Xll ACKOWEDGMENTS

I have received an enormous amount of detailed and helpfl fedback on this

manuscript. Among the many who kindly ofered me help I would like to ac
knowledge Gerard Alexander, Karen Ater, Frank Baumgartner, Tim Buthe,
John Carey, Daniel Carpenter, Elisabeth Clemens, David Collier, Pepper
Culpepper, Jefey Frieden, Peter Gourevitch, Anna Grzymala-Busse, Dava
shree Gupta, Alan Jacobs, Ron Jepperson, Michael Jones-Correa, Ira Katznel
son, Margaret Levi, Christopher Mantzavinos, Eileen McDonagh, Andrew
Moravcsik, Gerardo Munck, Dietrich Rueschemeyer, Jefey Sellers, Alec
Stone Sweet, and Julian Zelizer. I appreciate their generosity in sharing their
time and insights. I cannot possibly express te extent of my gratitude to a circle
of scholars who have, over many years, provided a wonderfl mix of prodding,
encouragement, and trenchant critique that helped to make this a better book:
Danny Goldhagen, Jacob Hacker, Peter Hall, Jim Mahoney, Theda Skocpol,
Allison Stanger, and Katy Thelen. I am lucky to have such fiends.
My children, Sidra and Seth, were merciflly unaware that dad was writing a
book at all (though Seth, in the past few months has become profcient enough
with words that he can accurately describe his fter's job as "sitting at the table
and drinking cofee") . They have, however, been exuberant expressions of the
wqnders of development, and teir rapid growth has been a constant reminder
tat writing this book has most defnitely been a slow-moving process. Finally, I
dedicate this book to my wife, Tracey. She is the master juggler who stands with
grace at the center of our personal three-ring circus, and of my lif.
Permission to use the fllowing materials fom previously published articles is
grateflly acknowledged.
Parts of Chapter One appeared in "Path Dependence, Increasing Returns, and
Political Science," American Political Science Review, Vol. 94, No. 2, June
2000, pp. 251-67.
Parts of Chapter Two appeared in "Not Just Wat, but When: Timing and Se
quence in Political Processes," Studies in American Political Development,
Vol. 14, No. l, 2000, pp. 73-93.
Parts of Chapter Three appeared in "Big, Slow-Moving, and . . . Invisible:
Macrosocial Processes in the Study of Comparative Politics," in James Ma
honey and Dietrich Reuschemeyer, eds. , Compartive Historcal Analysis
in the Social Sciences (Cambridge: Cambridge University Press, 2003),
pp. 1 77-207.
Parts of Chapter Fur appeared in "The Limits of Design: Explaining Institu
tional Origins and Change," Goverance, Vol. 1 3, No. 4, 2000, pp. 475-99.
Parts of the Conclusion appeared in "From Aea Studies to Contextualized
Comparisons," in Grzegorz Ekiert and Stephen Hanson, eds. , Capitalism
and Democracy in Centrl and Easter Eurpe: Assessing the Lgacy of Com
munist Rule (Cambridge: Cambridge Universit Press, 2003), pp. 353-66.
Intrduction
PLCING POLITICS IN TIME
For an economic historian, time has always been something
that is fndamentally distrbing, because there is no time in
neoclassical teory. The neoclassical model is a model of an
instant of time, and it does not therefre take into account
what time does .... I will be blunt: Witout a deep
understanding of time, you will be lousy political scientists,
because time is te dimension in which ideas and
institutions and beliefs evolve.
-Douglass Norh (1999, p. 316)
\ CAN BEGIN WITH A ANALOGY. Imagine that your fiend invites you to the
trendiest new restaurant in town, charmingly named "The Modem Social Sci
entist." A an added bonus, he infrms you tat he knows the chef well, and that
you will have a chance to tour the kitchen. Wen you arrive, the chef explains
tat te kitchen is divided into two parts. On te lef, she has all the ingredients
(which to your puzzlement she refers to as "variables"). These ingredients, she
insist, are the feshest available and careflly selected. On the right is an exraor
dinary profsion of measuring devices. You express astonishment at their com
plexit and detailed ornamentation, and te chef exlains tat each requires years
to learn how to operate properly.
Te chef proceeds to elaborate her culinary approach: good cooking, she says,
amounts to having the perfect ingredients, perfectly measured. Traditional cooks
have stressed how important te cooking process itelf is, including the sequence,
pace, and specifc manner in which the ingredients are to be combined. Not so,
says the proprietor of The Modem Social Scientist. A long as you have the cor
rect ingredients and they are properly measured, she insist, how, in what order,
and fr how long they are combined makes no diference.
Few would want to patronize a restaurant wit such a philosophy of cooking,
but most social scientist are working in tat kind of a kitchen. Disputes among
competing teories center on which "variables" in te current environment
generate important political outcomes. How does the distribution of public
opinion afect policy outcomes? How do individual social characteristics infu
ence propensities to vote? How do electoral rules afect the structre of part sys
tems? Yet the signifcance of such "variables" is fequenty distorted when they
are ripped fom teir temporal context. Contemporar social scientists tpically
2 INTRODUCTION
take a "snapshot" view of political life, but there is ofen a strong case to be made
fr shifing fom snapshots to moving pictures. This means systematically situat
ing particular moments (including te present) in a temporal sequence of
events and processes stretching over extended periods. Placing politics in time
can greatly enrich our understanding of complex social dynamics.
This book explores a range of temporal processes tat are common in political
l!f. It seeks to distinguish various processes tat unfld over substantial stretches
..me, to identif the circumstances under which such diferent processes are
likely to occur, and to highlight the signifcance of these temporal dimensions
of social lif fr our understanding of important political outcomes. In doing so,
I seek to demonstrate the ve hi h rice tat social science ofen a s when it
ignores the profund temporal dimensions of real social rocesses. The am 1-
thort, 1s to es out the ofen-invo e ut rarely examined declaration
that history matters.

It is no accident that so many of the giant fgures in the frmative period of

the social sciences -fom Marx, Tocqueville, and Weber to Polanyi andSchum
peter-adopted deeply historical approaches to social explanation. This stance
was not simply a quaint fature of a transitional stage to modern social analysis:
It was a key source of their profund insights about the nature of the social
world. Attentiveness to issues of terporality highlights aspects of social lif that
are essentially invisible fom an ahistorical vantage point. Placing politics in
time can greaty enrich bot the e lanations we ofer fr social outcomes f -
t$rest, an e very outcomes that we identif as wort explaining. The system
atic examination of processes unflding over time warrants a central position in
the social sciences.
'o ILLUSTRTIONS
Because I am eager to contribute to the eforts of social scientists working on a
wide range of matters, I initially discuss many of the issues addressed in this book
at a firly high level of abstraction. It may be of some help at te outset to briefy
outline to examples of recent scholarship tat ofer compelling examples of
the insights to be gained by shifing fom a "snapshot" to a moving picture of im
portant social processes. Daniel Carpenter's The Forging of Bureaucratic Auton
omy (Carpenter 2001 ) presents a striking demonstration of h attention to a
long-term sequence of causes (what I will call a "causal chain" in Chapter Thre)
can turn our understandings of social phenomena on their heads. He criticizes
the large and infuential literature on relations between legislatures and bu
reaucrats grounded in Principal-Agent theory (McCubbins, Noll, and Weingast
1987; McCubbins and Schwartz 1984). The literature argues that congressional
"prin

ls" have substantial political resources to assure that their bureaucratic

"agents" largely comply wit teir prefrences. Yet Carpenter persuasively demon-
PLCING POLITICS IN TIME 3
strates how these anal
.
ses substantiall underestimate the otential fr bureau- j\
cratic autonom because te ado t a cross-sectional approach to stu ying w a
should be understood as a long-term causal chair. Un er e ng con 1 10ns,
ambitious and entrepreneunal bureaucrat wer able over extended periods of
time to enhance their reputations fr innovativeness and competence, and de
velop strong networks of support among a range of social actors. These achieve
ments created a context in which Congress, fcing pressure fom below and
deferring to te expertise of leading bureaucrats, essentially asked fr what the
bureaucrats wanted. Viewed as a moment in time, one sees what looks like con
gressional dominance; viewed as a process unflding over time, the same cross
sectional evidence provides indications of substantial bureaucratic autonomy.
Thomas Ertman's The Birth of Lviathan (Ertman 1 996) ofers a compelling
exlanation of diferent patterns of state-building in early modem Europe that
emphasizes how the results of critical j unctures, and cross-national diferences
in the sequencing ofkey historical processes, generated durable variations in te
structures of nation-states, Ertman emphasizes the lasting repercussions of the
fnancing .metods adopted in European states that fced military competition
befre the rise of modem bureaucracies. Following predecessors such as Hintze
and Tilly; Ertman argues that the onset of military competition had a critical e_
fet on paterns of Eurean state deveopment. Crucially, howeer, Ertman
stresses that diferent states experienced intense militar confict at diferent times.
He maintains (p. 26) tat "diferences in te timing of te onset of sustained
geopolitical competition go a long way towards explaining the character of state
infastructures fund across the continent at the end of the 18t century."
States confonting militar competition fced the life-or-death challenge of
generating sufcient revenues to wage prolonged warfare. However, te available
repertoin:s of administrative response were conditional on te point in historical
development when this challenge appeared. According t Ertman, ''timing mat
terep because the range of'technical resources' available to statebuilders did not
remain invariant across this period." In the twelfh century, literacy was a very
scarce resource and sophisticated bureaucracy an unknown organizational tech
nology. In tis historical context, monarchs were frced to rely on systems of
. proprietary ofce holding and tax frming "which were much more benefcial
to [these ofceholders] than to their royal employers." By contrast, countries that
fced intensive military competition at a later time "fund themselves in a quite
diferent world" (p. 28). Literacy was much more common, and knowledge of
more modern bureaucratic organizational frms (as well as evidence of tax
frming's considerable drawbacks) was widely available. The result was that state
builders in these countries could work to construct "proto-modern bu. reaucra
cies based upon the separation of ofce fom te person of te ofceholder."
Why, ten, did not all states adopt te superior bureaucratic stuctures? Because,
Ertman argues, initial outcomes were strongly self-reinfrcing. Once a dense
network of institutions and interests developed around tax frming, especially in
4 INTRODUCTION
a context where monarchs ofen had immediate needs fr revenues, it became
virtually impossible to switch over to more modem frms of fnancing. In the
terms I will introduce shorty, te experiences of diferent countries were highly
path dependent. And these diferent paths mattered a great deal, Ertman main
tains, because the bureaucratic alternative was both more efective in waging
war and more conducive to the development of parliamentary institutions.
Ertan's powerfl analysis, like Carpenter's, is built on an understanding that
social processes, such as state building and the construction of bureaucratic au
tonomy, unfld over time. Just as Carpenter illuminates temporally extended
but crucial causal chains, Ertman employs elements of temporalit explored sys
tematically in the chapters to fllow: pat dependence (Chapter One), se
quencing (Chapter Two), and slow-moving processes (Chapter Three). Fr both
authors, the tum to history is not primarily a matter of employing narrative or
compiling historical evidence; it is the grounding fr theoretical claims about
how things happen in the social world.
"HISTORIC TURN" IN THE SOCIA SCIENCES;
A these illustrations attest, the social sciences have had a rich tradition of his
torical research. Scholarly communities devoted to extending such taditions
fourish in parts of the social sciences. Indeed, some (McDonald 1 996; Bates et
al. 1998) claim to witness a\'historic tum" in the human sciences as a wholYet
in spite of this activit there has actually been surprisingly limited attentiveness
to te specifcally temporal dimensions of social processes. In contemporary so
cial science, the past serves primarily as a source of empirical material, rather
than as a spur to serious investigations of how politics happens over time.
At least within political science, the adoption of a historical orientation has
generally filed to exploit its greatest potential contribution to the more system
atic understanding of social processes. ne can in fct distinguish three promi
nent "historic turns" in the social sciences. Each of these has added signifcantly
to our store of knowledge, but each has serious limitations as well. The frst, es
pecially prominent in the feld of American political development, might be
termed "histor as the study of the past." Here analysts study particular historical
events or processes, with a fcus on ofering convincing causal account of spe
cifc outcomes of interest. Such investigations ofen greatly increase what we
know about particular fcet of te political histories of specifc countries.
What is less clear, however, is how particular studies ft into some broader re
search program. Litte efort is made to suggest what, if anyhing, might "travel"
fom one investigation to anoter. Indeed, many historically oriented analysts
are uninterested in tis question, assuming te stance of most historians: that te
rich particularities of each event or process render it unique. Alternatively, these
analysts seem to assume implicitly that a discussion of, say, social movements in
PICING POLITICS IN TIME 5
the 1 920s generates clear implications fr our understanding of contemporary
social movements. Such an assumption is highly problematic. This frst histori
cal tum is ofen admired. Yet it is fir to say that the great maj orit of social sci
entists, whose principal interests lie in understanding contemporary society, sees
"history as the study of the past" as largely irrelevant to their own inquiries.
A second "tum" could be termed "histor as te hunt fr illustrative material."
Here te researcher's perspective could uncharitably be summarized as fllows:
I "I have a model of some aspect of politics. Let me search the past fr a good il
lustration of the model." jis te of exercise is especially common among
those aspiring to generate very general propositions about politics. Thus, ra
tional choice theorist-contemporary social science's most ambitious pursuers
of general propositions -comb the past fr examples of credible commitment
mechanisms or fr solutions to a particular class of collective action problems.
Given a wealth of available historical material and the ofen highly stlized ac
count generated in these exercises, te past may well ofer up te desired illus
trations. Here again, however, the analyses say litte or nothing about the
temporal dimensions of social processes. The motivation fr going back in time
is simply to get at examples that may not be available in the present.
/This is also the case of te third tum, which could be termed "history as a site
fr generating more cases." }istory becomes a source of data, especially fr
phenomena tat are relative
i
y uncommon in the contemporar environment
(Bartolini 1993, p. 14). There are deep controversies about this methodologi
cal move, whether it takes the frm oflarge-n stdies (e. g. , using pooled time se
ries) or small-n studies (e.g. , using some variant of Mill's methods of agreement
and disagreement) . 1 I share some of these critics' misgivings. My main point,
however, is a diferent one, and it applies to all three of the trns to histor tat
I have discussed. The best case fr connecting histor to the social sciences is
neither empirical (tum #1 ) nor methodological (turns #2 and ?),but teoreti
cal. We tum to an examination of history because social life unflds over time.
Real social processes have distincty temporal dimensions.
Yet an exploration of these temporal dimensions of social processes is pre
cisely the weakest link in social science's historical tm. We largely lack a clear
outline of why te intensive investigation of issues of temporality is critical to an
understanding of social processes. The declaration that "history maters" is ofen
invoked, but rarely unpacked. 2 Many of the key concept needed to underpin
analyses of temporal processes, such as path dependence, critical junctures,
10n the frmer see Shalev 1999. For diferent views on te latter see Lieberson 1985 and Mahoney
1999.
2There are notable exceptions, to which te current analysis is deeply indebted. Aong tese I
would single out Collier and Collier 1991; Katnelson 1997; Mahoney 2000; Mahoney and
Rueschemeyer 2003; Orren and Skowronek 1994; Skocpol 1 992; Stinchcombe 1 965, 1968; Thelen
1999, 2003. The excellent essays collected in Abbot 2001 cover some of the territory I take up here,
albeit in very diferent fshion.
6 INTRODUCTION
scgucncing, cvcnts, duration, Iiming, and unntcndcd conscgucnccs, havc rc-
ccivcdonyvcryhagmcntcdand imitcddiscussion.To asscrtthathistorymat-
tcrsisinsumcicnt,sociascicntistswanttoknowwhy,whcrc,how,andhrwhat.
A thcsocioogistKonad^minzadehasputit (1992, p.458) , wcscckthccon-
structionoIthcoricsoIcontinuityandchangcthatarcattcntivctoordcrandsc-
gucnccandthatacknowcdgcthccausapowcroItcmporaconncctionsamong
cvcnts. Jhis rcguircs conccptsthat rccognizc thc divcrsityoIpattcrnsoItcm-
poraconncctionsamongcvcnts.
Vhatisat stakc in this chort torchncour corctica undcrstandings oIthc
dihcrcntwaysinwhichhistorymattcrsincxpainingsociaphcnomcnaNost
important, cxamining tcmpora proccsscs aows us to idcnti and

cxpicatc
somcmndamcntasociamcchanisms. hymcchanismsmcan whatJonstcr
has tcrmcdpausibc, hcgucnty obscrvcd ways in which things happcn. ^
numbcroIschoarshavcrcccntycmphasizcd atwhcrc possibc-and itwi
not aways bc possibc-it is cxtrcmcy hcpmto idcnti|hcgucntyrccurring
causamcchanisms. JonLstcrhasputitmoststrongyinarguinglhatthcbasic
conccpIin thc socia scicnccs shoud bc that oIa mcchanism rathcr thanoIa
tcory . . ..

J|hcsociascicnccsarc ightycarsaway homthc sIagcat which it

wi bc possibc tormuatc gcncra-aw-ikc rcguaritics about human bchav-
ior. nstcad, wc shoud conccntratc on spcci|ing sma and mcdium-sizcd
mcchanismshrhumanactionandintcraction[stcr 1 989, p. viii).`
U
JhcmainhcusoIuisbookisonsociamcchanismsthathavcastrongtcmpo- _
dimcnsion.^moughhistoricayoricntcdschoarsarc[righdy)skcpticaabout
c prospcctshr gcncrating anythingikca gcncra thcoryoIpoitics, mostso-
iascicntistsrcmainin

tcrcstcdindcvcopi

ngatcastimitcdgcncraizations-

rgumcntsthatcantravcinsomchrmbcyondaspccihctimcandpacc.Jhc
currcntstudyscckstoaddrcssisgapbyoutiningmcchanismsathavcasbong
tcmporadimcnsion.xporingthccharactcroIthcscmcchanisms,andthcha-
turcsoIsociacontcxtsthatgcncratcthcm,canthussimutancousyaddrcssthc
dcsircoIanayststomovcbcyondasingccascandhcdcsircoIhistoricayori-
cntcd schoarstocapturchowhistorymattcrs. Jhcidcntihcationand carihca-
tion oIsuch mcchanismscan cnhancc ourabiito dcvcopargumcnts about
tcmporaproccsscsatarcbothconvincngandhavcatcastimitcdportabii.
Lxporingthcscmcchanisms,wisuggcst,cancadustorcasscsspromincnt
arcas oIsocia scicncc inguiry and convcntiona practiccs in ncw and hrtic
ways. t wi ohcn suggcst ncw hypothcscs rcgarding important subcct and
opcncxcitingpossibiiticshrcxtcndingcxisting thcorctica workin ncwdircc-
3For additional discussions see Scharpf 1 997, chap. 1, and Stinchcombe 1 991 .
4lndeed, one of the stiking features of much of te qualitative historical literature i n the feld of
Aerican political development, fr example, is the limited capacity of studies to actually build on
their predecessors, rater than simply piling up one afer another. For a discussion of the possibili
ties fr enhancing the qualit of research programs in comparative historical analysis, see the essays
in Part One of Mahoney and Rueschemeyer 2003.
PLACING POLITICS IN TIME
7
tions.A hcusonIhcsctcmporayoricntcdmcchanismssuggcstsncwgucstions
andrcvcasncwoutcomcsoIintcrcst-gucstionsandoutcomcsthat arcinkcd
to, butdistincthom, cxistingincsoIinguip. inay, thcintcnsivccxporation
oIthctcmporadimcnsionsoIsociaproccsscskrccsustorcthinkourstratcgics
oIsociainvcstigation.cannotcmphasizccnoughthatthisisnot abookabout
mcthods. `ct how you choosc toook hr ings dcpcnds hcaviIy on what you
think you arc ooking mr.cncc, many oIthc argumcnts cxporcd hcrc havc
substantiamcthodoogica impications a2003).
Jhcprincipa audicncc hr this book consist oIthosc intcrcstcd inthcat-
tcmpttodcvcopcaimsaboutthcsociawordatcanpotcntiayrcachacross
timc and spacc.1ct cvcnhr schoars who arc dubiousabout c ros ccts hr
gcncraizationorunintcrcstcdini pursuit,thcorcticacxporationsqhistorica
causationrcmainimportant. !tiscasytoundcrcstimatcccxtcnttowhichIhc-
orcticadiscussionsundcrpin, iIonyimpicihyandbywayoIdihsion, acm-
piricarcscarch.A ritz bcharpI( 1997, p. 29) puIsit,
inawordthatiscxcccdingycompcxandnwhichwcwiohcnbcstudy-
inguniguccascs,wcmusthavcagoodidcaoIwhattoookhriIwcwishto
discovcranythingworwhic. binccasingcdatapointcanbccxpaincd
by anynumbcroIrcgrcssionincs,posthoccxpanationsarctoocasytoin-
vcntandusuay[uncssinvcntcdwidthctraincdskioIcmastcrhisto-
rian) totay usccss. Jhc impication is that our scarch hr cxpanations
mustbcdiscipincdbystrongprior cxpcctationsanddatwcmustIakcc
disconhrmationoIsuchcxpcctationsasawccomcpointcrtodcdcvcop-
mcntoImorcvaidcxpanations.

ngcncra,wcwibcinamuchbcttcrpsitontocarryoutconvincingrcscarch

iIwc

inkthroughthcscprior

cctatio

s-

and

ourust

ihcationsIrh. oding
mcm-cxpicidy and carcmy.

ariing how various tcmporay groundcd

mcchanismsopcratcand suggcstin whcrcwcshoud cxpcctsuch proccsscs to
bc aI work can providc strongcr thcorctica undcrpinnings r c anayscs oI
cvcn thoscsccingtocxpainasingccasc. _
^chicvinggrcatcrcarity abouthowhistoryimpartsitschcctsonthcprcscnt
wiopcnuppossibiticshrmorcconstructivcintccctuadiaoguc. !twihc_
historicayoricntcdschoarsbcmorcchgyin communicatin_thcim_ortof
thcr rcscarch to cach othcr aswcasto ohcnskc tica coca ucs.Jhcsocia
scicnccs arc i h bakanizc n is contcxt, Ihcrc is much to bc
saidRrchortsthatsccktoarticuatccaspirationsandachicvcmcntsoIabody
oIrcscarchinaanguagcatmakcsscnsctomoscoutsidcmctribc.`Jhisisnot
5Even witin "tribes;' communication is ofen hindered by te lack of a common conceptual lan
guage tat allows scholars to recognize common interests -as, fr instance, when scholars studying
diferent areas of te globe lack a shared vocabulary to recognize teir overlapping concerns. On tis
see Mahoney and Rueschemeyer 2003.
8 INTRODUCTION
j ust about making oneself understood or about clearing away flse disagree
ments -altough tese are important benefts. At least as signifcant, greater
clarity and precision can reveal signifcant point of overlap and identif oppor
tunities fr usefl exchange as well as highlight genuinely contentious issues.
In tis context it makes sense to briefy situate the current discussion vis-a-vis
to prominent schools of thought in contemporary political science: "rational
choice teory" and "historical institutionalism." Such an efort is hazardous, since
tese are loose camps-"theoretical imageries," as Ronald Jepperson ( 1996) has
useflly put it-rather tan coherent theories. Yet given the tendency of many
scholars to orient their work in allegiance to one camp (and ofen in opposition
to oters) it may be usefl to make my own position clear: a fcus on te tem
poral dimensions of social processes largely cuts across this divide. The sysem-
atic examation of temporal process

can useflly draw on rich contibutions
f

m each tradition, while also highlighting certain limitations. It can help to

Cstinguish pc of ovea
e
, instances of ge

ine disagent, and substantial

areas in which the two traditions simply speak to diferent {although perhaps
complementary) questio
E
s. My claim is tat a fcus on temporal processes can
point to fuitl lines of theoretical, metodological, and substantive inquiry fr
those working witin each of tese "imageries" -although these lines will and
should remain distinctive in important respect.
One would think tat the issues at hand would be the natural terrain fr "his
torical institutionalism," wit its long-standing insistence that social science re
search should be historically grounded (Skocpol and Sommers 1980; Thelen
and Steinmo 1 992; Kaznelson 1 997). Yet those associated with historical insti
tutionalism h

e generally been more explicit in discussing the "institutionalist"

dimensions of teir fameworks tan te "historical" ones. They have concentated
on examinig tie impact of relatively fxed institutional features of the politi
cal landscape (such as constitutional arrangements and major policy struc
tures) . The specifcally historical component of historical institutionalism has,
at least until recenty (Hacker 2002; Thelen 2004) generally been lef unclear.
A already noted, if a justifcation fr a tum to history is given at all, it has more
ofen been methodological rater than theoretical in nature.
Thinking more explicitly about the role of time in politics will, however,
j ustif some of the key concerns and ofer support fr some key propositions
advanced by historical institutionalist scholarshiQ. It will provide stronger the
oical grounds fr emphasizing the ftickines_of inherited social arrange
ments, fr questioning fnctional explanations (that is, claims that social
arrangements exist because they meet certain needs of societies or particular
powerfl actors) , fr concentrating on issues of timing and sequence, and fr
investigating long-term processes of social change. More broadly, this investi
gation will vindicate historical institutionalism's interest in macrolevel social
phenomena, as well as its cautious stance regarding social science's capacit to
PIACING POLITICS IN TIME 9
develop broad generalizations about social processes that apply across sweep
ing stretches of time and space.
At the same time, I stress that rational choice analysis, broadly defned, ofers
essential analytical tools fr inv
;
stigating temporal processes. In this book I de
velop theoretical arguments organized around fur ma'or themes: pa
t

pendence, issues o 1mmg an sequence, the signifcance and distinctiveness of

"slow-moving" processes that require attentveness to exended eT
an pro ems o institutiona ori ins an c an e. On each of these points, work
drn om economic theory, fcusing on problems of strategic interaction
among calculating, rational individuals, has much to ofer.
Yet a fcus on the temporal dimensions of political processes highlights the
limits of choice-theoretic approaches as well as their stengs. Most important
is the way in which micromodeling exercises tat are centered on strategic in
teraction among individuals encourage a highly restricted feld of vision, bot in
space and time. Aong the tings that tend to drop out of such exercises are is
sues of macro structure, the role of temporal ordering or sequence, and a whole
host of social processes that play out only over extended periods of time and can
not be reduced to the strategic "moves" of "actors." Thus, a restricted feld of vi
sion adversely afects both the kinds of questions rational choice theorists tend to
generate and the kinds of answers they tpically provide. The critique, it should
be emphasized, is not the common (in my view ofen misguided) one about ra
tional choice theory's assumptions about human behavior, but about its re
stricted range of application. The implication is tus not that rational choice
theor should be rejected, but tat its scope should be placed in proper per
spective. Aalysts should fcus on establishing how insight fom rational choice
can be linked to other approaches, or where other approaches are simply more
appropriate fr addressing particular kinds of questions (Jepperson 1996) .
Clearer recognition of some of rational choice's blind spots may also fcili
tate eforts within that tradition to addres. s challenging problems in convincing
ways. In the pages that fllow I will ofen criticize strong tendencies associated
with particular techniques or theoretical approaches, while accepting-indeed
emphasizing-that there is nothing about these modes of inquiry that renders
these tendencies logically necessar. Analyst will sometimes seize on tis point to
suggest that the critique has litte import.6 This reaction is a mistake. The ques
tion, afer all, is not just what a particular technique or teory is capable of doing in
principle, but how and to what extent it is actally used in practice-a distinction
tat is almost always glossed over in general discussions of method and teor. To
paraphrase a recent observation by Ronald Jepperson, when fced wit a causal
account employing a long-term temporal structure, quantitative scholars (as well
6his response has been common, altough fr fom universal, in te reactions of rational choice
scholars to some of te arguments presented in tis book.
1 0 INTRODUCTION
as somc rationa-choicc thcorisL) might wc rcspond by cxcaiming, can
modcthat! Jhisripostc oscs mucholitsbitc, howcvcr, iIthcthcorctica im-
agcricsandmcthodsthcscanaystscmpoyrarcycadthcmtoconsidcrthcout-
comcsorhypothcscsingucstioninthchrstpacc [Jcppcrson 1 996).
JhcthrustoIthcargumcnbinPolitics in Time istorcconsidcrawidcrangc
oI thcorctica approachcs and mcthodoogica tcchnigucs hom thc vantagc
point oIissucsrcatcd totcmporaity. ndoing so, hopc tochacngc somcoI
thctriba,poarizcdcharactcroImuchcontcmporarysociascicncc.A hcuson
thc tcmpora dimcnsions oIsociaproccsscsprovidcscxciting possibiitics hr
shapingsomccommon-oratcastovcrapping-intccctuatcrrainhrscho-
arsworkingoutoIhighydivcrscrcscarchbaditions. nthcrcmaindcroIthisin-
troductionbricny summarizc mychortstomapoutthattcrrain.
ANALYTICAL FOUNDATIONS
JhchvcchaptcrsthatIrmmccorcoIthisbookcxporcsomccytcmporadi-
mcnsions oI socia and poitica proccsscs. Jhcy cxpicatc dihcrcnt ways in
which things happcnovcrtimcinsociaih, drawing ahcntionto proccscs that
arc unikcy to bc visibc wimoutspccihcayaddrcssing gucstionsoItcmpora-
ity. ncachcasc,sccktodcmonstratcwhysuchproccsscsarcikcytobcprcva-
cnt, thc circumstanccs undcr which thcy shoud bc cxpcctcd, and thc maor
impicationsRrourchortstoundcrstandsociaoutcomcs.
Chaptcr Lnc hcuscs on path dependence. a dcpcndcncc has bccomc a
Lddishtcrm, ohcnackingaccarmcaning.1ctthcrcarccxcitingncwpossibi-
itics hr a

pying thc notion in a morc rigorous way to thc anaysis oIsocia

proccsscs argucthatthis mcans hcusingon thcdynamicsoIscI-rcinhrcing
orpositivcIccdbackproccsscsinapoiticasystcm. ouchproccsscshavcvcryin-
tcrcsting charactcristics.Jhcy can bc highy innucnccd by rcativcy modcst
pcrturbationsatcarystagcs. Jhus, suchproccsscscanproduccmorcthanonc
outcomc. Lncc a particuar pa gcts cstabishcd, howcvcr, scIrcinhrcing
proccsscs makc rcvcrsasvcry dimcut. n omics, such dcashavc bccc
ncrcasmgy popuar m thcinvcstigation oIncw tcchnoogics [Nicrosoh v.
Nacintosh), tradc,cconomicgcography,andcconomicgrowth.
rawing on thc rcscarch oIhrian ^rthur and Lougas lorth, show how
thcsc idcas can bc cxtcndcd and modihcd to addrcss issucs ofccntra impor-
tancc to poitica scicntists. Jhcrc arc strong grounds hr bcicving at scI-
rcinhrcingproccsscswibcprcvacntinpoiticaiIc-arguabymorcpcnasivc
andintcnscthanthcyarcindccconomicsphcrc.Lncccstabishcd,pattcrnsoI
poitica mobiization, c instihtiona rucs oIc gamc, and cvcn citizcns'
basicwaysotthinkingaDoutcpoiticawordwiohcngcncratcscI-rcinhrcing
dynamics. LnccactorshavcvcnturcdLrdownaparticuarpath, thcymayhnd
PlCING POLITICS IN TIME 11
itvcrydimcuttorcvcrsccoursc. oiticaatcrnativcsthatwcrconccquitcpau-
sibcmaybccomcirrctricvabyost.
L about ath dc cndcncc icay suggcst that bcginnings arc cx-
trcmc_im_ortant. Soonc might ask: Vy cgin is iscussonwith pathdc-
pcndcncc Jhcanswcr is that an undcrstandngoscI-rcinhrcing proccsscs is
cxtrcmcyhcphhrcxporinga widcrangcoIissucsrcatcdtotcmporai[. x-
poring thc sourccs and conscqucnccs oIpath dcpcndcncc hcps us to undcr-
stand thc powcrIu incrtia or stickincss that charactcrizcs many aspccts ol
poitica dcvcopmcnt-mr instancc, thc cnduring conscgucnccs that ohcn
stcmhomthccmcrgcnccoIparticuarinstitutionaarrangcmcnL. Jncsca_u-
mcntscanasorcinvig!cte ana_sisoI_owcr insociarcatonsby soing
hoinc_uaitics oIpowcr,gcrha_smodcstinitia_,ancrcinhrccdovcrtim

c
andohcn comc to bcdccpycmbcddcdin or aniationsanddominantmodcs

0 po itica .ctin an
_
undcr
_
andin , as wc a s(n institutiona arran cmcnts.
Ia - cpcn cnccargumc
_
sasoprovidcauschandpowcrmcorrcctivcagainst
tcndcncics to assumchnctionaistcxpanationshrimportantsociaandpoiti-
caoutcomcs-thcsuppositionthatthccxistcnccoIcurrcntsociaarrangcmcnts
istobccxpaincdthroughrcIcrcncctothcnccdsthcyaddrcssIrthccurrcnty
powcrh. Norcovcr, an apprcciation oI thc prcvacncc oI path dcpcndcncc
hrccs attcntivcncss to thc causa signihcancc oItcmporay rcmotc cvcnts or
proccsscs. ath-dcpcndcnt proccsscs cxcmpi} what ^rthur Stinchcombc ha
tcrmcd historica causation, m which dynamics tricrcd b_ an cvcnL or
proccss aIcrc_r

ducc thcmscvcs, cvcn in thcabscncc oIt

_
c
rccurrcncc oIthcorigina cvcntorproccss. !inay, an apprcciation oIpositivc
Icdback aso justihcs ahcnIivcncss t issucs oI tcmpora ordcring. n path-
dcpcndcnt proccsscs, thc ordcr oIcvcntsmaymac a hndamcnta dHcrcncc.
n a thcsc rcspccts, path dcpcndcncc undcrscorcs thc dtstincty tcmpora di-
mcnsionsoIsociaproccsscs, ayingaIundationhrthcchaptcrsthatIow.
Lhaptcr Jwocxporcs issucsoItiming and seguence ingrcatcr dctai. bocia
scicntiststracingbroadnsoIpoiticadcvco_mcntacrossanumbcrgn
tricsohcnargucthatmctimiigandscqucnccoIparticuarcvcnLorproccsscscan
grcat dca. ochngs whcrccvcntA prcccdcs cvcntB wigcncratc a
hrcntoutcomcsthanoncswhcrcthatordcringisrcvcrscd. nrmansaccount,
hrcxampc, itisthcrcativctimingoIthcarrivaoImassitcracyandthconsct
olintcnsc miitary compctition that is crucia. hc conccpts oItiming and sc-
gucncing arc, howcvcr, morc ohcn invokcd than ccary thought through. Kc-
vicwing somc promincnt work in comgarativc historca, show that
scgucncinargumcnts havc bccn both prcvacnt and, o cn, aayticamud-
dcd.JhcrcarcactuayanumbcroIdistinctwaystomakcsuchcaims. LiHcr-
cnt|pcsoIscgucncingargumcntsarcikcytobcrccvantindihcrcntscttings.
Lari|ing thcsc dihcrcnccs and spcci|ing thc mcchanisms that ink caims
aboutscgucnccstoparticuaroutcomcsarcthcccntratasksoIthischaptcr.
f
1 2 INTRODUCTION
hcusontwobroadcasscsoIargumcnts. LnccassoIargumcntsaboutti

-
i_icncct(s;qs-intcractiohcctsgct
causaccsthatbccomcoincdatparticUarpointsintimc. orinstancc,
it bymattcrcda!rbiccpodpoiticswhcthcr
chwingorrightwingparticshappcncdtobcinpowcratthctimcwhcnacata-
cysmic cvcnt, thc Lrcat Lcprcssion, hit a particuar country. Just as a Ling
brickhasdstinctconscgucnccs whcnitarrivcsatt smtimcasanunIrtu-
nutcpcdcstrian,thcsitanciooproccsstinog_a|diI-
Ic;]im;_critica conscgnccndix's Lmous anaysis oIwhat
happcnswhcnthctwin q1occsscsofmdustriaizationanddcmocratizationoccur
atthcsamctimcisagoodcxampc.
JhchrstpartoILhaptcrJwocxporcsthcpotcntiaandpiasoIsuchcon-
Q(ra ar_ugL hr socia scicntists. istoricay oricntcd schoars ridy

point out thatcxpanations hr important socia outcomcs ohcn rcst at cast in

parton such conuncturcs.^tcsamctimc, howcvcr, thcrcappcartobcrca
imitstoourcapacitouscconuncturacaimstoscarchhrpattcrnsacrosscascs.
Vith modcstcxccptions,suchcaimswoudsccmtobc mostuscmhrundcr-
standing,ahcrmcLct,spccihcoutcomcsoIintcrcst.Jhcysccmcssikcy,how-
cvcr,toyicdanundcrstandingoImcchanismsthatcoudbcappicdinmutipc
scttings.
JhisisnotthccaschrthcsccondcassoIargumcntsaboutscgucncingat
considcr. Nany-probabymost-argumcnt about scgucncing tum out on
coscrinvcstigationtobcgroundcdincaimsaboutpositivchcdback.Jhismakcs
thcmamcnabctomckrmsofnaysisdcvcopcdinmyd
_
ssionoIpathdc-
pcndcncc.crc,timin_andsc_ucnccmattcrbccauscsI-rcinhcsscs
a_aparticuaraspcctoIpoiticaandsociaiIc

antranshrmthc

consc-
gucnccsoI_sinascgcncc.
..
1inking argumcnts about path dcpcndcncc to a Icus on scgucncing pro-
duccs powcrIu mcorctica syncrgics. ath-dcpcndcnt argumcnts about scI-
rcinhrccmcntcxpain why andwhcnscgucncingcanmattcr. ositivcccdback
proccsscsoccurringatparticuar timcscsscntiayrcmovcccrtain optionshom
thcmcnuoIpoiticapossibiitics. hydoingso,thcycangrcadyatcrthcconsc-
gucnccsoIcvcnts or proccsscs occurringataatcrstagc.^tthcsamctimc, thc
spccihchcusonscgucncinggcncratcsarichncwsctoIhypothcscsaboutpath-
dcpcndcntproccsscs. tcan draw attcntion tocontcstsovcr oiti aspacc in
wh_0 gotcntia com ctitors scck hrst-mgdvantagcs, w c cari in thc
_scy ong-tcrm im 1rn ia cIcats on t c oppornL
Lcingnitia oscrs orgroupsthatpointintimc. tdrawsattcn-
tiontothe sigifcane of @c-scJc morta

c
timin_oIcscchangcsrcativctocachocr.urdcrmorc,aIcusonhistorica
cgucnccssuggcsLhow argumcnts aboutpath dcpcndcncccanaddrcsscaims
aboutpoiticachangcaswcaspoiticaincrtia. orinstancc,path-dcpcndcnt
proccsscsmayopcratctoinstitutionaizcpoiticaarrangcmcntsthatprovctobc
PLACING POLITICS IN TIME 1 3
particuaryvuncrabctosomccvcntorproccsscmcrgingataatcrstagcinpo-
iticadcvcopmcnt.

Jhcscargumcntsaboutpath-dcpcndcntscgucnccscanbodrawonandcn-
anccargumcntsthatrationachoicccoristshavcdcvcopcdaboutthc tcm-
pora ordcring oI choiccs in highy institutionaizcd scttings. Vorking hom
^row'sparadoxoIvohng,whichsuggcsLcikcihoodoIcndcsscycinginmany
cocctivc choicc situations, rationa choicc thcorists havc argucd pcrsuasivcy
thatinstitutionaarran cmcnts ovcmin a c troanddccts
pcdurcscan_ro ucgsth@ owm cs.Jhcscinstitutionaarrangcmcntsarc
cruciabccauscincontcxtsvuncrabctocycingitcanbcdcmonstratcdthatthc
scgucncingamongatcrnativcchoiccswidctcrmincthcoutcomc.Jhiswhoc
inc oIargumcnt rcsL on institutiona mcchanisms that gcncratc _ath d
dcncc:stcpsinascgucnccarcirrcvcrsbcbccauscosin atc tivcsarcdro cd
fl rangcoIpossi coptions. hyshowinghowsuchirrcvcrsibiiticscanbc
gcncratcdmawidcrangcolsociacontcxts,howcvcr,itispossibctocxtcndmis
cruciainsighttoafr broadcrsctoIsociaphcnomcnaanthosccovcrcdinmc
itcraturcdcrivcdhom^rrowsworJ/cgucncingcanmattcr_noton_ c-
tccswiminc_isaturcs,buthrany sociarocccnt
cansthatkrsakcnatcrnativcsbccomcincrcasinyic
Itimc dcmonstratc, in comparativc historica anayscs pcwcrm ar-
gut scgucncingarc ohcn appicdnotto mc movcs olactors on a
micro scac, but to cxaminc thc impact oIargc-scac socia changcs such as
dcmocratization,industriaization,orstatcbuiding._
^striingIcaturcoImanyoIthcargumcntsdiscusscdinLhaptcrJwoismat
thcydrawattcntiontocngthy,argc-scachistoricaproccsscssuchasdcmocra-
tizationorstatcbuiding.LhaptcrJhrccturnstoamorcsystcmaticdiscussionoI
big,
g
ow-moving aspccts oIthcsociaword. Ithc prcoccupation oILhaptcrs
_ Lncad Jwo IS gucsonopora o

dcring, cspcciay thc signihcanccoI

bcginnings, thcccntraprcoccu ationhcrciscongdurc.Jhcchaptcrcx-
amincsa widcrangc o proccsscsatcannot
_
c undcrstoduncssanaysts rc-
mainattcntivctothcunhdingoIbocausaproccsscsandimportantpoitica
outcomcs ovcr cxIcndcd pcriods oI timc. Lontcmpora socia scicntists arc
strongyprcdisposcd tohcus on aspcctsoIcausaproccsscsand outcomcsthat
unhd vcry rapidy. 1ct many ings in c socia word takc a ong timc to
happcn-suchasthcsprcadoI itcracyinLrtmansanaysis,orthcchortsoIbu-
rcaucratstobuidthckundationshrautonomousaction inLarpcntcr'sstudy.
JhcLctthatsomcthinghappcnssowydocsnotmakcitunimportant.
Lhaptcr Jhrcc aso cxporcs a rangc oIdihcrcnt causa proccsscs and out-
comcsthatmayundovcrsubstantiastrctchcsoItimc. bomccausaproccsscs
and outcomcs occur sowy bccausc thcy arc incrcmcnta-it simpy takcs a
ongtimchrcmtoadduptoanying.nothcrs,ccriticaLctoriscprcs-
cnccoIrcshodchccL. bomcsociaproccsscsmayhavcithcsignihcanccunti
tcyattain acritica mass, which may cntriggcr maor changc.
Q
cr socia
14 INTRODUCTION
proccsscsinvovcconsidcrabctimcagsbctwccnthcappcaranccoIakcycausa
IctorandthcoccurrcnccoIthcoutcomcoIntcrcst. Jhismaybctrucbccausc
thcoutcomcdcpcndsonacausachain thattakcssomctimctoworkitscIout
(a causcs b, which causcs c .. ). ^tcrnativcy, causa proccsscs may tum on
structura Icaturcs that invovc transhrmations that arc probabiistic during
anyparticuarpcriod,whichmcansthatscvcrapcriodsmaybcncccssarybchrc
thc transhrmationoccurs.Lndcr conditionssuchasthcsc, thcsociaoutcomc
oIintcrcst may not actuay takc pacc unti wc ahcr thc appcarancc oIkcy
causaIctors.
^naystswhoIitobcattcntivctothcscsow-movingdimcnsionsoIsociaih
arcpronctoa numbcroscriousmistakcs. Jhcymayignorcpotcntiaypowcr-
m hypothcscs. Jhcyarcpartcuaryikcy to miss thcrocoImany socioogi-
cavariabcs,ikcdcmography,itcracy,ortcchnoogy. hcircxpanationsmay
hcusontriggcringorprccipitatingIctorsrathcrthanmorcmndamcntastruc-
turacauscs. ndccd,bytruncatingananaysisoIproccsscsunIdingovcrancx-
tcndcd pcriod oI timc thcy may cnd up invcrting causa rcationships-as
Larpcntcrsuggcsts in his crtiguc oIrincipa-^gcntthcory.crhaps mosthn-
damcntaoIa, thcy maymitocvcn dcnti| somcimportantgucstionsabout
poiticsbccauscdcrccvantoutcomcshappcntoosowyandarcthcrcIrcsim-
pyohthcirradarscrccns.
LhaptcrsuraivcintcgratcandcxtcndthcargumcntsoIthchrstthrcc
chaptcrsroughahcusonissucsoIinstitutionaoriginsandchangc._ucstions
about thc chccts oIinstitutions havcbccomc ccntra to thcorcticadiscussions
throughoutthcsociascicnccs.Norcrcccndy,sociascicntistshavcbccomcin-
tcrcstcd in cxpaining institutiona arrangcmcnL rathcr than simpy anayzing
thcirchccts. Nyanaysisisdcsigncdtoshowhowsystcmaticattcntion tothcar-
gumcnts oIthc hrst thrcc chaptcrs can iuminatc this maor topic in contcm-
porarysocia dcop.^t thc samc timc, in arguing hr a shih in hcus hom c
probcmoIinstitutionaselection to thc probcm oIinstitutionadevelopment,
mrthcrcxtcndmycritigucoIthcahistoricaprociviticsoImodcmsociascicncc.
A socia scicntists havc sought to cxpain institutiona outcomcs, thcrc has
bccnastrongtcndcncytocmpoy mnctionaintcrprctationsinwhichinstitu-
tionaarrangcmcnts arc cxpaincd by thcir conscgucnccs.nparticuar, what
tcrm actor-bascd mnctionaism Qicay rcsts on thc caim that institutions
takcthchrmthcydobccausc powcrh actorscngagcdin rationa, stratcgicbc-
haviorarcscckingtoproduccc outcomcs obscrvcd.unctionacxpanations
oI institutiona arrangcmcnL arc ohcn pausibc, but Lhaptcr our dcmon-
stratcshowthcadoptionoIancxtcndcdtimchamcrcvcasnumcrousprobcms
hr such accounts.unctiona intcrprctations oIpoitics arc ohcn suspcct bc-
causcoIthc sizabctimc ag bctwccn actors' actions and thc ong-tcrm consc-
gucnccsoIthosc actions.oitica actors, Icingthc prcssurcs oIthcimmcdiatc
orskcpIicaaboutthcircapacitocnginccrong-tcrmchccL, maypayimitcd
ahcntion to thc ong tcrm. Jhus c ong-tcrm chccts oIinstitutiona choiccs,
PLACING POLITICS IN TIME 1 5
which arc hcgucndy thc most proIund and intcrcsting oncs, shoud ohcn bc
sccnasthcby-prducts oIsociaproccsscsrathcrmancmbodyingthcgoasoIso-
cia actors. ^ sccond issuc rcatcd to tcmpora gaps bcwccn actions and out-
comcs conccrns unntcndcdconscgucnccs. Lvcn whcrcactors maybcgrcady
conccrncdaboutthcmturcinthcircHortstodcsigninstitutions,thcyopcratcin
scttings oIgrcatcompcxiand high unccrtain|. A aconscgucncc, thcy wi
ohcn makc mistakcs.orthcsc and othcr rcasons dcvcopcd in mydiscussion,
wcshoudgcncraycxcrciscconsidcrabcskcpticsmaboutasscrtionsthatinsti-
tutiona arrangcmcnts wircncctthcskicddcsignchoiccsoIrationa actors.'
nsIcad,wcshoudanticipatcthatthcrcwiohcnbcsizabcgapsbctwccnthcex
ante goasoIpowcrmpoiticaactorsand thcactuamnctioningoIpromincnt
institutions.
_uch_a_swoudnotmahcrmuchhrmnctionaaccountsiIinstitutionswcrc
casiyadaptcdinrcsponsctocurrcntnccdsorimprovcdundcrstandingsoIinsti-
mnashortcomings.Jhissuggcststhata ccntra issucmrinstitutionathcory
is thc roc oI adaptation, which may pay out mrough two important socia
mcchanisms- carningandcompctition- opcratingovcrtimc.^moughthcsc
arctwovcrysignihcantsociaproccsscs,argucthatncithcrisikcytoprovcad-
cguatctorcscucmnctionastaccountsinmanypoiticacontcxts.Jhisisinpart
bccauscthcscmcchanismswiohcnbcwcaknpracticc,butitisasobccausc
cscmcchanismsmustasoovcrcomcsubstantiabarricrstoinstitutionachangc.
JhissoncoIthcmostsignhcantimpicationsoIthccaricrdiscussionoIpath
dcpcndcncc,athoughmcrcarcadditionarcasonswhyinstitutionarcdcsignis
ohcndimcutinpoitcs.
ndccd,argucinLhaptcrivcatundcrstandinghownsttutionaarrangc-
mcntscanbccomcdccpycmbcddcdovcrtimcsuggcstscnccdtorchamcthc
topic as onc oIinstitutionadevelopment rathcr than institutiona choice. Vc
nccdto thinknotustaboutmomcnts oIinstitutonasccctionandmomcntsoI
institutiona changc, but oI proccsscs oIinstitutona dcvcopmcnt undng
ovcrsignihcantpcriodsoItimc.Jhcscproccsscsprohundyshapcthccircum-
stanccs undcr which modihcations to institutons arc ikcy tc occur, and thc
kindsoIchangcsthat arcikcy.^cusonmcdynamicsoIinstitutionadcvc-
opmcnt,arguc,cangcncratcguitcdistinctivcrcscarchagcndashrthoscintcr-
cstcdincxpaningpattcrnsoIinsttutionaoutcomcsinthcsocaword.
Jhcschvcchaptcrsprovidcthccorcomyanswcrtowhyhistorymattcrshr
socia sccntists. ndccd, wcarc nowin a position to rchamc thc gucstion ina
morchcpm way. Vhydosociascicntistsnccdtohcusonhowproccsscsun-
hd ovcr signihcant strctchcs oI timc irst, bccausc many socia proccsscs
arc path dcpcndcnt, in which casc kcy causcs arc tcmporay rcmovcd hom
7 Altough such claims can be, and ofen will be, supported by carefl inquiry. The point is that such
connections are ofen assumed or asserted witout justifcation, and tat exploring te limits of such
"rational design" accounts is teoretically fuitl.
1 6
INTRODUCTION
their continuing efect and a cental fcus of analysis is on "lost'' alternatives re
sulting fom te accumulation of selfreinforcing processes. Second, because
sequencing-the temporal order of event or processes -can be a crucial deter
minant of important social outcomes. Third, because many important social
causes and outcomes are slow-mo

ing-they take place over quite extended pe

riods of time and are only likely to be adequately explained (or in some cases
even observed in te frst place) if analyst are specifcally atending to that pos
sibilit. Finally, because the task of explaining institutional outcomes is better
famed as an issue of institutional developmentrather than one of institutional
selection. Institutional development, in tum, cannot be adequately treated With
out attending to issues incorporating an extended time fame, including the role
of time horizons, unintended consequences, learning and competitive selection
processes, and path dependence.
Chapter One
POSITNE FEEDBACK AD PATH DEPENDENCE
And te frst step, as you know, is always what matters most,
particularly when we are dealing wit tose who are young
and tender. 1bat is te time when they are taking shape
and when any impression we choose to make leaves
a permanent mark.
-Plato, Te Rpublic
IMAGINE A VERY LARGE URN CONTAINING TO BALLS, one black, one red. 1 You
remove one ball, and ten return it to the um along wit an additional ball of
the same color. You repeat this process until the um flls up. What can we say
about te eventual distribution of colored balls in te um? Or about a series of
trials in which we fll the um and ten start over again cne hundred times?
In each individual trial we have no idea what the eventual ratio of red to
black balls will be; it could be 99. 9 percent red, or 0, 01 percen1 red, or
anything in between. If we were to run one hundred trials, we would
probably get one hundred diferent outcomes.
In any particular trial, te ratio will eventually reach an equilibrium. Later
draws in a series contibute only minutely t te distibuton of balls in te
um. Thus te distibution settes down onto a stable outcome.
Sequence is thus crucial. Early draws in each trial, which have a consid
erable random element, have a powerfl efect on which of te possible
equilibria will actually emerge.
Mathematicians call this a Polya um process. m
fom 'the fct tat an element of chance (or accident) is combined with a deci"
sion rule tat links current probabilities to te outcomes of preceding (partly .
random) sequences. 2 Polya um processes exhibit positive fedback. Each step
1The fllowing discussion relies heavily on Aur 1994, which collect his groundbreaking essays on
increasing rr and pat dependence, as well a on te work of Paul David (1985, 1994, 2000), who
has been an equally important contibutr to te emerging literature on pat-dependent processes.
2his case depicts a specifc tpe of positive fedback process, in which te probabilit of a particu
lar "draw" precisely equals the rato between te two alteratives in te elisting population. Aur
( 1 994) has shown tat many of te fatures of tis case have a greater range of application, but not
all of tem. It is easy t model processes wit only two equilibria ( e.g:, Hill 1997), which probably
come closer to capturing te essence of many pat-dependent processes ih te social world.
1 8 CHAPTER ONE
aongaparticuarpathproduccsconscgucnccsthatincrcascthcrcativcattrac-
tivcncss oImat pa mr thc ncxt round. A such chccts bcgin to accumuatc,
thcy gcnctatcapowcrmcyccoIscI-rcinhrcingactivi.
ositivcIccdbackproccsscshavcguitcintriguingcharactcristics,whichhrian
^rthur[1 994) hassummarizcdashows.
1 . Unpredictability. hccausccarycvcntshavcargc chcctsand arc party
random, many outcomcs maybcpossibc. We cannot prcdict ahcadoI
timcwhichoIcscgossibccnd-statcswibcrcachcd.
2. Infexibility. Jhcmrcrintocproccss wcarc,thchardcritbccomcs
toshihhomoncpathtoanothcr. nappicationstotcchnoogy,agivcn
subsidytoaparticuartcchnigucwibcmorcikcytoshihthcutimatc
outcomciIitoccurscaryrathcrthanatcr. bumcicntmovcmcnt down
aparticuarpathmaycvcntuayockinoncsoution. '
3. Nonergodicity. ^ccidcntacvcntscaryinascgucnccdonotcanccout.
Jhcy cannot bc trcatcd [which is to say, ignorcd) as noisc,' bccausc
thcyhcdbackintomturcchoiccs. bmacvcntsarcrcmcmbcrcd.
4. Potential path inefciency. nthcongrun, thcoutcomcatbccomcs
cstabishcdmaygcncratcowcrpayohbanakrcgoncatcrnativcwoud
havc.Jhcproccssmaybcpathincmcicnt.
Jo thisonccan add a gcncra point oIparticuar intcrcst tosociascicntists.
thcsc arc proccsscs whcrc scgucncing is critica. aricr cvcnts mattcr much
morcthanatcroncs,andhcnccdihcrcntscgucnccsmayproduccdihcrcntout
comcs. Jhcscarcproccsscswhcrchistorymattcrs.
JhcoyaumiushationcapturcscsscntiaccmcntoIpamdcpcndcncc -
atcrmthatsociascicntisLarcincrcasingyincincdtoappytoimportantsocia
proccsscs. Laims oIpa dcpcndcncc havc hgurcd in bo cassic works oI
comparativcpoitics, suchasLipsctand okkan'sanaysisoIuropcanparsys-
tcms [Lipsct and okkan 1 967), and morc rcccnt anayscs on topics such as
aborincorporation inatin^mcrica [Loicrand Loicr 1 99 1 ), thcoutcomc
oI statc-buiding proccsscs in uropc [rtman 1 996) , and c Liurc oI thc
Lnitcd btatcs to dcvcop nationa hcath insurancc [ackcr 1 998). A in thc
oya um iustration, socia scicntists gcncray invokc thc notion oIpad dc-
pcndcncctosupportaIcwkcy caims. spccihcpahcrnsoItimingandscgucncc
mahcr,startinghomsimiarconditionsarangcoIsociaoutcomcsisohcnpos-
sibc, ar c conscgucnccs ma rcsut hom rcativc sma or contingcn_
vcnts,pa uarcourscsoIaction onccintroduccd,canbcvirtuayimpossibc
tqrcvcrsc, andconscgucndy,poiticadc t-
3This emerging stabilit represents a critical distinction between positive fedback processes and
chaotic processes tat may generate no stable equilibrium. For an interesting discussion of this quite
diferent famework, wit applications to politics, see Fearon 1996.
POSITIVE FEEDBACK, PATH DEPENDENCE 1 9
i@mor en ccbasiccontoursoIsociaiIc haumgartncr
andJoncs 1 993; LoicrandLoicr 1 991 ; kcnbcrry 1 994; Krasncr 8I
thcsc Icaturcsstand in sharp contrast topromincntmodcsoIargumcnt andcx-
panation in thc socia scicnccs, which attributc argc outcomcs to argc
causcsandcmphasizcthcprcvacnccoIuniguc,prcdctabcpoiticaoutcomcs,
thcirrccvanccoItimingandscgucncc,andthccapacioIrationaactorstodc-
signandimpcmcntoptimasoutions [givcnthcirrcsourccsandconstraints)to
thcprobcmsthatconhont thcm. Ipath-dcpcndcnccargumcnts arcindccdap-
propriatcinsubstantiaarcasoIpoiticaiIc,thishasconsidcrabcimpicatons.
Nyanaysisbcginswithagcncradiscussionoh dc_cndcnccthatsccksto
car@mcim_ortanIambjg[ticssurronnhcconccpt. th

ncand
acstigatc thc distinctivc charactcristics oIsocia proccsscs subcct to positivc
Iccdback. hcus hcrcon thcsc proccsscs bothbccausc thcy arc oIgrcat socia
signihcanccandbccauscsociascicntistsarcbcginningtodcvcoprigorousargu-
mcntsaboutthcircauscsandconscgucnccs. ositivcIccdbackdynamicscapIurc
twocy ccmcnts ccntra to most anaysL' intuitivcscnsc oIpath dccndcncc.
lirst, thcy ccary rcvca howthccostsoIswitchinghom onc atcrnativc toan-
othcrwi, inccrtainsociacontcxts,incrcascmarkcdyovcrIimc. bccond, and
rcatcd, thcy draw attcntion to issucs oI timing and scgucncc, distinguishin
hrmativc momcnts or conuncturcs hom thc pcriods that rcinhrcc divcrgcnI
paths _roccss in_ovi__ositvc_k, itis not |usta qucstion oIwhat
happcns, but oIwhcn it happcns. ssucs oItcmporaity arc at the hcart oIthc
anaysis.
hc howing scction rcvicws thc dcvcopmcntoIargumcnts aboutpositivc
cdback in thcsociascicnccdiscipinc whcrcthcy havcrcccivcd thcgrcatcst
hcnIion cconomics. JhisrcvicwsuggcststhcwidcswccpoIp0tcntiaapptca-
tions, cvcninahcdthatmightbc cxpcctcdtobchostictothc idca. Norcim-
portant, thcsc cconomicappicationsprovidc thc most anayticay dcvcopcd
discussions oIpositivcIccdback. conomists havcnot ony carihcd thcprinci-
pa impications oIpath dcpcndcncc but havc asoidcntihcd many oIthcspc-
cihcaspcctsoIaparticuarsocia cnvironmcnt thatgcncratcsuchproccsscs.
Jhc discussion oIcconomics prcparcs thc way hr an cxporation oIthc dis-
tinctivccharactcristicsoIpolitics. athcrthansimpyappyingcxtantargumcnts
in cconomics to poitica phcnomcna, wc nccd to considcrthc IcaturcsoIthc
poitica wordthatrcguircmodihcationsinthcuscoIpath-dcpcndcncccaims.
widcmonstratcthatargumcntsaboutpositivcIccdbackarcatcastasrccvant
to an undcrstanding oIpoiticsas thcy arc in o

th

as

th

iacs
ndccd,LctorssuchasthcpromincnccoIcocctivcactvtyinotis, tcccn-
tra roc oIhrma, changc-rcsistant institutions, thc possibiitics hr cmpoyin
poitica authority to magni| powcr asymmctrics, and Ihc grcat ambigui oI
many poitica proccsscsand outcomcs makcthisa domain oIsociaiIcthatis
cspcciaypronctopositivcIccdback.
20 CHAPTER ONE
JhchnascctonohcrsaprcmnaryasscssmcntoIwhatcscargumcntscan
contrbutctopotcaanayss.Jhcyprovdcanmportantcautona_an_ atoo
casy concuson oIc ncvtabg, natrancss, or mnctonaty oIobscrvcd
outcomcs. Lvcnthc ubguoIcams aboutc
g
ccntormnctVna ccmcnts
moms,saoncwoudbcan mportant corrcctvc. Norcsgnhcant, pa-
dcpcndcntargumcntsust| chortstostrctchctcmpora horzons opotca
anayss. Jhcy can rcdrcct thcgucstonssoca sccntsts ask, conbbutng toa
rchcr apprccaton oIthc ccntraty oIhstorca proccsscs n gcncratng vara-
ton n potca h. Jhcy can aso pont to promsing hypothcscs about thc
sourccsotbothpotcastabandpotcachangcnccrtaincommonpotca
schngs.ornstancc,cscag_h_htcnccdtocondcrh@os
bascd on tcmpora ordcrng-c possb dat c partcuar scgucncng oI
cvcnLorproccsscsmaybcakcypartoIthccxpanatonIrdvcrgcntoutcomcs.
Jhcy aso suggcst that n our scarch hr cxpanaton wc nccd to thnk about
causcs and chcctsat arc ohcn scparatcd n tmc, rathcrthan hcusng cxcu-
svcyonsynchronccxpanatons arsany 1 960; btnchcombc 1 968).
AIH DEPENDENCE AND OsiIiVFEEDBACK
^naysts arc ncrcasngy ncncd to nvokc thc conccpt oIpath dcpcndcncc,
buI ccar dchntons arc rarc. n practcc, usagc tcnds to huctuatc bctwccn a
broadcr and narrowcr conccpton. Vam bcwc, hr nstancc, suggcsts path
dcpcndcnccmcans that whathappcncdatan carcr pont ntmc w ahcct
thcpossbcoutcomcsoIascgucnccoIcvcntsoccurrngataatcrpontntmc
bcwc 1 996, pp. 262-63) . Jhsusagcmaycntai ony thc ooscandnotvcry
hcpmasscrtonat hstory mattcrs.A atcrnatvc,narrowcrconccpton oI
pathdcpcndcncchasbccnsuggcstcdbyNargarctlcv:
athdcpcndcncchastomcan, Itstomcan anythng, thatonccacoun-
t or rcgon has startcd down a track, thc costs oIrcvcrsa arcvcryhgh.
crcwbcocrchoccponts, butdccntrcnchmcntsoIccrtan nst-

|_
tona arrangcmcntsobstruct an casy rcvcrsa oIthc nta chocccr-
hapsthcbcttcrmctaphorsaIrcc,rathcranapath.romcsamc6ik,
thcrcarcmanydhcrcntbranchcsandsmacrbranchcs.^thoughtspos-
sbctotumaroundortocambcrhomonctothcothcr-andcsscntaI
thcchoscnbranchdcs-thcbranchonwhchacmbcrbcgnssconc
shctcndstohow. Lcv 1 997, p. 28)
nthsconccpton, whch wbcadoptcd hcrc, path dcpcndcnccrcIcrs tody-
namcproccsscsnvovng postvc Iccdback, whch gcncratcmutpcpossbc
outcomcsdcpcndngonthcpartcuarscgucnccnwhchcvcntsunhdrthur
1 994; Lavd2000).
POSITIVE FEEDBACK, PATH DEPENDENCE 2 1
^rthur, Lavd, andothcrs havc argucd that thc cruca lcaturc ola hstor-
ca proccss that gcncratcs path dcpcndcncc s positive fe'9k or scl-
rcnIrccmcnt) . Lvcn ths

~
cuar drcctonmakcs t
morcdmcuttorcvcrsccoursc.A lauLavd(2000, p.
.

contcntothcconccptoIpth dcpcndcnccasadynamc propcrty rchrs to thc

dca ol hstory as an rrcvcrsbc branchng proccss. bimiary, Jacob Hackcr
(2002, p. 54) argucsthat path dcpcndcnccrcIcrstodevelopmental trajectories
that are inherently difcult t reverse." nthc prcscncc olpostvc lccdback, thc
probabtyoIhrthcrstcpsaongthcsamcpancrcascs wthcacImovcdown
that path.Jhssbccauscthcrelative bcnchtsoIthccurrcntactv

comparcd
wth oncc-possbc optons ncrcascs ovcr tmc. Jo put t a dhcrcnt way, thc
costsoIswtchngtosomcprcvousypausbcatcrnatvcrsc.
^thoughsomcprckrdhcrcntdchntons, choosctocmpoythctcrmpath
dcpcndcncc nthsrcatvcy rcstrctcd scnsc, rchrrng to socaproccsscsthat
cxhbt posttvc Iccdbackandthusgcncratcbranchng pattcrns of hstorcadc-
vcopmcnt. Jhcmzzncssthathas markcdthcuscoIthsconccpt nsocasc-
cncc suggcsts thatthc grcatcrrangcohcrcdby ooscr dchntons hascomcata
hghprccnanaytca carty. @chasbccnavctmoIwhat bar-
torcacdconccptstrctchng bartor 1 970). ihcrcntgcsoItcmporaynkcd
scgucnccsarcgcncratcdndhcrcntways andhavcdhcrcntmpcatons ^-
bott 1 983, 1 990). JhcscdstinctvckndsoIsocaproccsscs, whchhavcbccn
bundcdtogcthcr,mustbcdisaggrcgatcdandsystcmatcaycxporcd. mtng
thc conccpt oIpa dcpcndcncc to s_rcnIrcing groccss n no way prc-
cudcscnvcstguhcrwaysnwhchscgucnccscanmattcrncxpan-
ngsoca!outcomcs. tdocscncouragcccarargumcntaboutdstnctcams. '
owcvcr such ssucsoIconccpt hrmaton arc utmatcy rcsovcd, thcrcarc
thrcccompcing rcasons fr hcusng spcca ahcntonon proccsscscxhbtng
postvchcdback. rst,suchproccsscs-nwhchoutcomcs nthccarystagcs
oIascgucnccIccdonthcmscvcs,andoncc-possbcoutcomcsbccomcncrcas-
ngy unrcachabcovcrtmc-charactcrzc many mponant par oIdc soca
4Like oters, I will sometimes describe these dynamics as generating irreversibilities, byt tis should
be read as shorthand fr "rising costs of reversal over time." It may be possible to imagine, or even
experience, a reversal in which some previously fregone alternative is recaptured. The point is that
the costs of doing so may increase sharply over time-so sharply as to make such a development
highly improbable.
5Anoter strategy would be to utilize a broader conception of path dependence and then disaggre
gate it, exploring how distinct tpes of pat dependence are generated in diferent ways, with difer
ent consequences (Mahoney 2000). Because research where pat dependence is invoked without
frther clarifcation is so prevalent, however, tis option seems more problematic. Ultimately such
defnitional disputes remain intractable. What is critical is that researchers should be clear and con
sistent about what they mean when tey employ the concept, and recognize te importance of dis
tnguishing diferent tpes of processes.
22 CHAPTER ONE
world. Second, social scientists are developing theory that makes the investiga
tion of te causes and consequences of positive fedback a particularly promis
ing area of inquiry. Third, a fcus on self-reinfrcing, path-dependent dynamics
turns out to be an essential building block fr exploring a wide range of issues re
lated to temporal processes. This fnal claim is advanced in the chapters to fllow,
while te frst to are developed in later sections of tis chapter-a task tat frst
requires a review of recent work on pat dependence in economics.
"INCREASING RETURNS" AD PATH-DEPENDENCE AGUMENTS IN ECONOMICS
Traditionally, economists have fcused on the search fr unique equilibria. The
goal is attactive, because it suggested a world of potential predictabilit and ef
fciency. Given knowledge of existing fctor endowments and prefrences, equi
librium analysis might point to a single optimal outcome. Moreover, because
economist assumed a context of decreasing marginal returns, this analytical
goal was potentially achievable. With decreasing returns, economic actions will
engender negative feedback, which will lead to a predictable equilibrium. ^sharp
rise in oil prices prompts increased conservation, exploration, and exploitation
of other sources of energy, leading to a fll in oil prices. Each step away fom
equilibrium is more difcult than the one befre. A Arthur ( 1 994, p. l ) sum
marizes, negative "fedback tends to stabilize the economy because any major
changes will be ofset by the very reactions they generate . . . . The equilibrium
marks the 'best' outcome possible under the circumstances: the most efcient
use and allocation of resources."
During the past twenty years, however, this decreasing-returns tradition has
fced a mounting challenge. Economists have exhibited a growing interest in
the idea of "increasing returns" -where each increment added to a particular
line of activit yields larger rather than smaller benefts. On a wide range of sub
j ects, including the spatial location of production, the development of inter
national trade, the causes of economic growth and the emergence of new
technologies, path-dependence argument have become prevalent. The ideas
developed in this research are not entirely new.6 Yet in the past fw years, promi
nent mainstream economists have embraced these ideas. Their work has re
ceived considerable attention in leading journals. Douglass North, who places
great emphasis on such arguments in his analysis of the development of modem
capitalism, was recenty awarded the Nobel Prize fr economics.
Arguments about technology have provided the most fertile ground fr ex
ploring the conditions conducive to increasing returns. A Brian Arthur and
6e concept of increasing returns received attention in the work of Adam Smith and (especially)
Alfed Marshall. In the twentieth century, an underground of "institutionalist" scholarship, includ
ing fgures such as Kaldor, Myrdall, and Veblen, contint1ed to explore these issues.
POSITIVE FEEDBACK, PATH DEPENDENCE
40
35
30

25 "
:
4 20

1 5 >

- q,ti^ -

"
a
1 0
5
0
2 3 4 5 6 7
Number of Users (in thousands)
Fig. 1 . 1 . Payofs wit increasing returs to scale.
8 9 1 0
23
1 1
Paul David have stressed, under conditions ofen present in complex, knowledge
intensive sectors, a particular technology may achieve a decisive advantage
(Artur 1 994; David 1 985) . A early edge may trigger positive feedback efects
tat may lock in tis technology, excluding competitors even if it is not neces
sarily the most efcient one in the long run. With increasing returns, actors have
strong incentives to fcus on a single alternative and to continue down a specifc
path once initial steps are taken in tat direction. Pat dependence arguments
have been applied to te development of the "QWR1' tyewriter keyboard,
te triumph of te light-water nuclear reactor in te United States, te bates be
tween Betamax and VS video recorders and DOS-based and Macintosh com
puter, early automobile designs, and competing standards fr electric current.7
Figure 1 . 1 , taken fom Arthur's work, summarizes te process. Each technol
ogy generates higher payofs fr ever user as it becomes more prevalent. In
other words, these technologies are subject to increasing returns. Because tech
nology B start wit lower payofs, however, early users gravitate to technology A.
This movement activates a process of positive fedback, improving te perfor
mance of technology A, which induces more new users to adopt it, which widens
the gap between technology A and B, encouraging yet more users to gravitate to
technology ^. The advantages of technology A rapidly become overwhelming,
even though technology B would have generated higher payofs fr all users if it
. had been the frst to reach a critical treshold of usage (here, 3 500 users). Th us
7Many of tese examples have been contested by critics who deny te empirical claim tat superior
technologies lost out. Since tese criticisms raise broader issues about the useflness of increasing
returns argument, I will postpone discussion until te end of tis section.
24 CHAPTER ONE
whcnancwtcchnoogyissubccttoincrcasing rcturns, bcingthc LstcstoutoI
thc gatc iI ony hr rcasons oIhistorica accidcnt) bccomcs critica. Vith in-
crcasing rcturns, actors havc strong inccntivcs to [cus ona singc atcrnativc,
andtocontinucmovingdownaspccihcpathonccinitiastcpsarctakcninthat
dircction.
lotatcchnoogicsarcpronctoincrcasingrctums.Lruciay,^rthurandavid
addrcsscd not onythccharactcristicsoIsuch proccsscs, but thcconditionsthat
givcrisctothcm. Lndcrstandingthcscconditionsiscsscntia,aswcshascc,bc-
causc anayticay simiar circumstanccs occur hcgucndy in thcword oIpoi-
tics. Jhcsc argumcnL thus providc a hundation Ir dcvcoping hypothcscs
aboutwhcnpositivchcdbackproccsscsarcikcytoopcratcinthcsociaword.
^rthur ( 1 994, p. 1 1 2) argucsthat kurIcaturcs oIatcchnoogy and itssoca
contcxtgcncratcincrcasingrcturns.
1 . Large set-up or fxed costs. Jhcsccrcatcahighpayohhrhrthcrinvcst-
mcnts in a givcn tcchnoogy. Vith argc production runs, hxcd costs
can bc sprcad ovcr morc output, which wi cad to owcr unit costs.
Vcnsctuporhxcdcostsarchigh,individuasandorganizationshavc
astronginccntivctoidcnti|andstickwitha singc option.
2. Leaming efects. KnowcdgcgaincdinthcopcrationoIcompcxsystcms
aso cads to highcr rcturns hom continuing usc. Vithrcpctition, ndi-
viduas carn how to usc products morc chcctivcy, and thcir cxpcri-
cnccsarcikcytospurmrthcrinnovationsinthc product ornrcatcd
activitics.
3 . Coordination efects. Jhcsc occur whcn c bcnchts an individua rc-
ccivcshomaparticuar activiincrcascasocrsadoptcsamcoption.
Itcchnoogics cmbody positivc netork exteralities, a givcn tcchno-
ogywibccomcmorcattractivcasmorcpcopcuscit. LoordinationcI-
hcL arccspcciaysignihcantwhcnatcchnoogyhas to bc compatibc
with ainkcdinhastructurc c.g. ,soharcwith hardwarc, automobics
withaninhastructurcoIroads,rcpairLciiticsand mcingstations). n-
crcascduscoIatcchnoogy cncouragcsinvcstmcnLinthcinkcdinha-
structurc, whichintm makcsmctcchnoogymorcattractivc.
4. Adaptive expectations. Ioptionsthat Litowin broad acccptanccwi
havcdrawbacksatcron,individuasmay Iccanccdtopickthcright
horsc.^thoughthc dynamichcrcisrcatcdtocoordinationchccts, it
dcrivcs hom thc scI-mhing charactcr oI expectations. rocctions
about hturc aggrcgatc usc pattcrns cad individuas to adapt thcir ac-
tionsinwaysthathcptomakcthosccxpcctationscomctruc.
JhisdiscussionoIIcchnoogyisimportantprimariybccauscitcarihcsasct
oIrcationshipscharactcristicoImanysociaintcractions. lcwsociainitiativcs-
suchasthccrcationoIorganizationsorinstitutions-usuaycntaconsidcrabc
POSITIVE FEEDBACK, PATH DEPENDENCE 25
start-upcosL,individuas,as wcas organizations,carnbydoing,thcbcnchtsoI
our individua activitics or thosc oIan organization arcohcn cnhanccd iIthcy
arc coordinatcd or ht with thc activitics oIothcr actors or organizations, it is
hcgucndy important to bct on c winning horsc, so wc adapt our actions in
ight oIour cxpcctationsabout thcactionsoIothcrs.
^though path-dcpcndcncc argumcnts about tcchnoog arc probaby thc
bcst known, cconomists havc appicd simiar anayscs in a striking rangc oI
cconomiccontcxts.hothKrugman( 1 991 ) and^rthur1 994) pointtothcrocoI
incrcasing rcturns inthc spatia ocation oI_roduction. Livcnthc importancc
oIphysica proximi|in many aspccts oIcconomic iI, aggomcration chccts
arcwdcsprcad. Jhatis,initiaccntcrsoIcconomicactivitymayact ikc amag-
nct andinnucncc thc ocationa dccisions and invcstmcnL oIomcr cconomic
actors. stabishcd hrms ah

act suppicrs, skicd abor, spcciaizcd hnanca

and cga scrviccs, and an appropriatc physica inhastructurc. Jhc conccn-
tration oIthcsc Lctors may in tum makc mc particuar ocation ahractivc to
othcr hrms that producc simiar goods. bodo socia nctwors, which ciitatc
thc cxchangc oInhrmation andcxpcrtsc. ncrcasing rcturns argumcnts hcp
cxpainthcprcvacnccoIpockctsotspcciaizcdcconomicactivi,hombiicon
Vacy to thchigh-cndtcxtcmanuLchrcrsoInorthcrntay. Krugman ,1 991 ,
p. 80) concudcs. Imcrc is onc singc arca occonomics in which path dc-
pcndcncc is unmistakabc, it is in economic geography-te ocation oI pro-
duction in spacc. Jhc ongshadowcast by history ovcrocation is apparcnt at
a scacs, hom thc smacst to thc argcst-hom thc custcroIcostumc cwcry
hrms in rovidcncc to thcconccntrationoI60 miion pcopcin thc orthcast
Lorridor.
Jhcsccaims coscy parac rcccnt anayscsoIintcrnationa tradc, whcrcar-
gumcnts about incrcasing rcturns havc gaincd widc acccptancc. Kcscarchcrs
bcan by hcusingoncconomictrcndsat appcarcd anomaous homcpcr-
spcctivcoItraditionabadcthcory-mostnotaby,ccxposionoIntrc-industp
tradcahcrVordVarKrugman 1 996). Icomparativcadvantagcrcsutshom
natura haturcsoIdihcrcntcountrics, onc woud cxpcct most tradc tooccur
bctwccn guitc dihcrcnt countrics, such as lorth-bou tradc cImanuLcturcd
goodshrrawmatcrias.Nostintcrnationatradc,howcvcr,islorth-lorth.c-
vcopcd cconomics tradc primariy wi othcr dcvcopcd countrics, incuding
cxtcnsvccxchangcswithinparticuar industrics. Jhispattcrnsuggcstapuzzc.
Vhyhavcbroadysimiarcountricsdcvcopcdhighyspcciaizcdnichccom-
parativcadvantagcs!
ncrcasingrcturnsprovidcdan ancr. nowcd c-intcnsivc scctors wi bc
p

oncto_ositivchcdback. Lountricsthatgainacadinapa icuarnic c,r

whatcvcrrcason,mayconsoidatcatcadovcrtimc.Jhcrcsutisahighdcgrcc
o spcciaization. vcncountricswsimiarinitiacndowmcrtsdcvcopdivcr-
gcntarcasoIcconomicshcngU. LomparaLvc advantagcisnotsimpygivcn, itis
26 CHAPTER ONE
ofen created trough a sequence of event unflding over time.8 The relevance of
increasing retrs processes to te economics of tade is now widely accepted. 9
Economist have also applied increasing-returs argument to economic change
more broadly. ^prominent development in discussions of economic growth has
centered on "endogenous growt" teory (Romer 1 986, 1 990). Economist in te
1 980s became puzzled by growth rates (notably in developed countries during
the post-World War II period) fr greater tan what measured increases in in
puts of capital and labor could explain. Romer and others argue that increasing
returns associated with economic applications of knowledge help account fr
the anomaly. Unlike capital and labor, many aspect ofknowledge are nonrival
their use in one frm does not prevent their use in another. ^ single gain in
knowledge can be applied in many settings and can lead to drmatic improve
ment in productivit. In short, economic growth generates positive feedback. ^
somewhat diferent analysis of growth based on increasing returns emphasizes
the importance of complementarities (Milgrom and Robert 1 990; Milgrom,
Qian, and Roberts 1 991 ) . Various economic activities (e. g. , in infrmation tech
nology) may be complementary to oter related activities. Improvements in a
core activity can tus spill over by improving related parts of the economy.
These improvements in turn may increase the attractiveness of the core activit.
Economists are now applying increasing-returns argument to a wide range of
important economic phenomena, but Douglass North's application to issues of
institutional emergence and change has te most profund implications fr so
cial scientists (North l 990a). Nort argues tat all te features that Arthur iden
tifed in investigations of increasing returns in technolog can be applied to
institutions. In contexts of complex social interdependence, new institutions
ofen entail high start-up costs, and they produce considerable learning efect,
coordination efects, and adaptive exectations. Established institutions will t
ically generate powerfl inducement tat reinfrce their own stabilit and fr
ther development (David 1 994) . In North's words:
Al fur of Artur's self-reinfrcing mechanisms apply, altough wit some
what diferent characteristics. There are large initial setup costs when the
institutions are created de novo . . . . There are signifcant learning efects
fr organizations that arise in consequence of the opportunit set provided
8It is wor noting tat this research on trade has been used to derive some contoversial policy im
plications. If frst-mover advantages are signifcant, fee trade may not be an optmal policy fr a
county whose tade parters are willing to subsidize emerging sectors. Under certain (resticted) con
ditons, a policy of"pickng winners" may make economic sense (Krugman 1996; Tyson 1 993). Tere
remains considerable dispute about te signifcance of such opporunites fr stategic interventon.
Kugman, fr instance, maintains tat tey wlappear relatively infequenty, not so much because
pat dependence is rare, but because governments will not be able to identif winners ex ante.
9As Krugman ( 1 996, pp. 1 09-1 10) notes, in the American Economic Association's classifcation sys
tem fr journal aricles one will now fnd "models of trade wit increasing returs and imperfect
competition" alongside te category fr "conventional trade models.''
POSITIVE FEEDBACK, PATH DEPENDENCE
27
by te institutional famework . . . . There will be coordination efects di
rectly via contracts wit oter organizations and indirecty by induced in
vestments trough the polit in complementary activities . . . . Adaptive l
expectations occur because increased prevalence of contracting based on a
specifc institution will reduce uncertainties about the permanence of that
rule. (North l 990a, p. 95)
North emphasizes tat not j ust single institutions are subject to positive feed
back. Institutional arrangements induce complementary organizational frms,
which in tum may encourage the development of new complementary institu
tions. For social scientist interested in paths of development, the key issue is
ofen what North calls "te interde en an institutional matrix." This
matrix, North emphasizes, "produce_massive increasing returns" (North l 990a,
p. 95) . Pat-dependent proc
sses will ofen be evident not only at the level of
individual organizations or institutions but at a more macro level that involves
confgurations of complementary organizations and institutions (Katznelson
1 997; Hall and Soskice 200lb; Pierson and Skocpol 2002).
This argument provides the core to North's sweeping reinterpretation of eco
nomic history. The central puzzle motivating North's inquiry is the limited
convergence of economic performance across countries over time. Neoclassical
teory suggest tat laggard counties should easily adopt te practices of high per
frmers, which would induce firly rapid convergence. But tis does not happen.
According to Nort, institutions, which he defnes broadly to include "te rules
of the game in a societ or, more frmally, . . . te humanly devised constraints
tat shape human interaction" (p. 3), explain te anomaly of continued diver
gence in economic performance. Once in place, instittions are hard to change,
and tey have a temendous efect on the possibilities fr generating sustained
economic grow. Individuals and organizations adapt to existing institutions. If
te institutional matix creates incentives fr piracy, Nor obseres, ten people
will invest in becoming good pirates. When instittions fil to provide incentives
to be economically productive, tere is unlikely to be much economic growt.
F other social scientists, Nor's insight is crucial fr two reasons. First, he
highlights the parallels between characteristics of technology and certain char
acteristics of social interactions. In tis context, it is wort noting that Artur's ar
guments about technology are not really about te technology itself but about
the characteristics of a technology in interction with certain qualities of related
social activit, such as incentives to coordinate wit others or adopt behaviors
based on expectations about the fture. Second, Nort rightly emphasizes that
instittional development is subj ect to positive feedback. Indeed, it is the role of
path dependence in explaining paterns of instittional emergence, persistence,
and change tat may be of greatest signifcance fr te social sciences.
The dialogue surrounding increasing returns and path dependence in
economics is te impassioned discourse of an emerging paradigm. Economists
28 CHPTER ONE
takoI ncwgrowcory, ncwtradcthcory, andsoon-abascdonargu-
mcntsinvovingpositivcIccdback.1ctdcspitcthcprcvacnccoIsuchargumcnts
and thcintccctuacxcitcmcntassociatcd with thcm, thcrc arc cxcccnt rca-
sontobcicvcatthcrangcoIappicaIionshoudbcatcastaswidcinpoitics
asitisincconomics. Jo undcrstandwhy,itishcphtoconsidcrthcma|orob-
|cctions toincrcasing-rcturnsargumcntsthathavcrcccndysurIccdincconom-
ics.Jhcdiscussionwicari|thcsourccsoIpadcpcndcnOcandidcnti|socia
mcchanismsthatmight ohct suchproccsscs. Jhis carihcation providcsausch
anayticabridgctoaninvcstigationopath-dcpcndcntproccsscsinpoitics.
naIrcchcritiguc, LicbowitzandNargois ( 1 995) raiscsomctoughgucs-
tionsaboutthc cconomicsitcraturconincrcasing rcturns. JwoaspcctsoIthcir
critiguc arc rccvant hcrc. Jhcy cmphasizc that ony rcmcdiabc paIh dc-
pcndcnccisrcayoIthcorcticasignihcancc,andthcycaimthatmarkctmcch-
anisms inrc mat rcmcdiabc pam dcpcndcncc is rarc. wi takc up cach
argumcntinturn.JO
oowingViiamson,1 993 ),LicbowitzandNargoisdtstinguishrcmcdiabc
and nonrcmcdiabc path dcpcndcncc. Jhc attcr occurs iIthcrcarc nofeasible
improvcmcnL in thc path, cithcr nowor n thcpast. lonrcmcdiabcpath dc-
pcndcncc stipuatcs at intcrtcmpora chcctspropagatc crror [p. 207) . Vi
hindsight,wc wishthat somcothcratcrnativchadbccnchoscn.1ct Licbowitz
andNargosgucstionwhcthcrnonrcmcdiabcpathdcpcndcncc hasprohund
impications. Iwcactcdasbcst wc coudwiththc inhrmation avaiabcat thc
timc, thcnthcmistakc wasunavoidabc,andwccannotrcasonabydcscribcthc
outcomcas incmcicnt.

Licbowitz and Nargois arguc that thc ony kind opath dcpcndcncc with
maorramihcatonsispamdcpcndcnccthatispotcntiayrcmcdiabc: pathdc-
pcndcncc, . . . whichj supposcsthchasibii,inprincipc,oIimprovcmcntsin
lhc path . . . is thc ony hrm oIpath dcpcndcncc that connicts with thc nco-
cassicamodcoIrccndcssyrationabchaviorcadingIocmcicnt,andcrchrc
prcdictabc,outcomcs [ibid) . Jhisdistinctionbctwccnrcmcdiabcand nonrc-
mcdiabcpathdcpcndcnccscruciatothcirargumcnt,bccauscLicbowitand
NargoisbcicvcIhat instanccsoIthcmorcthcorcticaytroubing, rcmcdiabc
kindoccurvcry inhcgucndy.
s thcir dismissa oI nonrcmcdiabc path dcpcndcncc convincing A
Viiamsonnotcs,hrpoicypurposcsrcmcdiabiiisikcyt bcanappropriatc
standard. ccognizingccxistcnccoIpa dcpcndcncc may not hcp poicy-
makcrsmuchiIthcydonotknowhowtoidcnti|itex ante. 1 1 hutthisobcction
1Note that te Liebowitz-Margolis critique depends on both parts of teir argument being true. The
signifcance of path dependence fr social scientists can be sustained if either te relevance of non
remediable path dependence or te prevalence of remediable path dependence can be sustained.
1 1 A noted befre, it is precisely fr this reason that Krugman and others question those making
broad claims about te implications of increasing returns arguments fr tade policy.
POSITIVE FEEDBACK, PATH DEPENDENCE 29
oscsitshrcciIourpurposcisinstcadtoundcrstand-pcrhapsex post-why as-
pccts oIsocicticsmovcin particuardircctions and c conscgucnccs oIsuch
movcmcnts. ^nd, oIcoursc, it is prcciscy thcsc gucstions about causaily that
arcthcccntraprcoccupationoImostsociascicntists.
JhcsccondpartoIcNargois-LicbowtzanaIysisisccaimatrcmcdiabc
path dcpcndcncc is rarc. Jhcir argumcnt is straighorward. onc oIto op-
tions issupcriorin thcong run butnotinthcshortrun, thcn markct arrangc-
mcnts wi gcncray assurc thc adoption oIthc supcrior path. Jhc abii| of
privatc actors to capturc mc rcturns hom ong-tcrm invcstmcnts prcvcnt bad
choiccs.nstitutionsoIpropcrtyrights,provisionshrpatcnts,andcxtcnsivccap-
itamarkctsinsurcthatoptionswithong-tcrmpromiscbutowshort-runpayoH
winoncthccssrcccivcthcsupportatthcydcscrvc.conomicactors,inshor|,
cacuatc in thc shadow oIthc mturc, and arc thus unikcy to indugc in my-
opic, short-tcrmmaxmizingbchavioratdcirownong-tcrmcxpcnsc.
Jhis argumcnt has considcrabc mcrit, ' but how much dcpcnds on thc
strcngth oIthcsc mcchanisms hr ovcrcoming short-tcrm thinking. ^thouh
icbowitzandNargoisarcmorcthanaihccompaccntaboutthccapacioI
dcscmarkctmcchanismstohyintcrnaizcthcconsidcrabccxtcrnaiticsthat
arc ccntra toincrcastng-rcturnsargumcnts, it ispcrhaps wisctocavc thosc is-
sucstocconomists. Jwoob|cctions, howcvcr, arc criticay important. Iirst, ar-
gumcntsaboutthcIrsightcdncssoImarkctssccmtoappytoonysome |pcsoI
pathdcpcndcnccinccconomy. JhcLicbowitz-NargoiscritigucIcuscson
te dccisions oIhrms toinvcst in particuar tcchnoogics or products. n most
oIthciustrationsdiscusscdcaricr[c. g. ,spatiaaggomcrations,tradcspcciaiza-
tion, cndocnous growm), howcvcr, many oIthc bcnchts oIincrcasing rcturns
arc cxtcrna to individua hrms and cannot bc my capturcd by individua in-
vcstors and cntrcprcncurs. Jhus, thc mcchanisms idcntihcd by Licbowtzand
Nargosarcunikcytocnsurcthatcbcstong-tcrmoutcomcwibcsccctcd.
Icrhaps morc important, thc Licbowitzand Nargois argumcnt has ihc
rccvanccto cdcvcopmcntoIinstitutions, whch arcasosub|ccttoincrcas-
tng rcturns. rivatc actors cannot obtain patcnts or cmpoy vcnturc capita to
capturc thcong-tcrmcconomicgainshom constructngkcycconomicinstitu-
tion. ndccd,cLicbowitz-Nargois argumcntsimpyassumcsdcprcscnccoI
institutionsatsupport cir kcy markct mcchanisms.Jhcirargumcntdocnot
sccm to havcmuchrccvancc hrlorth'sargumcntaboutcprcscnccinpar-
ticuarpoiticsoInctworksormatriccsoinstitutionsandorganizations.JhcIct
that thcy do not cvcn citc or's work is tcing. lorth maintains that path-
dcpcndcnt proccsscsoIinstitutiona dcvcopmcntarccruciatothccvoutionof
particuar markct cconomics. Te trsightcd hnancia markcts ccptra to thc
Licbowitz-Nargoisargumcntarc otimtcd hcp, howcvcr, in triggcring such
1 2Au explicitly recognized tis possibilit, altough as fr as I kno\ he did not systematically pur
sue the implications. See Artur 1994, p. 28, f 1 1 .
30 CHAPTER ONE
institutional development. Rather, to a large extent, they are product of that
development.
Te filure of te Liebowitz-Margolis critique to address issues of institutional
development in economies point tQ a more fndamental obj ection. Even if one
accept their analysis regarding the economic sphere, their argument still have
limited relevance fr te analyis of other kinds of social processes. However
strong market mechanisms fr "frsightedness" may be, they are almost cer
tainly fr weaker in politics. At the sa11e time, other sources of positive feedback
that may be relatively unimporant in te development of technology are highly
signifcant elsewhere. I explore both tese point in the next section.
MOVING FROM ECONOMICS TO POLITICS
Microeconomic theory has illuminated important features of the political land
scape in felds ranging fom te study of part competition, to the frmation of
interest groups and social movements, to voting and legislative. behavior. The
value of economists' theoretical exort is greaty enhanced, however, if te im
porters take carefl account of te distinctive fatures of te "local" environment.
A Terry Moe ( 1 990, p. 1 1 9) has put it in a related context, "te real problem is
to try to identif those essential features of politics that might serve as a funda
tion fr theory, a fundation that can take advantage of the new economics with
out being overwhelmed or misdirected by it." Argument drawn fom economics
must be sensitive to the quite diferent nature of the political world.
Politics difers fom economics in many ways. 13 The key is to specif which
pect are most rele
.
vant to a
.
n investigation of the sources and consequences of
.
th dependence. Following a brief summary of the distinctive tasks of the po
ltical arena, tis discussion is divided into two parts. The frst considers fur
prominent and interconnected aspects of politics that make this realm of social
life conducive to positive fedback: ( 1 ) the cental
.
role of collective action;
,)te high densit of institutions; (3) te possibilities fr using political autor
ity to enhance asymmetries of power; and ( 4) it intrinsic complexity and opacit.
Aer briefy explicating each, I will discuss their relevance to path dependence.
Each of these features makes positive feedback prcesses prevalent i n politics.
Second, I explain why te ameliorative mechanisms tat Liebowit and Mar
golis identif in economic sytems are ofen inefective in ofetting pat depen
dence in politics. Three characteristics of politics change te picture considerably:
the absence or weakness of efciency-enhancing mechanisms of competition
and learning, the shorter time horizons of political actors, and the strong status
quo bias generally built into political institutions. Each of these features makes
1 3The. fllowing discussion is particularly indebted t Lindblom 1 977, Moe 1984, 1 990, and Nort
1990b.
POSITIV FEEDBACK, PATH DEPENDENCE 3 1
positive feedback prcesses in politics paricularly intense. They increase the dif
fculty of reversing the course down which actors have started. Pat-dependent
processes are now central to economic theory, and te argument here is that
these dynamics will be at least as widespread and ofen more difcult to reverse
in politics.
Sources of Positive Feedback in Politics
A fndamental fature of politics is its fcus on te provision of public goods. 1 4
Such goods are distinguished by jointness of supply (where te production costs
fr te good are unafected or only modesty afected by te number of those
consuming it) and nonexcludabilit (where it is very costy or impossible to limit
consumption to those who have paid fr a good). These fatures, which are ex
tremely widespread in modem lif, make public goods -fom national defense
to environmental protection-difcult to provide trough markets. Nonexclud"
abilit creates incentives fr fee-riding, since individuals will receive te bene
fts of a public good whether or not tey contribute to it production. Jointness
of supply means tat private markets will underproduce the goods. in question
since private actors will tend to consider only te beneft to themselves.
The reason to emphasize tese characteristics of public goods is tat tey help
to explain a second fndamental feature of political systems: their key elements
are generally compulor rather than voluntary. The exercise of autority-to be
blunt, coercion-combined wit a complex array, of complementar institu
tions designed to circumscribe ard legitimate tat autorit, is necessary to gen
erate collective provision. Legally binding rules are not j ust a fundation fr .
political activit (like propert rights in the economy) . They are instead the ver
essence of politics (Lindblom l 977;Moe 1 990, 2003). The fcus on producing
public goods, and the consequent resort to coercive authority, has a number of
repercussions fr the character of political life, each relevant fr an assessment
of tendencies toward pat dependence.

THE COLLECE NATR OF POLITICS

A quick contrast wit economic markets can highlight the prevalence of collec
tive actiqn in politics. Suppose you are working fr a frm wit an annoying boss
and bad pay. You have a clear option: acting on your own, you can seekwork
elsewhere, either at one of a large number of oter frms or by setting up busi
ness on your own. Your abilit to move depends on the state of the labor market,
but te existence of competitive options sets clear limit on how annoying your
boss can aford to be and how bad the pay can get.
141n most cases, te goods in question are not "pure" public goods-a fct tat would complicate te
analysis but not alter the basic claims presented here. For discussions, see Mueller 1989, chap. 2,
and Cornes and Sandler 1996.
-
32 CHAYfER ONE
Or suppose you invent a great new product. Assuming that you can get fnan
cial backing (which you should be able to do-it is a great idea, and the market
generates a ready supply of venture capitalists), your prospects are good. Noth
ing stops you fom going into business or selling the idea to someone who will.
Either way, the new, superior product gets to see the light of day, and you reap
considerable benefts fom your innovation.
The setting of consumers, at least in the textbook case, is similarly atomistic.
My decisions are essentially independent of my expectations regarding the
choices of other consumers . 1 5 There is no need fr explicit attempts to coordi
nate behavior; the market simply aggregates the isolated decisions of individuals.
These highly stlized examples illustrate the relative fexibility, fuidity, and
atomization of economic markets. In contrast, political "markets" are generally
fr fom fexible ad fuid. In politics, the consequences of my actions are highly
dependent upon the actions of others. What I get depends not j ust on what I do,
but (mostly) on what others do. Following Olson's path-breaking work, students
of politics have long recognized the "logic of collective action" (Olson 1965) .
Most of the "goods" produced in politics are public goods; it i s difcult to limit
their consumption to those who helped provide them. A a result, individuals
will have a strong tendency to fee-ride. Creating conditions fvorable to collec
tive action is a principal issue in political life.
The problem here is not limited to the fct that the public sector produces
public goods. Given the reliance of politics on mechanisms of collective deci
sion backed by authority laws themselves have the character of public goods fr
those who beneft fom them. Achieving political infuence generally requires
collective action. In the words of Marwell and Oliver:
In the realm of politics and social movements, collective action gains ben
efts most ofen by afecting government policy. Collective actors bear the
cost of infuencing government ofcials, not the cost of actually providing
the good. Infuencing government policy almost always has very high j oint
ness of supply . . . those paying lobbyists fr tax loopholes are concerned
only with the cost of lobbying, not with the cost of the lost tax revenues.
The cost of cleaning up pollution may be roughly proportional to te num
ber of polluters, but the cost of obtaining laws requiring polluters to clean up
their own messes is not. A interest group or social movement campaigns
fr legislation of beneft to it members, but their costs are unafected by te
15 Altough tis represents a critical diference between economics and politics, one would need to
make a number of important qualifcations. The decisions of oter consumers clearly do afect the
price, supply, and qualit of the goods available to me. Furhermore, much economic activit, both
on the production and consumption side, involves signifcant externalities, which make te impli
cations of consumption interdependent. A already noted, tese conditions of independent con
sumption ofen do not apply to technologies involving network externalities. Fr a good discussion
of some of tese complications, see Hirsch 1977.
POSITIVE FEEDBACK, PATH DEPENDENCE
existence of oters who would also beneft fom the legislation. (Narwell
and Oliver 1993, pp. 42-43)
33
Or, as they summarize te same point more succinctly, "laws have high joint
ness of supply" (p. 45) . These circumstances generate major collective action
problems.
There is anoter reason why political action fequenty requires coordination.
Many of te goals tat political actors pursue have a "lumpy" or "winner-take-all"
qualit to them (politicians seeking reelection, coup plotters, and lobbyists ei
ter win or lose; legislaton eiter passes or is rejected). Unlike economic market,
where there is usually room fr many frms, fnishing second may not count fr
much in politics. Indeed-te Russian Menshiviks in 1 91 7 come to mind-it
can be extremely problematic. Aain, the efectiveness of my actions depends
heavily on the actions of others. This is less true of some aspects of politics
such as answering an opinion poll question or voting-than others. Even in vot
ing, however, te lumpiness of election outcomes, in the absence of a pure
system of proportional representation, means that if a person does not want to
"waste" her vote, her actions may well turn on what she expects others to do.
Thus, a crucial feature of most collective action in politics is the absence of a
linear relationship between efort and efect. Instead, . collective action fe
quently involves many of the qualities conducive to positive fedback (Marwell
and Oliver 1 993) . 1 6 A central reason is the prevalence of adaptive expectations.
Under circumstances where picking the wrong horse may have very high costs,
actors must constanty adj ust their behavior in te light of expectations of how
they expect others to act. Whether you put energy into developing a new part,
j oin a potential coalition, or provide resources to an interest group may depend
to a considerable degree on your confdence that a large number of other people
will do the same. In addition, many tpes of collective action involve high start
up costs. Considerable resources (material or cultral) need to be expended on
organizing befre the group becomes self-sustaining.
That collective action processes in politics are very ofen subjectto positive
fedback explains why social scientists have ofen been struck by the consider
able stability of patterns of political mobilization over tme. Lipset and Rokkan's
work on political parties in Europe exemplifes this dynamic: key historical
junctures produced major political cleavages. These political divisions became
organized into political parties. Having surmounted initial start-up costs and f
eled processes of adaptive expectations, these parties are reproduced through
time, generating "fozen" part systems (Lipset and Rokkan 1967) .
160lson's own Rise and Decline of Nations ( 1 981 ) i s built around exacty tis argument-interest
groups are difcult to fund but relatively easy to sustain, so we should expect them to be increas
ingly prevalent over time in democratic societes where severe exteral shock, such as war, have
been absent.
34 CHAPTER ONE
Recent work by Theda Skocpol on extensive voluntary associations in te
United States provides additional strong evidence of the organizational persis
tence tat can result fom positive fedback (Skocpol 1 999). Skocpol and a team
of researchers identifed all voluntary organizations in the United States that
have ever enrolled more than 1 percent of the Aerican population (or half that
amount fr single-gender groups), and tracked those organizations over time.
The result, covering f-eight such organizations since te 18 30s, reveal stiking
organizational continuities. Although some organizations crossed the 1 percent
threshold fr relatively
.
short periods, twent-six of te ffy-eight organizations
tat ever crossed the 1 percent threshold are still above it today. Of te twenty
six extensive voluntary associations in the United States today, sixteen had al
ready reached te 1 percent mark by te 1 940s, and a number of them stretch
back much frtler. A large number of te organizations that have fllen of.
Skocpol's list nonetheless stayed above it fr many decades. If one examines the
frty organizations on Skocpol's list funded befre 1 900, nineteen stayed above
the 1 percent mark fr at least fve decades; ten of the fr are still above that
treshold, a century or more afer teir funding. 17 In short, despite massive social,
economic, and political changes over time, self-reinforcing dynamics associated
with collective action processes -especially high start-up costs, coordination ef
fcts, and adaptive expectations -mean that organizations will have a strong
tendency to persist once they are institutionalized.
THE INSONAL DENSITY OF POLITICS
A much work in political science has recently stressed, eforts to coordinate ac
tors in the pursuit of public goods ofen require the construction of frmal insti
tutions. Once established, tese institutional constraints apply to all -those who
do not approve as well as tose who doand tey are backed up, ultimately, by
frce. The "exit'' option, while central to te workings of the market, is ofen un
available (or prohibitively costly) to actors who fel poorly served by existing po
litical arrangements.olitics involves struggles over the authorit to establish,
enfrce, and changJrules governing social action in a particular territory. In
short, much of politics is based on autorit rater tan exchange. Thus in pol
itics, institutional constraints are ubiquitous. Both frmal institutions (such as
constittional arrangements) and public policies place extensive, legally binding
constraints on behavior.\
Athough unortodox, te inclusion of public policies as well as frmal insti
tutions in this frmulation is important (Pierson 1 993; Pierson and Skocpol
2002; Moe 2003) . While policies are generally more easily altered tan the con-
17This evidence actually understates te degree of organizational persistence, since many organiza
tions that fll below te demanding 1 percent treshold noneteless contnue to have very large
memberships, and tey may also have existed as quite large organizations fr long periods befre ini
tially crossing tat treshold.
POSITIVE FEEDBACK, PATH DEPENDENCE 35
stitutive rules of frmal institutions, they are nevertheless extremely prominent
constraining fatures of the political environment. Policies, grounded in law and
backed by te coercive power of te state, signal to actors what has to be done,
what cannot be done, and establish many of the rewards and penalties associ
ated wit particular activities. Most policies are also remarkably durable (Rose
1 991 ) . Especially in modem societies, extensive policy arrangements fnda
mentally shape the incentives and resources of political actors.
That such institutions are gen
p
lly

subj ect to positive fedback has been im"
plicit in much recent research. / Scholars have emphasized how institutions
can help actors overcome various dilemmas arising fom collective-choicer
situa
.
tions -e
.
specially the need to coo
|

nate teir behavior by disciplining ex
.

pectations about the behavior of oter at is absent or downplayed in these

discussions, however, is a recognition at tese characteristics rendr processes
of institutional development path-dependent."
A already discussed, North's ?nalysis highlight how institutions induce self-
. ein
.
frcing
.
processes that makeev
.
ersals increasingly unattr
.
ac
.
ti.ve_
.

ov
.
ertime . In
.

ontexts o

complex social inte

dependence, ne in

titutions and policie

ofen

enerate h1

h ed cost, leammg efect, coordination efects, and adaptive ex-

ectanstitutions and policies may encourage individuals and organiza
tions to invest in specialized skills, deepen relationships wit oter individuals
and organizations, and develop particular political and social identities. 1 8 These
activities increase te atractiveness of existing instittional arrangements rela
tive to hypothetical alternatives. In institutionally dense environments, initial
actions push individual behavior onto pats tat are hard to reverse. A I will ex
plore in greater detail in Chapter Five, social actors make commitments based
on existing institutions and policies. A tey do so, the cost of reversing course
generally rises dramatically.
Economists typically refr to self-reinfrcing processes as involving a process
of "increasing returns": each additional expenditure of resources generates a
higher return tan the one befre it. A Mahoney (2001 ) has emphasized, this
efciency-fcused (beneft per unit of expenditure) terminology makes sense
fr te self-reinforcing processes tat interest most economist, but is not appro
priate fr many others. A we broaden our fcus to examine other social phe
nomena, the more neutal language of positive fedback or self-reinfrcement
1 8It has become common to refer t such consequences as "sunk cost." Wile intuitve, tis termi
nology is unfortunate. Wen economist refr to sunk costs tey mean costs tat cannot be recov
ered and should be regarded as irrelevant t current choices among options. By contast, te whole
point of pat dependence is tat tese previous choices ar releva
n
t to current action. In cases of in
creasing retur
n
s, social adaptations represent investment tat yield continuing benefts. Actors may
have powerl incentives to stick with a current opton because tey receive a continuing stream of
benefts fm investent already made in tat option. Massive new investents would be required
befre some teoretically superior alternative generated te same or a higher stream of beneft.
36 CHAPTER ONE
is necessary to avoid any implicit suggestion about efciency. This is especially
important in considering two possible sources of path dependence in politics:
power dynamics and patterns of social understanding.
POLITICAL AUTHORI1 AD POWER ASYMMETRIES
aThe allocation of political authority to particular actors is a key source of posi
ve fedback. Indeed, this represents a source of path dependen
.
ce quite distinct
fom those discussed by Arthur and North. Where certain actors are in a position
o impose rules on others, the employment of power may be .lf-reinfrciq.
Actors may utilize political authorit to change the rules of the game (both fr-
mal institutions and various public policies) to enhance their power. These
changes may not only shif the rules in their fvor, but increase their own ca
pacities fr political action while diminishing those of their rivals. And these
changes ma result in adaptations that reinfrce ese trends as undecided,
wea ommitte , or vulnerable actors j oin the winners or esert the losers.
Many political con ICts, rom t e azi seizure of power to the gradual process
through which the Labour Party supplanted the Liberals in Great Britain in the
early twentieth century, reveal this sort of dynamic. Disparities in political re
sources among contending groups may widen dramatically over time as positive
feedback sets in.
The disenfanchisement of Aican Americans in te post-Reconstruction
American South provides a clear and poignant example of how shifs in political
power can be self-reinfrcing. In Aexander Keyssar's marvelous study of the his
tory of sufrage in the United States, the end of Reconstruction precipitated a dy
namic process of shifing power relations that played out over a considerable
period of time (Keyssar 2000, pp. 1 07-1 6) . In 1 876, a contested presidential
election led to the removal of federal troops fom the South. In 1 878, Demo
crats won control of both houses of Congress fr the frst time in twenty years.
"The upshot of these events . . . ," Keyssar writes, "was to entrust the administra
tion of voting laws in the South to state and local governments" (p. 1 07) .
These breakthroughs, i t must be stressed, did not result i n immediate and
total victory fr Democratic "Redeemers" in the South. Instead, they ushered in
a "period oflimbo and contestation, of participation coexisting wit ef at ex
clusion" (p. 1 08) . In many parts of the South, the Republican Part "hung on,
and large, if declining, numbers of blacks continued to exercise the fanchise"
(p. 1 07):
Periodically they were able to frm alliances wit poor and upcountry whites
and even wit some newly emerging industial interest sympatetic to the
probusines policies of the Republicans. Opposition to the conservative,
planter-dominated Redeemer Democrats, therefre, did not disappear:
elections were contested by Republicans, by fctions within the Democra
tic Part, and eventually by the Farmers' alliance and the Populists. Conse-
POSITIVE FEEDBACK, PATH DEPENDENCE
quently, the Redeemers, who controlled most state legislatures, continued
to tr to shrink the black (and opposition white) electorate through gerry
mandering, registration systems, complicated ballot confgurations, and
the secret ballot (which served as a de fcto literacy test). Wen necessar,
they also resorted to violence and faudulent vote counts. (p. 107)
37
Social and political power was used over time to reinfrce and consolidate po
litical advantage. By the early 1 890s, maj or challenges to the Redeemers began
to dissipate, giving way to a durable system of planter hegemony. A late as 1 964,
only 10 percent of Aican Americans in Mississippi would be registered to vote.
A fcus on path dependence can serve to refcus the atention of social sci
entists on the role of power (Moe 2003; Thelen 1 999) . In the fmous commu
nity power debate of the 160s and 1 970s, Bachrach and Baratz ( 1 962) and
Lukes ( 1 974) argued persuasively that power asymmetries are ofen hidden fom
view; where power is most unequal, it ofen does not need to be employed
openly
'
alist critics essentially countered that it was impossible to sys
d1uate such claims (Polsby 1 963; Wolfnger 1 971 ) . Although he did not
fame the issue quite this way, Gaven ta ( 1 980) demonstrated that such power
asymmetries can refect positive feedback processes operating over substantial
periods of time. Processes of positive fedback can transfrm a situation of rela
tively balanced confict into one of great inequalit. Political settings where one
set of actors must initially impose their preferences on another set tfrough open
confict ("the frst fce of power") may change over time into settings where
power relations are so uneven that anticipated reactions and agenda control
("the second fce of power") and ideological manipulation ,''uethird fce")
make open political confict unnecessary. Thus, positive feedback over time
may simultaneously increase asymmetries of power and, paradoxically, render
power relations less visible.
THE COMPLEXI1 AD OPACITOF POLITICS
Economic theory is built around the usefl and plausible assumptions tat ac
tors know what they want, strive to get as much as they can, and are prett good
at doing so. Firms seek to maximize profts. The metric fr good perfrmance is
relatively simple and transparent. Prices send strong signals that fcilitate the
analysis of how various fatures of the economic environment afect frm per
frmance. Observable, unambiguous, and ofen quantifable indicators exist fr
many of these features. Workers can obtain firly good infrmation on the wages
and working conditions on ofer fom diferent frms. Consumers, too, are rea
sonably adept at navigating most aspects of the economic world. Links between
choices and outcomes are generally clear: take a new j ob and your income rises;
buy a car and your savings account balance shrinks. The qualit of goods is usu
ally evident in relatively short order, and repeated purchases allow consumers to
sample alternatives.
38 CHAPTER ONE
Of course, one can add many complications to tis simple picture of the
economic realm. The market is ofen highly complex and confsing. Yet the
presence of a unifing metic (prices), the absence of a reed to coordinate many
of one's economic decisions with tose of other actors, the prevalence of re
peated interactions, and the presence of relatively short causal chains between
choices and results greaty fcilitate te efort of economic actors to establish
priorities, construct sensible causal maps, and correct mistakes over time.
A I will explore in greater detail in Chapters Fur and Five, politics i s a fr,
fr murkier environment (Moe 1 990; Nor 1 990b). It lacks anything like the
measuring rod of price, involves the pursuit of a wide range of largely incom
mensurable goals, and consists of processes that make it very hard to observe or
measure important aspects of political performance. And, if we believe that a
system is not performing well, it is still more difcult to determine which ele-
ments in tese highly complex systems are responsible and what adj ustment
would lead to better result. The reliance on elaborate procedures to handle
collective-choice situations in politics is inescapable, but it undermines trans
parency-that is, it greatly increases transaction cost (Comes and Sandler
1996; Mueller 1 989) . The complexit of te goals of politics, and the loose and
difse links between actions and outcomes, renders politics inherenty ambigu
ous. A Nort has argued, "political market are fr more prone [than economic
markets] to inefciency. The reason is straightorward. It is extraordinarily dif
cult to measure what is being exchanged in political market and in conse
quence to enfrce agreements'' (Nort 1 990b, p. 362).
It is important to note that North is not simply arguing tat political decision
making is inefcient. Rather, politics gets stuck wit more difcult problems.
Were transaction cost are low, market mechanisms are likely to be efective.
They ofen break down, however, when transaction cost are very high. High
transaction costs are characteristic fr public goods (Comes and Sandler 1 996;
Mueller 1 989). Thus, it is complex and ambiguous issues and problems that
gravitate toward te public sphere.
Even if mistakes or filures in politics are apparent, improvement through
"trial-and-error" processes is fr fom automatic. Many participant in politics
(voters, members of interest groups) engage in activities only sporadically. Their
tools of action are ofen crude, such as the blunt instument of the vote, and
their actions have consequences only when aggregated. There may be long lags
and complex causal chains connecting these political actions to political out
comes. The result is tat mistaken understandings ofen do not get corrected.
The point is not that learing never occurs in politics. Rater, learning is very
difcult and cannot be assumed to occur. Instead, understandings of te politcal
world should temselves be seen as susceptible to pat dependence. Drawing on
work in bot cognitve psycholog and organizational teory, researchers argue
tat actors who operate in a social context of high complexit and opacity are heav
ily biased in the way tey flter infrmation into existing "mental maps" (North
POSITIVE FEEDBACK, PATH DEPENDENCE 39
l 990b; Denzau and Nort 1 994; Aur 1 994). Confrming infrmation tends to
be incororated, while disconfrming information is fltered out. Social interre
tations of complex environment lik politcs are ofn subject to positve feedback.
The development of basic social understandings involves high start"up costs and
learning efects-as Philip Converse notes, infrmation must be stored in an ef
fctive way in order to fcilitate te processing of additional infrmation (Con
verse 1 991 ) v The need to employ mental maps induces positive feedback.
The pat-dependent nature of views of the social world, evident at the indi
vidual level, is even more apparent at the group level. A Alexander Wendt
notes, "identities and interests are not only learned in interaction, but sustained
by it'' (Wendt 1 999, p. 3 3 1 ) . Ideas are fequenty shared with other social actors
in ways tat create network efects and adaptive expectations. Sociologist have
emphasized that the development of norms or standards of appropriateness is a
collective, selfreinfrcing process. Social interactions involving activities as in
nocuous as a handshake "chronically reproduce" themselves (Jepperson 1 991 ;
see also Wendt 1 999, pp. 1 84-89). Every time we shake hands, te strength of
that norm is reinfrced.
The same argument can be applied wit considerable frce to collective
understandings -of how the world works, what is to be valued, what an individ
ual's interests might be, and who tat individual's fiends and enemies might
be-in short, to collective ideational constructs ranging fom policy paradigms
to fll-fedged ideologies (Hall 1 993; Berman 1 998; Bly 2002). Robert Wth
now's subtle analysis of te comparative development of ideologies has eleganty
shown how emerging worldviews, once tey reach a critical mass, can generate
a set of culture-producing institutions, organizations, and specialized actors tat
greaty fcilitate te spread and reproduction of that ideology (Wuthnow 1989) :
Wuthnow demonstrates how relatively brief periods of historical openness are
ofen fllowed by processes tat select and then institutionalize a particular
track of ideological development. His account of how "communities of dis
course" ofen,come to share, institutionalize, and reproduce a similar ideolog
is strikingly consistnt with the famework suggested here. Berman ( 2003) has
made a similar argument about te spread of radical Islam, in which extremists
have used their control over key institutions of cultural production to fment a
revolutionary transfrmation in citizens' worldviews -even ir te absence of a
revolutionary overtrow of the state itelf.
North's work on "mental maps" thus converges with long-standng views of
those studying political culture as well as te recent contributions of cognitive
science. 1 9 Once established, basic outooks on politics, ranging fom ideologies
19Consider Karl Mannheim in his fmous essay on generations: "It is of considerable importance fr
the frmation of consciousness which experiences happen to make tose all-important 'frst impres
sions.' . . . Early impressions tend to coalesce into a natral vie of te world. All later experiences
ten tend to receive teir meaning fom tis original set, wheter tey appear as tat set's verifca
tion ard flflment or as its negation and antitesis" (Mannheim 1952).
40 CHAPTER ONE
to understandings of particular aspects of governments or orientations toward
political groups or parties, are generally tenacious. They are path dependent.20
"Social systems,'' as Wendt notes, "can get 'locked in' to certain patterns by the
logic of shared knowledge, adding a source of social inertia or glue that would not
exist in a system without culture" (Wendt 1 999, p. 1 8&) .
There are, then, compelling reasons to believe that political processes will
ofen be marked by dynamics of increasing returns. Tendencies toward positive
fedback characterize fur processes central to political environment: collective
action, institutional development, the exercise of authority, and social interpre
tation. In each case, tere are reasons to anticipate tat steps in a particular direc
tion can triger a self-reinfrcing dynamic. This conclusion should be underlined.
By itself, it suggests why positive feedback is a critical concept fr those who seek
to understand the sources of political stabilit and change. If a recognition of te
signifcance of self-reinfrcing processes is shaking up economics, then those
studying politics have at least as great a need to consider it implications.
There is also reason to believe that these effects in politics are ofen particu
larly intense. In the remainder of tis section I briefy outline some reasons why,
given the onset of some process of positive feedback, it is fequently more dif
cult to reverse course in politics than it would be in economics. These issues are
pursued in greater detail in Chapters Four and Five.
Mechanisms for Rversing Course
One of the reasons economists were slow to worry about path dependence was
because tey believed te market provides two powerfl mechanisms fr exiting
problematic paths: competition and learning. Competitive pressures in a mar
ket society mean tat new organizations with more efcient structures will de
velop, eventually replacing suboptimal organizations (Alchian 1 950) . Leaming
processes within frms can also lead to correction. Firms learn fom their own
experiences, as well as those of other frms, and can correct mistakes over time
(Williamson 1993).
It is worth emphasizing that neither of these mechanisms represents a guar
anteed corrective in the pat-dependent contexts explored by Arthur, North,
and others. Options that gain a head start will ofen reinfrce themselves over
time, even if they have serious shortcomings. What I wish to stress, however, is
20Indeed, as marketers know well, such path-dependent cognitive efects are evident even in the less
ambiguous world of consumption. This is why advertisers covet the attention of youngsters who have
yet to make defnitive (and resilient) choices. A telling recent example is the new marketing efort of
the National Fotball League, which is alarmed by indications that youngsters are increasingly
drawn to basketball and soccer. A frmer M1 executive now working on special events speaks te
language of increasing returns: "It's all about getting a fotball . . . into a kid's hands as soon as you
can. Six years old, if possible. You want to get a fotball in their hands befre someone puts a bas
ketball in their hands, or a hockey stick or a tennis racquet or a golf club" (Seabrook 1 997, p. 47).
POSITNE FEEDBACK, PATH DEPENDENCE 41
that these corrective mechanisms are even less efective when one shifs fom
frms in private markets to te world of political institutions (Moe 1 984, 1 990).
This i s clearest fr mechanisms of competition. While models of competition
may be helpfl fr understanding some important aspects of politics (such as in
ternational relations and elections), there can be lite doubt that political envi
ronments are typically more "permissive" than economic ones (Krasner 1 989) .
A just discussed, te complexity and ambiguit of politics create serious prob
lems fr learning arguments. It may be appropriate to argue tat politics some
times involves learning processes, in which responses to public problems proceed
in a tial-and-error fshion (Lindblom 1959; Hecla 1 974; Hall 1993). There is lit
te reason, however, to tink tat tis acts as a selection mechanism wit anything
like the efciency-enhancing properties of market competition in economics or
Darwinian natural selection in biology. Because political reality is so complex
and the tasks of evaluating public performance and determining which options
would be superior are so frmidable, such self-correction is ofen limited.
Even where learning does occur, refrms noneteless fce all the barriers to
change that are characteristic of systems exhibiting positive feedback. Long
movement down a particular path will have increased the costs of switching to
some previously fregone alternative. Furthermore, in politics the pursuit of
such change fces two additional obstacles: the short time horizons of political
actors, and the strong status quo bias associated wit the decision rules govern
ing most political institutions. These fctors will ofen make path-dependent ef
fects particularly intense in politics.
TIME HORIZONS
A statement attributed to David Stockman, budget director during the Reagan
administration, is unusual among political decision makers only fr its candor.
Asked by an adviser in 198 1 to consider pension refrms to combat Social Se
curit's severe long-term fnancing problems, Stockman dismissed te idea out of
hand. He explained that he had no interest in wasting "a lot of political capital
on some other guy's problem in [the year] 201 0" (quoted in Greider 1 982, p. 43) .
Many of the implications of political decisions -especially complex policy
interentions or major institutional refrms -only play out in the long run. Yet
political actors, especially politicians, are ofen most interested in the short-term
consequences of teir actions; long-term efects tend to be heavily discounted.
The principal reason is te logic of electoral politics. Because te decisions of vot
ers, which determine political success, are taken in the short run, elected of
cials generally employ a high discount rate. They will pay attention to long-term
consequences only if these become politically salient, or when they have litle
reason to fear shor-term electoral retribution. Keynes once noted that in the long
run, we are all dead; politicians have special reason to take that message to heart.
We know relatively litle about the time horizons of diferent political actors.
A interesting literature is developing on "credible commitments" -the attempt
42 CHAPTER ONE
of political actors to create arrangements tat fcilitate cooperation (North and
Weingast 1989; Shepsle 1 991 ; North 1993). This research suggests that particu
lar institutional designs (such as independent central banks), empowering par
ticular kinds of political actors, may succeed in lengthening time horizons. In
general, however, such mechanisms will ofen be oflimited efectiveness in pol
itics. A noted, te marketplace possesses some stong mechanisms fr lengten
ing time horizons -especially the basic continuit of frms over time and the
presence of capital markets. Such mechanisms in politics are generally fr
weaker. Monitoring political behavior over time is difcult because the interac
tions that go into generating perfrmance are so complex, and even the indica
tors of perfrmance are tpically so limited. The relatively rapid turnover of
critical positions also makes it hard to hold actors accountable. Politics, in short,
lacks the characteristic property rights that fcilitate the linkage of actors' deci
sions over time in the economic sphere. In many cases, the long term is essen
tially beyond the political horizon. A statesman, Bismarck said, is a politician
who thinks about his grandchildren.
The diferent nature of time horizons in politics and in economics matters a
lot. This can be seen by revisiting te Liebowitz and Margolis critique of path
dependence. They properly point to key market institutions as a protection
against certain kinds of remediable path dependence. If it is believed tat long
term beneft will be greater using option B, ten investors should gravitate
toward that option even if in the short-term it will perfrm more poorly than op
tion A. Thus, they argue that market mechanisms should allow te more ef
cient outcome (B) in fgure 1 . 1 .
In politics the outcome may well be diferent. Assume that te crucial decision
maker is a politcian up fr reelection in two years. In tis context, efect afer te
election cycle do not count fr much. 21 A politician fcusing on the short-term
payof would choose option A The diference in time horizons has profund
consequences. If time horizons tend to be short, ten we can expect that long
term costs and benefts will have a limited efect on the chosen path. Further
more, once on a particular path, political actors will generally have powerfl
incentives to stay on it. Switching costs are tpically borne in the short run, and
te beneft will generally only accrue in the long run, that is, to someone else.
THE STATUS QUO BIAS OF POLITICAL INSTITUTIONS
In the economic realm, an individual wit a new idea fr a product need only
secure te fnance to put it on te market. If enough consumers (choosing inde
pendenty) fnd it sufcienty appealing, the product will be a success. Change
21These long-term efects will count if an actor wit longer time horizons (such as an interest group)
is able to make tem relevant to politicians-e.g., through campaign contributions or votes. The
question is whether such mechanisms are anywhere near as efective as te capital markets opera
tive in te economic sphere. In my view, tere are stong reasons to be skeptical of this, but it is
clearly an issue deserving considerable attention.
POSITIVE FEEDBACK, PATH DEPENDENCE 43
can be engineered through competition against existing products. Similarly,
those with property rights over a frm are generally in a strong position to remake
their organizations as they choose. Lines of autority are clear, and the relevant
decision makers are likely to share the same broad goal of maximizing proft.
By contrast, key features of political lif, bot public policies and (especially)
frmal institutions, are change resistant. Policies and institutions are generally
designed to be difcult to overturn fr two broad reasons. First, tose who design
institutions and policies may wish to bind their successors. Moe ( 1 990) terms
this the problem of "political uncertaint." Unlike economic actors, political ac
tors must anticipate that their political rivals may soon control the reins of gov
ernment. To protect themselves, tese actors therefre create rules that make
existing arrangements hard to reverse.
Second, in many cases political actors are also compelled to bind themselves.
The key insight of the "credible commitment" literature is that actors can ofen
do better, even in te relatively short run, if tey remove certain alternatives
fom their fture menu of options. The economy of a country will grow fster,
fr instance, if a monarch can credibly commit himself to refain fom expro
priating an excessive amount of te hard-earned wealth of his subj ects (North
and Weingast 1 989). This can be done if he accedes to parliamentar control
over the power to tax. Like Ulysses preparing fr te Sirens, political actors ofen
bind themselves, restricting their own feedom in order to achieve some greater
goal. To constrain themselves and others, designers create institutions that are
sticky. Stickiness is built into the design of political institutions to reduce uncer
tainty and enhance stability, fcilitating frms of cooperation and exchange that
would otherwise be impossible.
Formal barriers to institutional refrm are thus ofen extremely high, such as
unanimity requirements in te European Union and multiple supermajorities
to alter te Aerican constitution. Of course, these obstacles may fcilitate frms
of cooperation and exchange tat would otherwise be impossible. The relevant
point here is that tis institutional stickiness characteristic of political systems re
infrces the already considerable obstacles to movement of an established
path.22 Combined wit the weakness of competitive mechanisms and learing
22An important characteristic of political systems runs counter to tis line of argument. Because pol
itics is a powerfl system fr mobilizing coercive power, governments may at times be in a position
to orchestrate a "j ump" fom one pat to anoter. Government, by employing sanctions, can coor
dinate adjustments in a way tat markets might never be able to achieve. Fr instance, te British
government was able to enact a shif to te metic system tat would have been difcult or impossi
ble to engineer trough te more atomistic mechanisms of the market. And goverments are clearly
capable, on occasion, of mobilizing resources fr more dramatic changes in course. Such possibili
ties, however, should not be exaggerated. The metic example represents a relatively modest in
stance of reversing path dependence. Switching costs were low; te problem was essentially one
of coordination-inducing everyone to make te switch at te same time. For tis task, the author
itative rule-setting capacities of goverment are of great assistance. For reasons already discussed,
4 CHAPTER ONE
processes, as well as te short time horizons characteristic of politics, the bias
means tat tendencies toward path dependence in political development are
ofen particularly intense.
Politics difers fom economics - in many Ways. Applying tools of economic
analysis to politics is treacherous unless these diferences are systematically con
sidered. In the case of arguments about path dependence, attention to the char
acter of politics suggests a striking result. The political world is unusually prone
to positive feedback, and te capacities fr reversing course are ofen weak. Bot
te prevalence and intensit of tese processes in politcs suggest tat pat depen
dence arguments ofer an important tool fr understanding political dynamics.
PATH DEPENDENCE AND THE STY OF POLITICS
To summarize briefy, in settings where self-reinfrcing processes are at work po
litical life is likely to be marked by fur fatures:

ultiple equilibria. Under a set of initial conditions conducive to posi

ve fedback, a range of outcomes is generally possible.
: ontingency. Relatiely small events, if occurring at the right moment,
can have large and enduring consequences.
. A critical role fr timing and sequencing. In these path-dependent
processes, when an event occurs may be crucial . Because early parts of
a sequence mater much more than later part, an event that happens
"too late" may have no efect, although it might have been of great con
sequen

e if te timing had been diferent.

.
Inertia. Once such a process has been established, positive feedback
will generally lead to a single equilibrium. This equilibrium will in turn
be resistant to change.
There are also good reasons to think that positive fedback processes are wide
spread in politics, since they will be characteristic in institutional development,
collective action, the exercise of authority, and the emergence of our under
standings of the political world. This has fndamental theoretical implications.
We need to change bot te kinds of questions we ask about politics and the
kinds of answers that we generate. Many of these implications will be explored
in greater depth later, but a brief summary here will clarif te organizational
lte chapters to fllow.
.
"
path-dependent arguments point to the )ignifcance of sequencin
tleporal order in which social events or processes unfld. Most variable-
it is much less evident tat governments will generally be willing or able to engineer shifs t a dif
ferent pat when switching costs are high. Cases of fndamental or revolutionary refrm in well
institutionalized political systems attract our attention precisely because they are so rare.
POSITIVE FEEDBACK, PATH DEPENDENCE 45
oriented research assumes a
.
. 9rld without ositive feedback, where histor
washes out and sequence is irrelevant. We only nee to now the values of vari
ables at the moment of interest, not te sequence through which these fctors de
veloped. In path-dependent processes, however, positive fedback means that
history is "" These processes can be_highly nfuenced b relative!
modest perturbations at early stages. "Small" events ear y on may ave a big im
pact, while "large" events at later stages may be less consequential. To put it an
other way, outcomes of early events or processes in the sequence may be
amplifed, while the signifcance oflater events or processes is dampened. [,
) when a particular event in a se uence occurs will make a bi
e
. A crucial
imp 1cabon o pa - epen ence arguments is that early stages in a sequence can
place particular aspects of political systems onto distinct tracks, which are ten
reinfrced through time.
Second, a fcus on path-dependknt processes suggests the need to develop
analyses that may incorporate substantiastretches of time. For this reason, as
with their highlighting of sequencing, they justif a turn to history. At one level,
of course, all social scientists agree that "histo matters." The existing condi
tions that infuence current social outcomes came into being at some point i
te pas|Those earlier processes are relevant to a fll understanding of contem
porary social events/Yet the standard assumption is that fr most purposes we
may safely put such issues aside, Looking back leads to the fmiliar problem of
infnite regress. A exploration of each preceding event leads to the conclusion
tat some other preceding occurrence was also part of te . chain of necessary
events, and so on. Social scientists need to break through the seamlessness of his
tory somewhere, and the present is as good a place to do so as any. George
Homans ( 1 967) compared the situation to that fced by mine sweepers who
needed to know the magnetic charge of a ship. Such a charge resulted fom an
infnite range of small fctors accumulated over the ship's lifetime. For practical
purposes, however, a simple expedient could be used: the current charge of the
ship could be measured. Ifthe task is to understand the ship's vulnerabilit to
mines, one can simply cut through the Gordian knot of historical regress. '

For many purposes, this is an appropriate approach. Social scientists ofen

ve good reason to fcus on synchronic causality-to try to understand how
riations in current variables afect present social outcomesAWhere processes
of positive fedback are signifcant, however, such a strategy may be problem
atic. Pat-dependent arguments rest on what Artur Stinchcombe has termed a
conception of "historical causes" ( Stinchcombe 1 968, pp. 1 03-1 8; see also
Harsanyi 1 960; Ikenberry 1 994; Jervis 2000, p. 97; Lieberson 1985 ) -some ini
tial event or process generates a particular outcome, which is ten reproduced
through time even though the original generating event or prcess does not recur.
21Fr a discussion of Homans's argument, see Krapp 1983, pp
,
43-5.
46 CHAPTER ONE
This is very diferent fom the more tpical search fr invariant relationships
among fctors, in which te analyst assumes tat if adding x to a setting causes y,
then te removal of x should remove y as well. A Stinchcombe puts it, a "his
toricist explanation . . . is one in which an efect created by causes at some pre
vious period becomes a cause of that same e ect in succeeding period. In such'
arguments, the problem of explanation breaks down into two causal compo
nents. The frst is the particular circumstances that caused a tradition to be
started. The second is te general process by which social patterns reproduce
themselves."
In such a process te crucial obj ects of study become te fctors tat set de
velopment along a particular path-and which lie in the past-and the mecha
nisms of reproduction of te current pat, which at frst glance mig seep.
c?mmonplac

, perhaps almost invisibe or at least analytically uninteresting.

This distinctive approach to explanation can be clarifed trough refrence to
the Polya um illustration intoduced at the outset. Suppose we sought to under
stand why te distribution of balls on the hundredt trial was 96 red/4 black. We
would not be very happy wit a Homans-style explanation: "because afer 99 tri
als the dis
.
tribution was 95/4, and thu
5
there was a better than 95 % probabilit
that the next draw would also be red.'JWhat we would want to know would be
te initial event that established this trajectory, and the mechanism (in this
case, a process of selection governed by a particular decision-rule) that made
that trajectory self-reinfrcing. Thus, where historical causation is at work, ex
planation requires the examination of considerable stretches of time. This issue,
along with oters related to the investigation of long-term social proceses, is
tn dept in Chapter Three. f
tird\crucial implication is that pat-dependent arguments provide a plau
sier to fnctionalist explanations of social outcomes, which too ofen
go unchallenged. Altough not always explicity stated, fnctionalist arguments .
are prevalent in the social sciences. They are common, fr instance, among
tose who emphasize the rational choices of individual actors tat underlie po
litical activit, and te reasonably efcient nature of collective responses to so
cial needs (Keohane 1984; Shepsle 1 986; Weingast and Marshall 1 988) .
Functionalist arguments take te fllowing frm: outcome X (an institution,
policy, or organization, fr instance) exists because it seres the fnction Y. In a
world of purposive actors, it may indeed be the case tat te efect of an insti
tution have someting to do wit an explanation fr its emergence and persis
tence. Aguments about positive fedback, however, suggest te large dangers in
any assumption tat an existing institution arose or continues to exist because it
serves some particularly usefl purpose. ]gugin fnctionalist terms ahQl
ap existi_!instjtion. polic or social organization may be a good w to derive
causal Joteses, but fnctional account are fr fom being the only plausible
Qes. Many alternatives to te outcome in question might have been possible,
and a dynamic of positive
.
fedback may have insitutionalized a particular op-
POSITIVE FEEDBACK, PATH DEPENDENCE 47
tion even though it originated by accident, or the fctors that gave it an original
advantage have long since passed away. Rather than assuming reltive efcienc
a an explanation, we have to go back and look. Thus recognizing the possibility
of path dependence necessarily draws social scientist to an investigation of his
tory, if only to evaluate te validit of fnctionalist assertions. A I explore in
Chapters Four and Five, an investigation of path dependence can provide a
basis fr developing alternative hypotheses about institutional origins, stabilit,
and change.
Consider one example. A prominent theme in recent research in compara
tive political economy is the idea of "varieties of capitalism." Despite increasing
international economic interdependence, which seems . to . generale pressures
toward convergence, the advanced industrial societies continue to exhibit fn
damental diferences in their core institutional structures (Soskice 1 999; Hall
1 999; Berger and Dore 1 996; Hollingswort and Boyer 1 997). To date, however;
this literature has doe a better j ob of identifing and describing tis diversity
than it has of explaining what generates and sustains it. Hall and Soskice (200l a)
have made an important
.
step frward by emphasizing the role of institutional
complementarities. The benefts of particular economic institutions and orga
nizations are increased if tey ft well in an environment populated by specifc
kinds of institutions and organizations.
The "varieties of capitalism" analysis persuasively illuminates distinct equi
libria in diferent economies, but it does not address te question of how these
distinct equilibria emerge. From te curre1 analysis, one can easily see why te
elaborate roducti ms ofmodemecono i ou be sub" ect to o
Start-up costs, not j ust fr new frms, b1t (more fndamentally) fr
te key organizations and institutions tat link private actors, are enormous. Or
ganizations, and te frmal and infrmal arrangements (bot public and private)
that help to structure teir interactions create, as North would put it, densely
linked institutional matrices. Economic and social organizations and political
institutions (both basic constitutional arrangement and public policy fame
works) have coevolved over extended periods of time. Coordination efects are
widespread; particular courses of action are encouraged or discouraged because
of anticipated actions of others in the system. Firms have developed sophisti
cated stategies suitable to te particular institutional matrix tey confont
that' is, tremendous amount of learning by doing have occurred over time in
these complex systems. In short, national economic systems are highly path de
pendent. They are likely to exhibit substantial resilience, even in the context of
major exogenous shocks such as recent changes in te global economy.
Highlighting the possible limits of fnctionalist explanations reveals a fnal
teoretical beneft of recognizing te signifcance of pathdependent processes:
tese arguments provide te basis fr a revitalized efort to investigate issues of
power in social lif. Functionalist arguments tat start fom the benefts tat par
ticular actors derive fom instittions and infer tat their power accounted fr
48 CHAPTER ONE
those arrangements typically ignore important feedback processes that may gen
erate the same observed outcome in a completely diferent way (Hacker and
Pierson 2002) . A j ust emphasized, an understanding of pat-dependent dynam
ics may suggest alternative explanations fr arrangements that are too easily at
tributed to power relations. At the same time, the possibility that self-reinfrcing
processes may magnif power imbalances over time, while simultaneously ren
dering those imbalances less visible, indicates that "snapshot" views will ofen
miss important element of power relations. In short, both these observations
point to deep-rooted difculties in the treatment of power in contemporary so
cial science: sometimes analyst see power when it is not there, and sometimes
they do not see it when it is.
Some Initial Concers
Because the preceding discussion provides the basis fr much of what fllows, it
may be helpfl to stop and briefy address some possible obj ections to the line of
argument developed so fr.
OLD WINE IN NEW BOTLES?
Of course, there is a long tradition of attention to history in the social sciences.
Particularly fr tose who want answers to critical questions that grow out of the
experiences of real polities, the trn to history has been common. Issues of tim
ing, sequence, and critical junctures fgure prominenty in this body of work.
Among many such studies, Gerschenkron's study of industrialization and state ,
building (Gerschenkron 1 962) and Lipset and Rokkan's analysis of te forma
tion of part systems (Lipset and Rokan 1967) are two classic examples. Indeed,
it is fir to ask whether incorporating the concept of positive feedback and path
dependence into te study of politics is akin to the man who discovered that he
had been speaking prose all his life. Is path dependence merely a trendy name
fr old ideas?
Discussions of path dependence would be worth having if they did no more
than fcus the attention of fd-prone social scientists on the insight and contin
uing relevance of this earlier body of work. Yet there is every reason to believe
that the concept can do more. Awareness of the dynamics of positive feedback
processes can greaty sharpen our understanding of why particular junctures
(and which aspects of those j unctures) are critical and why timing ofen counts
fr so much in politics. Most of tis earlier work was vague on this crucial point.
The specifc characteristics of positive feedback provide a key to making sense of
the complex mix of stabilit and change tat characterize so many political
processes. To repeat, Arthur's work on path dependence is groundbreaking not
simply because he described te characteristics of these processes, but because
he identifed conditions likely to generate tem. Building on North's work,
we can now begin the process of adapting these arguments to the study of poli-
POSITIVE FEEDBACK, PATH DEPENDENCE 49
tics. Doing so requires carefl attention to the distinctive fatures of the politi
cal world, such as its intrinsic ambiguity, the prevalence of highly sticky institu
tions, the prominence of collective action problems, and te prospects fr
utilizing policy autority to amplif asymmetries of power. A investigation of
self-reinfrcing processes can generate sharper hypotheses, based on more ex
plicit social mechanisms, about the sources of divergent pats and social inertia.
ARE PATH-DEPENDENT ARGUMENTS JUST DESCRIPTIVE?
Without carefl attention to te identifcation of the mechanisms at work,
analyses of path dependence can easily become descriptions of what happened
rather than explanations fr why it happened. Thelen ( 1999, p. 391 ) nicely sum
marizes this concern:
Arguments about the "feezing" or "crstallization" of particular institu
tional confgurations . . . obscure more than they reveal unless they are ex
plicitly linked to complementary arguments tat identif the mechanisms
of reproduction at work. Witout these, they are at best incomplete, fr
they cannot explain why these paterns persisted and how they continue to
dominate the political space. . . . [ Sometimes] authors invoke Stinch
combe's arguments about "sunk costs" and "vested interests" that make em
barking on alternative paths costy and uncertain. But such references,
though a promising starting point fr te analysis, cannot themselves re
place the analysis; these concepts need to be applied, not j ust invoked.
Among other things, we need to know exacty who is invested in particular
institutional arrangements, exactly how that investment is sustained over
time, and perhaps how those who are not invested in the institutions are
kept out. 24
The key, as Thelen suggests, is to specif the mechanisms that reinfrce a pr- _
icula
.
r path or trajectory. itout this,

a
.
th-dependent arguments degenerate
mto litte more than a descnpt10n of stability.

At the same time, a fcus on these mechanisms should help us to explain vari
ation across setings. Although I have argued that many aspects of politics will
promote strong tendencies toward self-reinfrcement, not all aspects of political
life are subj ect to positive fedback. One can think about this in terms suggested
by Hannan and Freeman ( 1989, p. 1 06), who discuss te "mixing of diferent tes
of outcomes in [al spatial or temporal distribution." Where there are long runs
24Stinchcombe ( 1965, p. 1 67) made a similar observation: "It is considerably more difcult to ex
. plain why many types of organizations retain structural peculiarities afer their fundation without
flling into tautologous statement about 'tradition; 'vested interests,' or 'flkways' not being change
able by frmal regulation. The problem is to specif who it is tat carries 'tradition' and why tey
carry it, whose 'interests' become 'vested,' under what conditions, by what devices, whose 'flkways'
cannot be changed by regulation, and why. This problem is at te very center of sociological theory."
Stinchcombe 1965, p. 1 67.
50 CHAER ONE
of one outcome, one can describe these distributions as "coarse grained," while
te opposite pattern would be fne-grained. Pat-dependent processes will tpi
cally generate coarse-grained paterns of outcomes rater than fne-rained ones.
Consider a concrete example. A Mayhew notes (2002, p. 1 29 ) , despite a re
alignment literature tat suggest one or anoter political part should gain the
upper hand as a "naturl" govering majorit, te part wit te advantage in part
identifcation rates has shown almost no advantage in presidential elections since
1 900. In other words, the outcomes of presidential elections are "fne grained"
rather than "coarse grained."25 A plausible explanation is tat pesidential can
didates, at least in te American context, retain considerable fexibili in adopt
ing nategies and appeals t increase teir electability. A part losing one
presidential election is at lite systematic disadvantage in te next. In other
words, presidential elections do not appear to be highly path-dependent. Part of
te reason fr tis outcome, however, is tat losing parties are relatively fe_o
j etison unpopular policy positions and adopt more populr ones-including
cnes introduced by te incumbent part. Thus, presidential elections may be
fne grained in part because p)itend to be coarse grained-that is,
marked by very substantial stretches of stability (Huber and Stephens 200 l; Hall
and Soskice 200la) . Parties maintain teir electoral competitiveness by adapt
ing to the policy successes put in place by teir competitors.
We should expect consideble vriation in te extent to which strong pat
dependence is evident in various element of the social world, depending on the
presence or absence of the kinds of fctors mentioned here, as well as the pres
ence or absence of countervailing fatures of the social landscape. It is here that
recent teoretical developments in work on path dependence have the greatest
promise, fr tey ofer precisely t his -a set of propositions about te tpes of cir
cumstances that promote positive feedback. Specifing mechanisms helps us to
develop hypotheses about where we might expect (or not expect) to encounter
similar dynamics -in oter words, it helps us to develop portable claims.
CONTINGENCY AND DETERMINISM
{ain, Thelen ( 1 999, p. 385) has raised a reasonable concern that path
dependent models can be "too contingent and too deterministic" -tat is, too
open at the font end or critical j uncture and too closed at the back end (once
the critical juncture has passed) .
\
Critics have argued that pat-dependent argu
ments seem to suggest a very high level of contingency at critical j unctures,
while the "mechanisms of reproduction" tat fllow generate an overly static
25Interestingly, tis pattern is not always evident cross-nationally. Patterns in some counties are
"coarse-grained;' revealing extended periods of single-part dominance (Pempel 1990). Based on
te cur
rent analysis, one could useflly fcus on what fctors in tese diferent polities generate pos
itive feedback fom an electoral victory, and make it difcult fr a losing party to adapt in a compet
itive fshion.
POSITE FEEDBACK, PATDEPENDENCE 5 1
view ofthe social world (Katznelson 2003; Schwartz, n. d. ). To take the starkest
illustration, Aur's Polya um processes are totally random at te outset-the
smallest perturbation can make all te diference-and yet tey all sete on a
particular equilibrium and ten essentially stop. Positive feedback processes
seem to generate only brief moments of "punctuation" in a largely fozen social
landscape.
A critical feature of pat-ependent processes is te relative "openness" or "per
missiveness" of early stages in a sequence compared with the relatively "closed"
or "coercive" nature of later stages (Mahoney 2001 ; Abbott 1997) . Viewed ex
ante, such processes can produce more tan one possible outcome. Once a par
ticular pat gets established, however, self-reinfrcing processes are prone to
consolidation or institutionalization. "Critical junctures" generate persistent
paths of political development.
26
It is necessary to stress these fatures of path-dependent processes while avoi:..
_ing a tendency to overinterret their imHcatiros. These fatures of compara
tively "open" processes at the outset, fllowed by a more constrained choice-set
once reinfrcement sets in, are precisely the fatures that make this type of
process distinctive and underpin the claim that temporal sequence is crucial
(Mahoney 2001 ) . At te same time, however, such argument rarely, if ever,
suggest that "anyhing goes" at te initial branching point. Rater, the claim is
that there may be more tan one alternative (ofen, as in the discussions of path
dependent technologies that have been common in economics, the analysis will
specif two), and that cc||irc|ysmall fctors may push the outcome
.
one way or
the other. Aong the same lines, claims tat processes are pat-dependent do not
require that te fctors leading to one path rather than the other be truly ran
dom, or beyond the reach of teory. Athough sometimes these j unctres are
treated as highly contingent or random, generally analysts seek to generate con
vincing explanations fr why one pat rather than anoter was chosen:
The exlanations, however, will ofen emphasize event or processes tat
seem "small" when compared wit te large efect: (maj or, lasting divergences
across cases) that they produce once positive fedback processes have had a
chance to amplif te initial repercussions. In analyzing these nonlinear
processes we cannot assume that "large" efects are te results of'1ae" causes
(Abbott 1 988). To put it anoter way, if one imagines a counteractual in which
an alternative outcome emerges, the size of the change needed to generate te
diferent outcome will be smaller-perhaps much smaller-at te onset of a
self-reinforcing process than it will be at a later date. Hacker's analysis of the
26
Atough analyses invoking te language of "critical junctures" sometimes fcus on large-scale,
dramatic events, those qualities are neither necessary nor sufcient to generate path-dependent dy
namics. In fct, te point)n path-dependent analyses is tat "causes" may ofen seem relatively small
compared wit teir efects . . Wat makes a particular juncture "critical" is tat it tiggers a process
of positive fedback.
52 CHAPTER ONE
development of health-care policy in the United States ofers an excellent ex
ample (Hacker 2002). Hacker does not argue tat the filure of the United
States to adopt national health insurance during the New Deal period was
highly contingent-there were good reasons why this was the likely outcome.
What he does argue is that this initial development generated powerfl positive
fedback, institutionalizing a set of private arrangements that made it much
more difcult to make a transition to national healt insurance at a later point
in time. A a result of these self-reinfrcing processes, the scale of the counter
fctual needed to imagine an alternative pat, Hacker suggests, is much greater
in the year 2000 than it was in 193 5 .
I f the suggestion that path-dependent arguments imply hypercontingency at
the outset strikes me as a red herring, so does the suggestion that they point to a
world of stasis. Nothing in path-dependent analyses implies tat a particular al
ternative is permanenty "n" fllowing the move onto a self-reinfrcing
path. Identifing self-reinfrcing processes does help us to understand why or
ganizational and institutional practices are ofen extremely persistent-and tis
is crucial, because these continuities are a striking feature of the social world. As
serting that the social landscape can be permanently fozen is hardly credible,
however, and that is not the claim. Change continues, but it is bounded
change -until something erodes. or swamps the mechanisms of reproduction
that generate continuit. Douglass North summarizes the key point well:
"At every step along the way tere [are choices ] -political and economic -tat
provide . . . real alternatives. Path dependence is a way to narrow conceptually
the choice set and link decision making through time. It is not a stor of in
evitabilit in which the past neaty predict te fture" (North l 990a, pp. 98-99) .
The claims in path-dependent arguments are that previously viable options may
be freclosed in the afermat of a sustained period of positive feedback, and cu
mulative commitments on the existing path will ofen make change difcult
and will condition the frm in which new branchings will occur.
Indeed, as I explore in greater detail in Chapters Two and Five, some of the
most interesting developments in the work on path dependence fcus specif
cally on these issues of"downsteam" development in path-dependent processes.
A recenty emphasized by both Mahoney and Thelen, identifing the particu
lar fedback loops (or "mechanisms of reproduction") at work will ofen provide
key insights into the kinds of events or processes that might generate major sub
sequent change points (Thelen 1 999, 2003; Mahoney 200 1 ) . Such j unctures
are usually atributed, ofen ex post, to "exogeneous shocks." We should expect,
however, tat these change points ofen occur when new conditions disrupt or
overwhelm the specifc mechanisms that previously reproduced the existing
path. Thus, a clear understanding of the mechanisms of reproduction provides
an instrument fr te investigation of change.
Similarly, analyses of historical sequences may fcus precisely on te dy
namic downstream consequences of particular paterns of instutionalization fl-
POSITNE FEEDBACK, PATH DEPENDENCE 5 3
lowing a critical j uncture. As I suggest i n the next chapter, analysts sometimes
wish to know about te elimination of particular alternatives at a key moment
not because the outcome at that stage is permanenty locked in, but because te
removal of certain options trough path-dependent processes creates diferent
outcomes at a later choice-point in a historical sequence. Thus path-dependent
arguments ofen involve a more complex position on issues of stability and
change than the simple "contingency fllowed by lock-in" frmulation suggests.
When sequences involve self-reinfrcing dynamics, we can expect periods of rel
ative (but not total) openness, fllowed by periods of relative (but not total or
permanent) stability.
A common thread runs through my discussion of each of these concerns
about path-dependent arguments: the explication of specifc mechanisms that
generate path dependence is the key to making this a fuitl line of theorizing
about the sources of social stability and change. And this in tm is te main jus
tifcation fr taking te time to work trough key issues about these mechanisms
in economic analysis, and te implications of shifing fom te economic realm
to the political. In the chapters tat fllow I will provide many illustrations of
these mechanisms at work in important political processes.