityof
newness" clock back towards zero.
Assumption 8. Structural reorganization produces a liability of newness.

In order to make this argumentconcrete, we

consider its implications for one kind of
parametricmodel for liability of newness. A
variation of the Makehammodel fits data on
age-variations in organizational death rates
well (Freemanet al., 1983; Carrolland Delacroix, 1982). This model has the form
r d(t It o) = a + Pe --Wl-. o),

where to is the time of foundingand y is positive. The liability of newness in this model is
expressed by f3,because the initialdeathrate is
a + f8 and the asymptotic death rate is a.
Imaginean organizationcreated at time to that
successfully changes its structureat tn, that is,
it enters the state "new organization"at that
time. The argument that the liability-ofnewness clock is set back towardszero implies
that its deathrate at timet approximatesthatof
a new organizationwith the same structure.In
particular,suppose that the deathrate of a new
organizationwith structurelike this one has the
following age-dependencein death rates
rd(t Ito) = a' + f'e-v'(' -to),
:
where y' 0. Then for the case of an organization born at to that switches to this structureat
tn > to, the death rate is given by
rd(tltn) = a' + /3'e-V'('-'t).
That is, developmentover the period(to,tn) has
no impacton its death rate, other things being
equal.
The argumentin the preceding paragraphs
can be viewed as one way to formalize some
long-standing notions about organizational
crises. Childand Kieser (1981:48)put the issue
as follows: "To some extent, a crisis successfully overcome may representa new birth,
in the sense that changes initiated are sufficiently radical for a new identity to emerge."
We suggest that such questions be viewed in

terms of shifts in age-dependenciesin organizational death rates.

Externalprocesses may also account for the
tendency of death rates to decline with age.
For example, we mentioned the tendency for
organizationsto acquire legitimacy simply by
virtue of longevity as well as the fact that it
takes time for organizationsto develop enduring exchange relations with key actors in the
environment. Some sorts of changes in strategy and structurestrainexternalrelations, especially when the changes imply a shift in ostensible goals. But, simple structuralreorganization, without any apparentchange of goals,
does not rob an organization'shistory of its
value for publiclegitimacyand does not necessarily upset exchange relations with the environment. Old organizations can presumably
count on their existing exchange partnersfor
support during and following such structural
change.
If the liability of newness reflects internal
processes, the death rate will jump with
structuralchanges. In contrast,if the decline in
the death rate with age reflects mainly the operationof externalprocesses of legitimacyand
exchange, the death rate will not jump when
structuralchanges do not imply a change in
basic goals. That is, argumentsabout internal
and external processes lead to different predictions about the effect of structuralreorganizationon the deathrate. Thereforethe study of
such effects may shed light on the relative importance of internaland externalprocesses in
accountingfor age variationin the deathrate in
selected organizationalpopulations.
Finally, there is no reasonto suspect thatthe
death rate declines with durationin the state
"reorganization."Quite the contrary-as the
length of time over which reorganizationis attempted increases, the costs (especially the
opportunitycosts) of reorganizationincrease.
As the fractionof organizationalresources devoted to reorganizationincreases, the capacity
of the organizationto producecollective products declines along with its capacity to defend
itself from internal and external challenges.
Hence protracted periods of reorganization
disruptorganizationalcontinuity and increase
the risk of death.
Assumption 9. The death rate of organizations
attempting structural change rises with the duration of the reorganization.

A model consistent with this assumptionis the

classic Gompertzmodel:
re(t It r) = Oek(t-t),

where tr is the time of entering the state of

reorganizationand k > 0. This sort of model
can perhapselucidate anotherclaim in the or-

STRUCTURAL INERTIA AND ORGANIZATIONAL CHANGE

ganizations literature. March (1981:567), referring to the, work of Hermann (1963) and
Mayhew (1979), states that
. . . organizations facing bad times will follow riskier and riskier strategies, thus simultaneously increasing their chances of survival and reducing their life expectancy.
Choices that seek to reverse a decline, for
example, may not maximize expected value.
As a consequence, for those that do not survive, efforts to survive will have speeded up
the process of failure.
It is hard to imagine how an action can both
increase a survival probability and increase the
death rate in conventional models for the death
rate (since life expectancy is a monotonically
decreasing function of the death rate). However, the framework introduced above is consistent with this sort of pattern.
Consider the case in which the death rate of
organizations in some environment rises precipitously at a certain moment t1 (due perhaps
to some discontinuous change in the environment). The death rate of organizations that retain their structures, rd, will gradually decline
to an asymptote that is considerably higher
than the asymptotic rate in the old environment.
Suppose that some organizations in the
population attempt structural change at t 1.
Consider two kinds of trajectories of death
rates by age. The dashed trajectory in Figure 2
depicts the death rate of an organization that
successfully implements the new structure at
t3.. The dotted trajectory pertains to an organi-

161

zationthat revertsto the old structureat t4. In a

collection of histories like those in Figure 2,
one would see that strategicaction to promote
survivalexposes an organizationto great risks
(thereby "reducingits life expectancy"). But,
because the death rate declines rapidly with
duration in the new structure, a successful
transformationeventually leads to a lower
death rate (seeming to "increase chances of
survival")--even lower than the death rates of
organizationsthat retainthe originalstructure.
However, it is not clear that structuralchange
necessarily increases unconditional life expectancy. This depends on the various rates.
Still, introducingthe competingrisks of death
and reorganizationallows one to deal systematicallywith this complicatedproblem.

ENVIRONMENTALCHANGE,
SIZE, AND INERTIA
Assumption 5 states that large organizations
are less likely than smallones to initiateradical
structuralchange. Does this mean that larger
organizationshave greater inertia, as Downs
(1967) and others have claimed? If inertia is
equated with low absolute rates of initiating
structural change, it does. When inertia is
viewed in comparativeterms, as we argue it
should be, the relationshipof size to inertiais
more complicatedthan the literaturehas indicated.
According to Assumption 7, the death rate
declines with size. This statementis equivalent
to the propositionthat time-scales of selection
processes stretch with size, as we noted earlier. One way to visualize such a relationshipis
to consider environmentalvariationsas composed of a spectrumof frequenciesof varying
lengths-hourly, daily, weekly, annually,etc.
Small organizations are more sensitive to
/..
high-frequency variations than large organizations. For example, short-termvariationsin
the availabilityof creditmay be catastrophicto
small businesses but only a minor nuisance to
giant firms. To the extent that large organizations can buffer themselves against the effects of high-frequencyvariations,their viabilFigure 2. Hypothetical death-rate functions for a ity depends mainly on lower-frequencyvariapopulationof organizationsexposed to a tions. The latter become the crucial adaptive
shift in selectionpressuresat t1. The solid problem for large organizations. In other
decreasing curves represent the death words, the temporal dimensions of selection
rates of organizationsthat retain their environmentsvary by size.
strategiesand structures.The risingsolid
We proposedabove that inertiabe definedin
curve representsthe death-ratefunction terms of speed of adjustmentrelative to the
of organizationsthat undergoattemptsat temporal pattern of key environmental
reorganizationat t2. The dashed curve changes. Althoughsmallorganizationsare less
representsthe new (betteradapted)strategy and structureat t3. The dotted curve ponderous than large ones (and can therefore
representsthe death-ratefunction of or- adjust structures more rapidly), the enviganizationsthat revertto their old strate- ronmentalvariationsto which they are sensitive tend to change with much higher fregies at t4.
:1

12

V3

=4

162
quency. Therefore, whether the adjustment
speeds of small organizationsexceed those of
large ones compared to the volatility of relevant environments is an open question. One
can easily imaginecases in whichthe reverse is
true, in which elephantine organizationsface
environmentsthat change so slowly that they
have relatively less inertia than the smallest
organizations.

COMPLEXITYAND INERTIA
The complexity of organizational arrangements may also affect the strength of inertial forces. Althoughthe term complexity is
used frequentlyin the literatureto refer to the
numbersof subunitsor to the relative sizes of
subunits,we use the termto referto patternsof
links amongsubunits.FollowingSimon (1962),
we identifya simple structurewith a hierarchical set of links, which means that subunitscan
be clustered within units in the fashion of
Chineseboxes (what mathematicianscall a lattice).
Hierarchicalsystems have the propertythat
flows (of information,commands, resources)
are localized: an adjustmentwithin one unit
affects only units withinthe same branchof the
hierarchy.Simon (1962)arguedthat hierarchical patterns appearfrequentlyin nature("nature loves hierarchy")because the probability
that a complex assembly is completed in an
environment subject to periodic random
shocks is higher when stable subassemblies
exist, as in a hierarchy. More complex
structures do not have many stable subassemblies and thus are vulnerable to shocks
duringthe whole developmentalsequence.
Recent work on populationecology supports
Simon's argument.For example, May (1974),
Siljak(1975),and Laddeand Siljak(1976)show
analytically and with simulation experiments
that ecological networksare destabilizedwhen
links (of predation,competition,or symbiosis)
are introduced.Both the numberof links and
the complexity of the patternaffect stability.
We think that similar arguments apply to
structuralchange within organizations.When
links among subunits of an organizationare
hierarchical,one unit can change its structure
without requiring any adjustment by other
units outside its branch. However, when the
pattern of links is nonhierarchical,change in
one subunitrequiresadjustmentby manymore
subunits. Such adjustmentprocesses can have
cycles; change in one unit can set off reactions
in otherunits, which in turnrequireadjustment
by the unit that initiated the change. Long
chains of adjustmentsmay reduce the speed
with which organizationscan reorganizein re-

AMERICAN SOCIOLOGICALREVIEW
sponse to environmentalthreats and opportunities.

Althoughslow response does not necessarily

imply a lower rate of attempting structural
change, it seems likely that this is the tendency. As we noted above, a slow speed of
response increasesthe likelihoodthat the environmentwill have changedbefore an organization can complete a process of reorganization.
Knowledge of this fact may dissuade organizationalleadersfrominitiatingchangeand may
serve as a powerful objection to proposed
change by parties who benefit from the status
quo.
Complex systems have slow response times
not because they are any slower than simpler
systems in detectingenvironmentalthreatsand
opportunitiesbut because the process of adjustment takes longer. In terms of the framework developed in earlier sections, this argument implies:
Assumption 10. Complexity increases the expected duration of reorganization.

Thatis, once a complexorganizationhas begun

structuralchange, it will tend to be exposed to
a longerperiodof reorganizationthana simpler
organizationattemptingsimilar changes. Assumptions9 and 10 imply:
Theorem 5. Complexity increases the risk of
death due to reorganization.

A complete analysis requires consideration

of the effects of complexity on rates of initiating change and of its effects on success in implementingchange (as we discussed above in
the analysis of the effects of size). We are not
yet ready to make any claims about effects of
complexity on these rates. Still, the result in
Theorem5 suggests that population-ecological
analysis might be more appropriatefor explainingchange in populationsof complex organizationsthan in populationsof simple ones
because complexity increases inertia by at
least one mechanism.This result, like that on
size, disagrees with the conventionalwisdom.
CONCLUSIONS
We have attemptedto clarifywhen it is reasonable to assume that organizationalstructures
have inertia in the face of environmentalturbulence. We have argued that selection pressures in modern societies favor organizations
that can reliablyproducecollective action and
can account rationally for their activities. A
prerequisitefor reliable and accountable performance is the capacity to reproduce a
structurewith high fidelity. The price paid for
high-fidelityreproductionis structuralinertia.
Thus if selection favors reliable, accountable

STRUCTURALINERTIA AND ORGANIZATIONALCHANGE

organizations,it also favors organizationswith
high levels of inertia. In this sense, inertiacan
be considered to be a by-productof selection.
Our argumenton this point may be considered
an instance of the more general evolutionary
argumentthat selection tends to favor stable
systems (see Simon, 1962).
Of course, the claim that selection favors
organizationswith high inertiais not a warrant
for assumingthat most organizationshave high
inertia. Selection pressures often may not be
strong enough to screen exhaustively for the
"most fit" organizations.Moreover, most organizationalpopulationsare replenishedmore
or less continuouslyby an inflow of new members. Younger organizationstend to have less
inertiathan older ones, and new organizations
are more likely to adopt structuresthat differ
greatly from those that would dominate any
steady-stateof the process subjectto selection
and closed to new entries.
Organizationalselection operates on many
dimensions besides reproducibility of
structure. If selection pressures on specific
featuresof structureare sufficientlystrong,organizations with the characteristicsappropriate to the environmentare favoredeven if they
have relatively low levels of reproducibility.
By the same token, environmentsin which
changeis turbulentand uncertainmay not constitute a systematic regime of selection. The
traits that are favored may shift, frequently
enough that no clear trend emerges. Such settings may favor organizationalforms that can
take quick advantageof new opportunitiesand
the appearanceof new habitats. The capacity
to respond quickly to new opportunitiespresumably competes with the capacity to perform reliably and accountably (Brittain and
Freeman, 1980; Freeman, 1982). Such
dynamics may dilute the importanceof reliability and accountability in organizational
selection. We will addressthese issues in subsequent papers.
For all of these reasons, it is not sufficientto
assume that selection processes favor organizations with high inertia and to proceed as
thoughobservedpopulationscontainonly such
organizations.These considerationslead naturally to considerationof systematic variation
within populations in the strength of inertial
pressures. Existing theory provides some insights into these matters. One line of reasoning, which we pursued, suggests that inertial
pressures increase with age-that organizations tend to ossify as they grow older. We
suggest that the more fundamentalprocess is
that reproducibilityincreases with age. It follows from our general perspective that the
death rate declines with age.
The effects of size on inertiaare problematic

163

in our revised theory. It is widely agreed that

larger organizationsare.more ponderousthan
small ones. We think that analysis of the effects of size on inertia must consider several
kinds of transitionrates. One is simplythe rate
(in an absolute time scale) of attemptingfundamentalstructuralchange. Anothertransition
concerns success in implementing change.
Thereis also the effect of attemptingchangeon
the death rate. We argue that small organizations are not only morelikely thanlargeones
to attemptchange, but are also more likely to
die in the process. Withoutfurtherinformation
on the magnitudesof the rates, it is not clear
whether small or large organizations have
higher overall rates of successfully implementingchange. Ouranalysissuggeststhat it is
prematureto conclude that ecological theory
may be appliedmore readilyto smallthanlarge
organizations. Clearly this matter deserves
more theoretical and empiricalattention.

REFERENCES
Aldrich, Howard E.
1979 Organizations and Environments. Englewood Cliffs, NJ: Prentice-Hall.
Arrow, KennethJ.
1974 'The Limits of Organization.New York:
Norton.
Astley, W. Grahamand Andrew Van de Ven
1983 "Centralperspectivesand debates in organization theory." AdministrativeScience
Quarterly28:245-73.
Becker, Gary S.
1975 HumanCapital.Secondedition.New York:
ColumbiaUniversity Press.
Blau, Peter M. and W. RichardScott
1962 Formal Organizations. San Francisco:
Chandler.
Burawoy, Michael
1979 ManufacturingConsent: Changes in the
Labor Process Under Monopoly Capitalism. Chicago: University of Chicago
Press.
Brittain,Jack and John Freeman
1980 "Organizationalproliferationand densitydependentselection." Pp. 291-338 in John
Kimberlyand Robert Miles (eds.), Organizational Life Cycles. San Francisco:
Jossey-Bass.
Carroll,Glenn R.
1983 "Concentration and specialization: dynamics of niche width in populations of
organizations."Unpublishedmanuscript.
Carroll,Glenn R. and Jacques Delacroix
1982 "Organizationalmortalityin the newspaper
industries of Argentina and Ireland: an
ecological approach."AdministrativeScience Quarterly27:169-98.
Chandler,Alfred D.
1977 The Visible Hand:The ManagerialRevolution in American Business. Cambridge:
Belknap.

164
Child, John and Alfred Kieser
1981 "Developmentof organizationsover time."
Pp. 28-64 in WilliamH. Starbuckand Paul
C. Nystrom (eds.), Handbook of OrganizationalDesign, Volume 1. Oxford:Oxford
University Press.
Coleman,James S.
1974 Power and the Structureof Society. New
York: Norton.
DiMaggio,Paul J. and WalterW. Powell
1983 "The iron cage revisited: institutional
isomorphismand collective rationalityin
organizationalfields." AmericanSociological Review 48:147-60.
Downs, Anthony
1967 InsideBureaucracy.Boston:Little, Brown.
Edwards,Richard
1979 Contested Terrain:The Transformationof
the Work Place in the TwentiethCentury.
New York: Basic Books.
Etzioni, Amitai
1975 The ComparativeAnalysis of ComplexOrganizations. Second edition. New York:
Free Press.
Freeman,John
1982 "Organizational life cycles and natural
selection processes." Pp. 1-32 in BarryM.
Staw and Lawrence L. Cummings(eds.),
Researchin OrganizationalBehavior, Volume 4. Greenwich,CT: JAI Press.
Freeman, John, Glenn R. Carrolland Michael T.
Hannan
1983 "The liabilityof newness: age-dependence
in organizationaldeath rates." American
Sociological Review 48:692-710.
Freeman,John and MichaelT. Hannan
1983 "Niche width and the dynamicsof organizationalpopulations."AmericanJournalof
Sociology 88:1116-45.
Hannan,Michael T. and John Freeman
1977 "The populationecology of organizations."
AmericanJournalof Sociology 82:-929-64.
Hawley, Amos H.
1968 "Humanecology." Pp. 328-37 in David L.
Sills (ed.), InternationalEncyclopedia of
the Social Sciences. New York:Macmillan.
Hermann,C. F.
1963 "Some consequences of crisis which limit
the viabilityof organizations."Administrative Science Quarterly8:61-82.
Janowitz, Morris
1960 The ProfessionalSoldier.Glencoe, IL: Free
Press.
Ladde, G. S. and D. D. Mijak
1976 "Stability of multispecies communitiesin
randomly varying environments."Journal
of MathematicalBiology 2:165-78.
Lawrence, Paul and Jay Lorsch
1967 Organization and Environment. Cambridge:HarvardUniversity Press.
March,James G.
1981 "Footnotes on organizational change."
Science Quarterly
Administrative
26:563-97.

AMERICAN SOCIOLOGICALREVIEW
March,James G. and Johan P. Olsen
1976 Ambiguity and Choice in Organizations.
Bergen, Norway: Universitetsforlaget.
McKelvey, Bill
1982 Organizational Systematics. Berkeley:
University of CaliforniaPress.
May, Robert M.
1974 Stability and Complexity in Model
Ecosystems. Second edition. Princeton:
PrincetonUniversity Press.
Mayhew, L. B.
1979 Surviving the Eighties. San Francisco:
Jossey-Bass.
Merton, Robert'K.
1957 Social Theoryand Social Structure.Second
edition. Glencoe, IL: Free Press.
Meyer, John W. and Brian Rowan
1977 "Institutionalized organizations: formal
structureas myth and ceremony." American Jourpalof Sociology 83:340-63.
Nelson, RichardR. and Sidney G. Winter
1982 An Evolutionary Theory of Economic
Change. Cambridge:Belknap.
Parsons, Talcott
1960 Structureand Process in ModernSociety.
Glencoe, IL: Free Press.
Perrow, Charles
1979 Complex Organizations:A CriticalEssay.
Second edition. Glencoe, IL: Scott Foresman.
Pfeffer, Jeffrey and GeraldSalancik
1978 The ExternalControl of Organizations:A
Resource Dependence Perspective. New
York: Harper& Row.
Scott, W. Richard
1981 Organizations:Rational,Natural,and Open
Systems. Englewood Cliffs, NJ: PrenticeHall.
Simon, HerbertA.
1962 "The architecture of complexity." Proceedingsof the AmericanPhilosophicalSociety 106:67-82.
?iljak, D. D.
1975 "When is a complex ecosystem stable?"
MathematicalBioscience 25:25-50.
Stinchcombe,ArthurS.
1%5 "Social structure and organizations."Pp.
153-93 in JamesG. March(ed.), Handbook
of Organizations.Chicago:RandMcNally.
1%8 ConstructingSocial Theories. New York:
Harcourt,Brace & World.
Thompson,James D.
1967 Organizations in Action. New York:
McGraw-Hill.
Williamson,Oliver E.
1975 Marketsand Hierarchies:Analysisand AntitrustImplications.New York:Free Press.
Weber, Max
1968 Economy and Society: An Outline of InterpretiveSociology. Three volumes. New
York: Bedmeister.
Weick, Karl
1976 "Educational organizations as loosely
coupled systems." AdministrativeScience
Quarterly21:1-19.