Pierson Skocpol 2002

Political Science: X ? J ~ 7 I PAUL PIERSON AND THEDA SKOCPOL - - - -

1 Historical institutionalism in I I Contemborary Political Science

Ira Katznelson and Helen V. Milner, editors

Columbia University

W. W. Norton & Company American Political Science Association

, NEW YORK ( LONDON WASHINGTON, D.C.

Like the character in Molicre's play who spoke prose all his life without knowing it, contemporary scientists are familiar wit11 leading examples of historical institutionalist rescarch without necessarily realizing that they exemplify a cohcrent gcnrc-mucll as tlo works in the other two major rcscarch apl~roaclies practiced in clnpirical political science, survey- based behavioralis~n ant1 rational-clioicc modeling. tiistorical institutionalists analyzc organizational and institutional configurations where others look at particular scttings in isolation; and they pay attention to critical junctures and long-tcnu processes whcre others look only at slices of time or short-term maneuvers. Researching important issues in this way, historical institutionalists make visible and understandable the overarching contexts and interacting processes that shape and reshape states, politics, and public policymaking.

Stephen Skowronek's The Politics Presidents Make (1993), for example, reveals recurrent cycles in the nature a11c1 success of presidential lead- ership throughout U.S. history. Anotlicr long-tern1 study in American politics, John Mark Hansen's Gaining Access: Congress and the Farm Lobby, 1919-1981 (1991), dcvelops a nlodel of interest group interaction with government and uses it to explain the emergence, persistence, and ul- timate eclipse of the political influence of national farmers' associations. Ranging across nations as well as time, Peter A. Hall's 1986 book Govem- ing the Economy: The Politics of State intervention in Britain and France explains how institutions and orga~lizations intersect to shape not just the policies of governments but also the strategies and alliances of interest groups and public intellectuals. Painting on even grander canvases, Shap- ing the Political Arena: Critical lunctures, the Labor Movement, and Regime Dynamics in Latin America by Ruth Berins Collier and David Col- lier (1991) and Birth of the Leviathan: Building States and Regimes in Early Modem Europe by Thomas Ertman (1997) explain regime dynamics and the varied formations of modern national states. In the study of international relations, Activists Beyond Borders by Margaret Keck and Kathryn

Sikkink (1998) analyzes the historical roots and contelnporary poliferation of t ra~ l~na t iona l ad\,ocacy networks, \vhile Judith Goldstein's Ideas, Inter- ests, and American ' k d e Policy (1993) shows how the institutionalization of policy ideas at an earlier j ~ ~ n c t ~ ~ r e had lasting conseqoences for subse- quent U.S. trade measures.

TlIese are but a few of Inany possible citations, for historical institutionalist studies have cumulated to provide wide-ranging as well as causally precise understandings of sue11 important matters as transitio~ls to democracy,' tlic elnergcnce and derl~ise of authoritarian reginles,? the iiitersec- tion of domestic and international politics,z the origins and development of lvelfare states,' social identities in politics,i the ~o l i t i c a l dvnamics of gender rigl~ts," the developme~lt of eeo~io~r l ic regimes,' and the causes and consequences of social illoveme~lts and revolutions."

Obviously, studies using historical-institutionalist strategies of a~lalysis vary in many iillporta~lt ways. Some are explicitly comparative, while 0th- ers analyze trends within just olie macrocontcxt. Soine offer s~iggestive interpretations (e.g., Hart 1994). while others develop explicit ~~ io t l e l s framed in general terms (e.g., Hansen 1991; Luong 2002). Sonic Iiistorical- illstitutionalist studies draw exte~lsively from primary sources (e.g., G a m ~ n 1999), while others synthesize findings from secondary publications (e.g., Skocpol 1979; Downing 1992). And sollle deploy a r g u n c ~ l t s about strate-

1. See, for ex:unplc, Anderso~~ 1999; Baloyra 1987; Brattoll and V:ul de MJ;~llc 1995: Diamond 1999b: Downing 1992; Gor~ld 1999: Haggard and Kaufinan 1995; R~escheme~er , Stephens, and Stephens 1992; and Yashar 1997. A comprehensive review of hypothesis testing and cun~~~la t ive theoretical developmel~t in this field appears in Maho~~ey 2002.

2. Examples include Chehahi and Linz 1998; Doyle 1986a; Ekiert 1996; Im 1987; Maho~~ey 2001; and S ~ ~ ? d e r 1998. For further review of the literature, see Mal~oney 2002.

3. Examples include Friedherg 2000, inclGourevitch 2 0 0 2 . B T / T 1 _ 0 1 T f r E T E M C B T / T . 4 9 3 0 9 . 6 - 5 9 ( t e s t i n g ) ] T J 0 . 0 1 3 9 2 d 3 o 9 T c 6 6 . 9 7 T j 0 T c 0 1 T 3 r 7 8 T c 0 . 0 5 3 - 1 . 5 0 9 T d ( 3 . ) T 9 0 0 9 . 6 2 5

2002.BTg.dr0Doy3 Tmsp0 9.625 66.J0.01392141.300.9993 Tm(2002.BT/T1_0 1 TfrETo6la915.320.7.625(testing )]TJ0.01392d3o9 Tc.12532 0 0 9.625 )Tj64 66.97Tj0 Tc ~~?d0 9.625 r:62103g5 206.039a7<</m9ng Tm0 9.625 Maho~Tj64 66.a2201 200d6.J-

ists, whose studies thus contribute empirical substance to debates raging among political theorists. What is illore, research ager~das sl~aped by historical institutionalists may draw considerable interest from formal theorists and behavioralists. 'I'he literature on the development of modern welf~l-e states exemplifies ~nultiple forins of bridging. Normative concerns about equalib, democracy, and liberty are frankly acknowledged and explored in this literature. And because research illvariably focuses on major, real- world developments and dilemmas-as historical institutionalists prefer- choice theorists and survey researchers have been drawn into academic dialogues remarkably free of sterile paradigm disputes. Theoretical and methodological dialogues have been higllly fruitfiil, and particular studies often combine inlodes of analysis (see the literature reviews in Anienta 2002; Pierson 2000d). With historical institutionalists playing a central role, the entire research co~nmunib has used a full rallge of methods and research designs to develop ever sharper and broader explanations-and richer normative understandings-of the origins and dynamics of national syste~ns of eco~lomic reg~lation and social provision. Everyone seems to re- alize that theoretical eclecticism, inl~ltiple analytic techniques, and a broad colliparative and historical purview works best.

I Tracing Historical Processes

Historical institutionalists may ask big questions and bridge divides within and beyond academia, but how do they go about developing explanations? Wc have already alluded to what is perhaps the most distinctive feature of this approach: U\;lether particl~lar ~vorks use comparisons or analyze various aspects of oilc theoretically justified case, historical institutio~ialists take history seriously, as something much more than a set of facts located ill the past. 'To ~mderstand an interesting outcorne or set of arrangements usually mealis to analyze processes over a substantial stretch of years, maybe eve11 illany decades or centuries. Scholars working in this tradition have developed c o ~ n ~ e l l i n g ~nethodological and theoretical iustifications for historically grounded i~~vestigatio~~s, by which they mcan investigations that look not just at the past but at processes over time.

Soine reasons for taking history seriously concern methodology and are alre;~dy recognized (if not always practiced) by political scientists of many perstiasions. Extending the tirne frame of social inquiry obviously widens the range of experience available for examination. This simultaneously makes more data available and generates greater variation in outcomes. Such widening of the empirical terrain is especially important for political scientists, because many matters of great interest-eapeciaily macrophenomena sucli as revolutions, state building, democratization, the construction of welfare states-occur relatively infrequently, or o111y par-

tially, within any particular slice of time. Historically informed iiivestiga- ti011 also sensitizes investigators to teinporal I~ouiidary conditio~is or pcriod effects, with respect to claims about causal relationships. By cxanlining a wider range of historical settings, an analyst can consider t l ~ e ~)ossibilit]\. that supposedly universal effects in fact only llold under particular circumstances.

Historical investigations can also contribute to causal inferciice. Be- cause thcorctically grounded assertio~~s of causal relationships imply temporal relationships among variables (either that one precedes the other or that both happen at esseiitially the same tinle), examining historical sequences is extremely useful for testing such assertions (Rucschemeyer and Stephens 1997; Mahoiley 2000b). Optinlally, assertio~~s of causality should be borne out ilot just by a correlatioil between tivo variables but also by a tl~eoretical account showing why this linkage should exist and by evidence suggestiiig support for the theorized linkage. Althougll social research does not always achieve this ideal, efforts to syste~natically trace social processes can make an essential contributio~~ to the rigorous assessment of claims about social causation (Hennett and George 1997; Hall 2002). Here the relatively small nuniber of cases in many llistorical i~~s t i t~~ t iona l studies call become an advantage, allowing for exactly the sorts of detailed exa~nina- tions of processes needed to evaluate clainis about causal inechanislns.

Important as these gencral methodological advantages are, historical institutionalists are also pushing beyond the111 to theorize about historicuf dimelzsions of cuz~satiolz. As we are about to see, historical institutionalist.,~Iists have recently begun to underline the theoretical power of their dvizamic approach to social change and politics. Without the kincl of attentiveness to temporally sl~ecified process that is a distinctive hallmark of historical- institutionalist scholarship, important outcomes may go unobserved, causal relationships may be misunderstood, and va l~~ah le hypotheses may never receive consideration.

A central example of why history may be causally critical involves clai~ns about path dependence that are common in historical institutionalist schol-

I arship (see, e.g., Collier and Collier 1991; Ertman 1997; Hacker 1998; Shefter 1977; Huber and Stephens 2001). Path dependence can be a fad- dish term, lacking clear meaning, but in most liistorical-institutionalist scholarship, it refers to the dynamics of sclf-reinforcing or posilive feedback

I processes in a political system (Pierson 2000a; cf. Mahoney 2000b). A clear logic is involved in such path-dcpcndent processes: Outcomes at a critical

! jurlcture trigger feedback mechanisms that reinforce the recurrence of a i articular pattern into thc fiiture. Such processes have very interesting : characteristics. They can be highly influenced by relatively modest pertur- ' ,.

bations at early stages. Once actors have ve~itured far down a particular

path, however, they are likely to find it very difficult to reverse course. Po- litical alternatives that were once quite plausible inay become irretrievably lost. Thus, events or processes occurrillg during and immediately following critical junctures emerge as crucial.

There are strong theoretical grounds for believing that self-reinforcing processes are prevalent in political life. Once established, patterns of political mobilization, the institutional rules of the game, and even citizens' basic ways of thinking about the political world will often generate self- reinforcing dynamics. In addition to drawing our attention toward critical junctures or formative moments, arguments about path dependence can thus help us to understand the powerful inertial "stickiness" that character- izes many aspects of political development. 'Fhese argunlents can also reinvigorate the analysis of power in social relations, by showing how inequalities of power, perhaps modest initially, are reinforced and can become deeply embedded in organizations, institutions, and dominant modes of political understanding. Path-dependence arguments also provide a useful and powerful corrective against tendencies to assume ftinc- tionalist explanations for important social and political outcoines. Perhaps most important, an appreciation of the prevalence of path dependence forces attentiveness to the temporal dimensions of political processes. It highlights the role of what Arthur Stinchcornbe (1968) has termed historical causation in which dynamics triggercd by an event or process at one point in time reproduce themselves, even in tlie absence of the recurrcnce of the original event or process.

A11 appreciation of increasing returns dynamics is one iinportant justification for the focus on issues of timing und sequeizcing that constitutes a second important theoretical rationale for examining historical processes. In path-dependent processes, the order of events inay make a fundamental difference. Historical institutionalists tracing broad patterns of political development across a number of countries often argue that the timing and sequence of particular events or processes can matter a great deal (Anderson

I 1986; Gerschenkron 1962; Kurt11 1979; Shefter 1077; Ertman 1997). I11 Ertman's analysis of Europeail regime formation, for example, it is the sequence of two processes-the expansion of literacy and the onset of mili-

I tary competition-that is crucial to paths of state building. The prevalence of literacy powerfully affected the bargaining position of those who would collect state revenues. Where cou~ltries faced intense military challenges prior to the era when literacy became widespread, the character of the state's fiscal apparatus was likely to be very different than when the order of these two processes was reversed. Crucially, because of path-dependent effects these differences were long lasting; that is to say, a country's bureau- cratic structures, established at an early juncture, did not simply adjust once literacy rates rose to a higher level.

Jacob Hacker's study (2002) of the development of U.S. social policy presents a secoild compellillg example of this kind of argument. Hacker

explores the distinctive paths of de\,elopment of p b l i c versus private, employer-based systems of health care and pension provision in the United States over the past century. Although both policy sectors have roughly equd mixes of public and privatc spending, they de\,eloped along funda- mentally different lines. In health care, public interventions have served as supplements or compensations added to a prior set of institutions featuring private provision. By contrast, in pensions, a public core system of retire- ment provision developed first, with private arrangements playing a sup- plementary or complementary role. 'l'hrough processes of policy feedback, these different sequences of public and private intervention generated quite different interest group environments, shifting botli the policy preferences and political resources of crucial actors like employers. Moreover, the differei~t pattcr~ls of developi~~ei~t produced very different distributional outcomes, profoiindly shaping the contours of contemporary political struggles over social provision. Any scholar who merely discusses these contemporary struggles without awarcness of thc history that shaped the terrain of preferences and actors will miss much of central causal relevance to explaining politics ant1 policymakiiig today.

Waldner's stiidy (1990) of state building and economic in late-developing cotintries rests on a similar claim. He contrasts Syria and Turkey, wlierc statc building and mass incorporation occurred simultaneously, with Korea and 'liiwan, where greater elite consensus allowed state building to prccedc mass incorporation. In tlic lattcr cases, state elites were able to institutionalize structures that facilitated long-terin economic de- velopmc~it before the cl~allenges of bargaining with nonelites had to be confronted. In the foriner cascs, patterns of state building were fundamen- tally altercd b y tlie immediate need to win support from broader social groups. 'I'lius Waldner identifies two distinctive paths of late state development, generated by different sequences of state building and incorporation.

Like these arguments of Ertman, Hacker, and Waldner, most proposi- tions historical iilstitutioilalists offer about the causal i~lipact of sequences are grounded in claims about self-reinforcing or positive feedback processes (Pierson 2 0 0 0 ~ ) . Relative timing, or sequence, matters because self-reinforcing processes, playing out over time in political and social life, transform the consequences of later developments. Path-dependent arguments about self-reinforcement explain why and when sequencing can matter. Increasing returns processes occurring during particular periods generate irreversibilities, essentially removing certain options from the sub- sequent menu of political possibilities.

Although many path-dependent arguments stress the institutionalization of particular configurations that subsequently prove difficult to dis- lodge, the focus on sequencing illuminates how arguments about path-dependent processes can be incorporated into explanatio~ls of politi-

,+ cal change as well as political inertia. For instance, path-dependent processes may operate to institutionalize specific political arrangements

PIERSON A N D SKOCPOL HISTORICAI. INSI'ITCJ'I'IONAI.ISM ( 703

that ultimately prove vu1neral)le to somc disl~lacing event or process einerging at a later stage in political devclopmeilt (Collier and Collicr 1991; 1,uebbert 1991). The Colliers' work on labor incorporation in Latin America provides an excellent example. In some of their cases, labor was excluded froin political participation at a critical juncttire, and positive feedback lcd to the institutionalizatio~~ of regimes wherc organized labor had little access to political influence. Over time, however, key changes in social and ecoi~omic life made it impossil~le to maintain these systeins of labor exclusion. Eventually, orgaiiized labor had to be iilcorporated. When this took place at a later stage of political development, llo\vever, it necessarily occurred 011 quite different tcrms and with different consequences thaii was true for countries cxperieucillg early incorporation. 'The key point is that options available for earl). i~~corporators were not available for late incorporators.

It is highly instructive to contrast these arguments about path- dependent sequences with arguments ratioiial-choice theorists have inadc about sequences withill highly instit~itionalized settings (Shepsle 1986). U'orking froin Arrow's impossibility tlleorel~l, w l ~ i c l ~ suggests the prospect of endless cycling in many collective-choice situatioiis, rational-choicc tlic- orists have argued persuasively that iiistit~itioiial arrailgements governii~g ageiida control and decision-~naking I~roced~ires can produce stable outcomes. By dcinonstrating the crucial rolc of scqucncing as choiccs arc made, these arguments rest on the cquivalei~t of a patl~-del~eiidcnt meclla- nism: Steps in a sequence arc irreversible because institutional rules causc forgone alternatives in early rounds to bc dropped from the range of possible later options (as, for cxan~ple, in con~nlittee voting rules that requirc a sequence of binary choices, with losers eliminated). By showing how such irreversibilities can be generated in a wide variety of social contexts, however, it is possible to expand this crucial finding to a far broader range of social phenomena than those covered by the rational-choice literature sterllmiilg froin Arrow's work. In comparative historical analyses, analogous arguments are often applied to large-scale social changes stich as deinocra- tization (Collier and Collier 1991; Collier 1999), industrialization (Ger- schenkron 1962; Kurth 1979), state building (Ertman 1997; Shefter 1977), or welfare state developnleilt (Huber and Stephens 2001). Sequencing matters not only for collective choices within legislatures but also potentially for any sociopolitical process ~vhere self-reinforcelllent means that forsaken alternatives become increasingly unreachable with the passage of time.

Historical institutionalists also employ timing and sequence argu- inents to highlight conjunctures-interaction effects between distinct causal sequences that become joined at articular points in time (Arnin- zade 1992; Orre11 and Skowrollek 1994; Sewell 1996). 'The ability to identify and explore such colljunctures is a major advantage of the more macroscopic inclinations of historical institutionalism. For example,

Skowronek's account (1993) of U.S. presidential leadersl~ip emphasizes the interaction of slowly developing modern institutional capacities and the particular position of an individual president within the sequential rise and decline of a dominant political coalition. 'The explanatory power of Skowronek's analysis therefore derives from l~igl~ligl~ting the intersection of long-term develol~ments with recurrent political cycles. Other historical institutional accou~lts havc focused on the political effects produced when separately unfolcling processes conjoin. 'Thiis Ertrnan (1997) stresses links between military coinpetition and developing social capacities for bureau- cratic governance, Shefter (1977) analyzes the interaction of state building and party formation, 2nd Skocpol (1979) attributes social revolutionary outbreaks to conj~irlctures of donlestic and interilational conflicts. These analysts focus on distinct sociopolitical processes that becoine linked in different and causally crucial ara),s depending on relative timing. 'The causal centrality of such coilj~li~ctures, it should be stressed, would never be noted by analyses investigating or t l~eor iz i i l~ about single processes in isolation.

Another theoretical justification for foci~sing on historical process is to draw attention to lengthy, large-scale, but often very slow moving social processes (Pierson 2002). Historical institutionalists seek to be attentive to the unfolding of both causal processes and important ~oli t ical outcomes over extended periods of time. Most political scientists are strongly predis- posed to focus on aspects of causal processes and outcomes that unfold very rapidly. Yet many tliings in the social world take a long time to happen. Seine causal processes and outcon~es occur slowly because they are incremental; it simply takes a long time for thein to add up to anything. Changes in pensioll systeins, for exai~nple, are not fully translated into lev- els of public spending for a half-century or more. A second possibility is the presence of threshold effects; marly social processes inay have little signifi- cance until they attain a critical mass, which may then trigger major change (McAdam 1982; Goldstone 1991; Baulngartner and Jones 1993). Alternatively, slow-moving ~~rocesses may involve transformations that are probabilistic during any particular period and therefore several periods may be necessary before the transformation occurs. Under such circumstances, the social outcome of interest rnay not take place until well after the appearance of key causal factors. Particularlj. when it foc~ises on macroscopic processes, historical-institutionalist research is often primarily interested in such structural preconditions for particular outconles, rather than the specific timing of those outcornes (Collier 1999; Moore 1966; Rueschemeyer, Stephens, and Stephens 1992). When either structural causes or threshold effects are at work, analj,sts adopting a short time frame are likely to focus erroneously on the more idiosyncratic or precipitating

u'

PIERSON A N D SKOCPOL 1 HISTORICAI . ~ N S ~ . I ~ I . ~ ~ T I O N ~ I . I S M 705

factors that trigger outcomes. Because some crucial social conditions may change only slou.ly, analysts studying a narrow time frame will be strongly inclined to take thern as fixed and therefore irrelevant to their causal accounts (Rueschemeyer, Stephens, and Stephens 1992; Kitschelt 199 1) .

Another possibility is that causal processes iilvolve chains wit11 several links, which require some time to work theinselves out. To the extcnt that causal chains of this sort are at work, analyses must frame their studies on a broad time scale. Collier and Collier's influential work on labor incorporation in Latin America presents arguments of this kind, in which the ulti- mate outconles of interest reflect a sequence of key developments over extended periods of time (Collier and Collier 199 1). Indced, this type of claim about long-term, multistage causal processes is often invoked in work 011 state building (Flora 1999a, I,) or de~nocratization (Luebbert 1991; Collier 1999).

Swank (2001a) offers an instructive recent example. In assessillg the impact of political institutions on welfare state retrenchment, he criticizes the view that frag~nented institutions \vill limit cutbacks by increasing thc number of veto points available to defenders of the status quo. Swank ar- gues that this is t r ~ ~ e as far as it goes hilt notes that the long-term, indirect effects of institutional fragmentatioil run largely in the other direction. Not only does institutional fragmentation limit the initial expansion of tllc wcl- fare state, I>ut it also reinforces social heterogeneity, inhibits the growtl~ of e~lcompassing interest grol~ps, and weakens ciiltural commitments to universalist social policies. All of these long-terin effects strengthen the \vclfare state's oppo~le~lts and weaken its advocates. Thus nlany of the most iinportant effects of institutional fragmentation work themselves out only in- directly and over extended periods of time. Ahistorical analyses t)picall\. seek to consider the effects of institutions while holding constant other variables, but these variables are in part the long-term consequences of institutional structures. Ahistorical investigations are therefore likely to sys- tematically rnisread the impact of institutional structures on the politics of the m:elfare state.

Analysts who fail to be attentive to these slow-moving dimensions of social life may ignore potentially powerfill hypotheses. They are particu- larly likely to miss the role of many sociological variables, like demography (Goldstone 1991), literacy (Ertman 19971, or tecl~nology (Kurt11 1979), as

I i u.ell as the impact of other slowly building pressures such as international

military competition and fiscal overload (Skocpol 19791. Their explanations may focus on triggering or precipitating factors rather than deeper

1 causes (Kitschelt 1991). Perhaps most fundamental of all, they may fail to I

ei:en identify some important questions about politics because the relevant outcomes happen too slowly and are therefore simply off their radar

I screens.

I11 the ways we have just surve!.ed, tl~eoretical attentiveness to historical processes represents a formidable conlparative advantage of historical institutionalism, especially sincc this attentiveness is linked to macroscopic analysis focusing oil institutions and organizations in addition to aggregates of people. Most research in the behavioral tradition uses surveys that offer a snapshot in time. And e\:en when surveys are repeated to offer a longitudinal series, it is rare indeed for behavioral analysts to consider changing institutional contexts, critical conjunctures, or path-dependent large-scale processes as causally rele\ant to the clianging inodes of individual behavior theq. prohc. l h e r e are exceptions, certainly, such as Carmines and Stiin- son's dissection of the processes by which racc became a transfornliilg issue in U.S. partisan politics, or Piitnam's consideration of the iillpact of \2'orld War I1 and other watershed historical events on "life course" develop~nents for latc-twentieth-cei~ttir~ U.S. adults (Carmines and Stimson 1989; Put- nam 21)00). Rut for the most part, s(ir\,cy a~ialysis relies on o~ic-ti111c data about attitudes and self-reported bcliaviors to explorc i~itlividual-level hypotheses about mass patterns. Change over timc oftell e ~ ~ t e r s the discussio~i only speculatively, as wl~en a researcher rcports differences behvccil Lvork- . .

ing womcn and Ilousewives at a monlcnt in time, and thcil go on to spec- ulate that the changing proportions of such persolls i l l the population ma), signify a continuous social change.

To I>e sure, some rational-choice scholars have turilcd to Ilistorical case studies in recent years. Rut most of the analytic advantages we have outlined are al>sent in this work, because the past enters only in a highly restricted scnse, as what might be termed illustrative history, the mining of the historical record for outcomes that can be "explaine~" by prticular rational-choice models. Previotisly established models may be applied in interesting ways to examples in the past (cf. Bates et al. 1998), but the tools of game theory turn out to be poorly suited to analyzing conjunctures or exploring slow-moving macroprocesses. Rational-choice theory faces considerable difficulties in rnoviilg froin the micro- to the meso- or macrolevels of analysis that are typically featured in works that analyze processes over long stretches of time (Elster 2000; Munck 2001b). Rational choice's reliance on theory-driven agendas and on the identification of empirical terrain favorable for its preferred methods leads relentlessly back to an emphasis on the micro. And the results of game theory quickly becorrle indeterminate or unmanageably coinplex as one increases the number of actors involved (indeed, in game theory the problem of indeterminacy is often rife even at the microlevel). The fact that man! macroprocesses take considerable time to play out presents a further difficulb, since game theory generally requires that all the relevant actors, preferences, and payoffs be established and fixed simultaneously at the beginning of a game. 111

jo sa!la!lr:.i lo sartr!Zal aleis a l t l f ~ n . ~ at\!inrr!is!p a je~aua2 sa~njsnl+s [euo!~ ' ( ~ b b l . ~ . \ J E H

-E:Z!LIE~IO prre 'srro!inl!isrr! le ru~oj 'sn!n!~od 30 s u o ! ~ r ~ ~ ~ ~ S ~ u o ~ ~ o q uo pasnn 3,661 l p ! ~ p ~ ~ aas 'srro!ldaoxa lied ioj) sls!ioa~~l ~~!OLIS-IEIIO!~EI 'iq palous! .i1351~l uaaq seq ssaooicI les!iols!q LIO snooj 2uoils s!q 'aslraplradap tiled 110 (0b61) Y J O - ~ ~ -01 "?L[ ( [ ( ) 0 ~ suarldais ptre laqnH f0661 L I ~ S J ~ ~ L I V - 2 ~ 1 ! d q ) Xs!lod le!sos 4uEl~od~r! S ~ ~ J J O N sel211oa al!dsap 'asuelsu! iod ,uo!lssi!p s!ql u! sa.zoul pa$!~u!l

I'LI" ([C)C)z a 3 ! ~ w s [[eH f(,661 suarldais pue 'syiem 'a2ne7 'l[aqnsi!y) apeur seq $1 'padcI!nba iallaq suraas Xioaql ao!otIo-1etro!lei alaq.r\ uaj\2 '01 i u ~ o u o s s leo!l!locI a.i!icled~troo rpoq u! y~o.zz iuanai 'Xlle~!ur!s I

'LIMO SI! uo annpoid 04 ,ilay!~ uaaq a.\eq plno'zz ssal81rog ir ! lQ~~o!eru e JeilM prroXaq sepua8e l!alp anne'ipe 01 alqe traacl 'sni[nrrc1q o , ~ ar[l lraa'qaq Xeldla+rr! y8noiq+ 'a.ieq saaU!rrruroo 1cuo!ssa18rroo ptre sllnos [elapaj aqi u! rrrs!n!loe Xn!lod l o sa ieno~py . S ~ ! S l e t ~ o ~ j t i ~ ~ ~ s i ~ ! +JLI!IS![I asarl+ 2u!ldnon u! aIol le!nnin e 8u1 -Xelcl snogez!ue2lo 'straz!j!n jua81nttra .i[.\\arr 1[pM 'ssal81rog pue silnon ayi rraa"rz1aq rro!lnelaiu! rre 30 jlnsal arIl se poolslaprrn aq isnur I! ieql saleqs -trottrap 311 'aur!~ 30 rlojaljs papunjxa ue la.\o prre sanrrat\ n[d!i[nru y8noiyi rro!inlot\al s!~p 2rr!nel~ .altltnexa poo2 e sap!.\old saleis pai!un ayi u! uo!s -1retlxa .iloieln2al pala113 set1 jeqj .,rro!jnloAsl slrl2!1,, arli l o (b661) s!d[eue S,Y3!Ulam '(EZ(,(,~ '!nit\ :Zh(j1 10d3OJs '2661 ~3\10[ P\1E l ~ l ~ ~ l ~ % l ~ l I ~ ~ )

s8rr!fias [elro!irq!istl! alcl!+[urrr rr! sloine [e1ro!~ez!rre~10 a[cl!+[nrrr 8uoure 1 I Xelcllaitr! arp uo sas~roq .illeo!dXi 'aorreistr! 103 'irratrrdo[a.zap Xn~[od . s . n uo

1In.masaH .tro!ielos! rr! rro!irq!isrr! ierp l o s!s,\'lerre rre rro .ilnlos jsai lerp spa j -In [etro!$rri!lsi1! 1110([t: SIII!I?[D ,i111?111 f o l2rMOd 3111 iq110[1 04 , in~rapt~aj ~ U O ~ + S e s! ala[[,I, 'o[) uo+jo s~l?[ot[nmo!ot~n 11:110!~c.1 ST 'm)!i~?imiuoo 30 341s [euo!+ez!u -1?21o .to [er1o!ill+!lsll! slYtr!s I? 110 snooj .$nler ys!letIo!ilri!islr! ~ t : o ! ~ o j s ! ~

.ysedrtr! ~I!sIII::.' ~ !n[ [ l J.>I?.I~ ~ I I P DICI!S!.Z s11.1n~r:cI nsotli nJerrr oi ttl!e s+s!leuo!+uj -!jSll! [l?,J!lOjSll\ 0s .(()-s ' ~ ( ) ( ) z ) LL.IO!~\I?t13C[ [t!!30S ~ [ ~ L C I ! S 30 110!jl?~31~%2 -3qi 111o.q ~ J A ~ I I I X I J I ~ .XI? sassnoo~d ~,\!jno[[on DSDI[$ [I! ~D,ZIO,ZII ! s ~ ~ P J ~ I ! [ [ C S I I E ~

.JIP . ' ' j11q 'olclootl jo .~o!.\et~nq 3 1 ~ r?!.i ~ > ~ ~ I I ~ X ) I C I nsl11oo jo 311: . . . srro!l -11j!js11! rlnn.\\jncl pt~l: ~~!rll!.\\,, Yrr!~~nooo snssnoo~cl nsoq1 *I! sirltl rrosladda[ 1>[1?11oy qr ~ S I I O ! ~ I I ~ ! ~ S I I ! nlc'I!i~~~tr~ 30 l.>l?tl[[r! i[r!o!r~o;, arll 1l21loitp pnssaiclxa olr? sionjjo nsnr[i jo i t r r ? r t ~ ierli ~ I I I : S ~ ~ ~ J J J YI I !~ ! I I I I [ J~ [IIII? ~ ~ I I ~ ~ I I ~ I C ~ D Inya -mot1 o,ietl sn.\losrtratli ptry slr:~rp!.z!ptr! I(I)!I[M (11 scI!r[s~ro~~r:[n~ 11111: sanlnosal l o srr~oiiecl nip ietp n.in!lacl oslt? . iaq~ 111k1 .( 1661 J ~ I ~ ! - I ) . K ) I . \ c ~ ~ x ~ 1 x 1 ~ UO!+EA

-!$our lelIp!,i!ptr! 30 ~ I I I I O ~ X 31c1!~1n?1t\ 1 1 4 1 . ~ + ~ m p s u o o ac) .~[~+I:NI!$~II pluorls sasrunn leqi alcl!nrr!~tl nrli itlao.>r: s~s!lt:tro!jr~j!jstr~ les!lo+s!r[ *.ilrr!c+~a:) .lnt\aI lntro!irli!istr!.roilI! lo ~ I ? I I ~ ! ~ ~ Z ! I L I ? ~ I ~ . I J ~ ~ I ! 3114 $1: 211!jt:~ac[o sasrrt:o nloldxa [XII? ~1[2![t[2!11 01 PtIai S ~ S . ~ [ C U I ? 'nSl?3 l311j!3 111 ' (z~j( j1 J~LIOIMOYS f ~ 6 6 1 lIl?lU

-$I:;[ izh(j[ ~II!U.\ \OCI f s ~ 6 1 .i[l!,l, fI?'-)~bl ~ I / < o c ~ :5)861 lIOSl"L"1V "8.2) LIO!$

-etruoj aitqs .lo ((,#(,I . \ \ ou t l i " i~ ) STIO!~I?III.IOJXII~.I~ [crl+nq1ai11! 8rr!z!1uapo1n se r ~ o r r ~ s ~ r ~ ~ r ~ r d o ~ a . i o ~ ~ la.in10.131?m1 isotrr arp a[joe+ osle derrr s+s!lctro!l -nl!istr! Ieo!lois!H . ( ( ) u ~ z U O S U I ~ I \ T put 'zrret) '1ot1no~~ '.$!.a) s[>[ny [~110!4

- I ? Y ! L I ~ ~ J O (I! sa2irer1o l o ('1,7661 1!n1fi fzh(j1 ~ I I ~ I ~ L U L U I f8661 . ~ a y n e ~ "2.a) I I

t?a.re ens$! .telnx+~t?cl I? (11 sirra~rrtlola.iap /;n!locl rro 'alclrrrexn 103 '2n!ie~~tra.7 I i -troo 's11.303 l,?.\a[o~1111 e slclo[>e urs!lerro!iui!isrr! [eo!lo$s!q tr! rpieasal rlnnly

qsalain! jo sarrron+no n q i o r~!e[dxa ol srro!lein8yuoo lerro!inl -gsu! 04 2u!11ajai sa1qqlm asn 04 sarrr!l iaqlo +e 'sa.qasr~rar[+ sirrarua2rre11e leuo!~n~!~su! aqi u!e1clxa ol s! 1eoS led!nrr!id ar[+ sarrr!+auroS .suo!leln8yuon leuo!in$!isu! lapeolq se [law se S L I O ! ~ I ~ ~ ! ~ S L I I qy!oadqo .G![!qelsrr! l o X$![!qe$s pue 'pedru! 'srr!8!io aqi 2u!qold *aru!j ia.zo srro!lrq!lsu! jo au![nap prre as!l ayi ale8gsat\rr! s+s![euoqni!+su! len!iojs!H .sa2ueqs la32!11 o q ~ slolne 103 s8u!rrado papuaiu!un 2u!~eai3 rrayo llayloLre auo rp!.u inasiaiu! sassnnoid prre surleal Ieuo!lni!isu! ald!ilriur r p ~ ~ p u! s X e ~ ol uo!+uaGe asoln Xed .iayjJ .sixailton lapeolq 'Xq panuanHLr! ptle ' L I I 2rl!jc~ado '~ro!lnelalu! rr! suogn+gsu! a~cl!~[rarr au!rrrexa j u ~ p sns/ilcue 1at\a1oloerrr lo -osnttr op .411en!db 'isellrron Aq 'qs!letro!+rl~!lsu! Ien!iois!H .alrr!i e ie s n ~ n l 30 ins airo uo snnoj ( r a p sielorlos an!ot1n-11:~10!p8 .sloioe lenbntru 2 t ro~~re al22ruls 30 qnnpoid 8rr!dolat\ap arll SF SLIO!~II~!~S~I! 33s ~ I I I : ' S D D ~ I I O S ~ I prre ~a .uod 30 sanlreieq Xsearrr~ ac[oltl 'n[!r[.fi~ren~r ' s ~ s I ~ I ' : u o ! ~ I ~ ~ ! ~ s L ~ ~I!:"IIO~SIH ' S I ? I ~ I ~ I I ~ ~ ! ~ I troqne at\!l.~[[o;, ol "s[ro!l11[os,, L ~ I I I ! I c ~ ! ~ ! I I ~ > ~ ap!t\o~cl ierp nttre2 nip 30 s a ~ n l LIO snooj 04 pr1ai sle[o[[;,s ao!oqo-Ietlo!iey .rr!etrlnl snnrIainjj![, juej~ocl -ti[! '( I X - Z ; ~ '6661 uo[ar[,l,) s2u!puy pue stro!isonl, ,i1e~~rn~tra1tl~troo attros 110 pa21a~rloo a;zt:r[ sacI!lls lrral,~jj!11 30 ys![erro!qlq!istl! t12rtoql trat\a $ 1 1 ~

,onlra!.?s [e;,!l!lod X~eiod~trn+~roo rr! ~~oeoiclcle paleqs .illx?o~(l e panlxr! y ~ U S I ~ I " I O I ~ I ~ ~ ! + S L I I 'ast~ns s!ql LII ~yrrr:cl!o!~~ecI [eo!i!lod pno1:j sirr!eijsuon pue 's~ro~jtlo 'sat\!lrla.x~! nrli 8~r!lninnlis ptre saxuosni 8rr!ieoo[ -1e u! a101 I E ! ~ I I J ~ I? Xeld sntryr~ol pa![,oc[ura , i ~ ~ ~ u o ~ ~ ~ z ! ~ 1 ~ 2 i o l1?114 < J L L I I ~ S ~ N ~

~ ~ S ! ~ E U O ! ~ I ~ ~ ! ~ S I I I ~KI!JO~S!LI pue S ~ S ! I B L I O ! ~ I I ~ ~ ~ S L I ! ao!otlo-11?rro!l1?1 [[lot1 'paieio -tiela .ilqe set1 (6661) uaIariLI, s y .(9661 .to[LeLl, ~ U C [[cH) s luxua: \o~1 pas 01 8l1!yeu1 rro!s!nap a.\!iels!2a1 rrroq 8t1!2tre1 s.xsnoo.td adells prre lo!ner[ -aq 1e3!l!10d aouanHu! suo!inl!isu! . ~ o q az,ilt?tre sie[or[ns .(rrt?rrr .iepoi aoua -!as 1en!l!1od 111 . M O ~ I S~S![I?LIO!~I~!~SII! [[t? ~ J C alp\ :urs!rre!satr.iay jo p!es anuo i r o x ! ~ p l e y n ! ~ i e r p ,iepo+ antIa!ns [en!i![od qonm 30 lies rrm ~ S - I E I O ~ ~ S

jo XI!UIPJ s!L[+ 103 rrearrI u s ~ ~ u n o ~ ~ i ~ ~ ~ ~ s u ~ s o p ~ e q . u 4 1 1 ~ ~ u ' y arp se [ [ a ~ st! IaqeI si! l o +led prrooas aqi .ic1 paz!~alneietln s! ~t1s!le~ro!+rl+!+srr! [es!lois!FI

o r . ~ n ~ ~ n m ~ d [1?3!10iS!l[ jo sinadse .iuettr jo uoge -lap!suon sno!ias 01 ,iloarp an!oqn-[e~ro!+e~ II! sn[oeisqo leal ale 3ln1p 'ilotls

capitalism that operate in fu~ida~nentally different manners. Outcomes are generated not by somc universal operating principles characteristic of a gi\:c~i b p e of actor or realm of activity but by intersectio~is of organized practiccs (such as labor markets, firm structures, or crucial policy arrangements). 'l'hese practices will often have originated at different times and thcn for~lied configurations that advantage key actors (such as employer associations and labor unions). These actors, ill turn, will generally work to maintain the configuration with 0 1 1 1 ~ incremental adjustliients even when economic, c~iltural, and geopolitical circumstances shift.

I s s~~es of long-term institutional developr~ient are also ceiitral to historical i~istit~itional research agendas (l'helen 1999). Historical institutionalists are typically suspicious of fi~nctional explanations, in which institutional outco~ncs are explained by their consequences. In such functional ac- co~~i l t s , i~~~stitutions de\:elop because of their capacib to solve ccrtain collective problems. 'l'he implicit or explicit claiill is that rational actors produced these outcomes in order to solve these problems.

As suggested above, concern with issues of instit~itional develop~nent within historical institutionalis111 is strongly lii~ked to theorizing about the caiisal relevance or origins, sequences, aiid temporal proccsses. F~~nct ional accounts of institutions seen1 inost plausible when i~ivestigations take a - one-tiinc snapshot, because long-term effects and inconsistent "layers" of inst i t~~tio~lal development (see Schickler 2001) are not so readily noticed i l l such studies. I11 most cases, synchronic analysts simply probe an extant institution to uncover benefits for part ic~~lar actors. Analysts tlien infer that these benefits explain the institution, implying that the actors who are presentl\ advantaged (or their forebears) created the institution to produce the benefits. This is a plausible hypothesis, but it is only a hypothesis and over-time i~ivestigation of institutional origins and dynamics often does not bear out such lines of reasoning.

By examining issues of institutional origins and change over an extended time frame, historical institutionalists have been able to highlight a number of potential problems for f~inctional accounts (see Thelen 1993, 1994, 1999; Pierson 2OOOb for relevant findings and arguments). Func- tional interpretations of politics are often suspect because of the sizable temporal gap between actors' actions and the long-term consequences of those actions. Political actors, facing the pressures of the immediate or skeptical about their capacity to engineer long-term effects, may pay lim- ited attention to the long run. Thus the long-term effects of institutional choices, which are frequently the most profound and interesting ones, should often be seen as the by-products of social processes rather than the realization ("congealed preferences" as Riker put it) of actors' goals.

A second issue concerns unintended consequences. Even where

actors may be greatly concer~led about the future in their efforts to design institutions, they operate in settings of tremendous complexity. 4 s a conse- quence, they will often make mistakes. Thus institutions ma! not be functional even in a context of far-sighted actors, because they do not operate as intended. Although \videly acknowledged to be significant in actual politics, political scientists often treat unintended consequences as an error term or simply ignore them by failing to investigate institutions over time. In cross-sectional studies of institutions, the issue of unintended consequences vanishes from view, since either the long-term consequences of institutional choices or the original factors generating the institutional choice will be outside the scope of the analysis. By contrast, historical institutionalists examining institutional development often stress the surprising long- term consequences of earlier political choices and conflicts (Anderson 1986; Luebbert 1991; Skocpol 1992; 'l'helen 1999, 2002).

Historical institutionalists, finally, emphasize the ways in which institutions are remade over time ('l'lielen 1999, 2002). Because of strong path- dependent effects, institutions are not easily scrapped when conditions change. Instead, institutions will often have a highly "layeredn quality (Schickler 2001; Stark and Bruszt 1998). New initiatives are introduced to address contemporary demands, but they add to rather than replace preex- isting institutional forms. Alter~iatively, old instit~itions may persist but be turned to different uses by ~ie\vly ascendant groups. In either case, the original choices are likely to figure heavily in the current functioning of the institution. l h u s institutions will rarely look like optimal solutions to present collective action problems.

Clearly, attentiveness to history and the investigation of meso- or macrolevel institutional corifigurations are highly complementary strategies of analysis. 'I'racing politics through time is very helpful for identifying the boundary conclitions for particular theoretical claims. Even more significant, the emphasis historical institutionalists place on conjunctures and sequencing draws attention to the temporal connections among social processes, and highlights the importance of meso- or macrolevel analysis of institutional configurations. Furthermore, while path-dependent or increasing returns processes can play out at a microlevel (e.g., in the way individuals de\:elop and reinforce particular mental maps of the social world), they are often most significant at the meso- or macrolevel. Particu- lar sets of institutions and organizations are often mutually reinforcing or complementary-the presence of each enhances tendencies for the development of others-in a manner that can be seen as a kind of coevolution- ary or selection process playing out over considerable stretches of time. A good example is the recent research in comparative political econoiny focusing on varieties of capitalism (Hall and Soskice 2001b; Estevez-Abe, Iversen, and Soskice 1999; Soskice 1999). This work stresses how differellt political econolnies have developed along quite different lines because of the coevolution of nlutuall!- reinforcing institutional and organizational

<,

1.

PIERSON A N D SKOCPOL ~ ~ I S T O R I C ~ ~ I . INSTITU'TIONALISM 1 711

structures. Patterns of fir111 organization, systems of interest intermediation and key social and econoinic policy structr~res contain institutional corn- plementarities that place national political economies on distinct, path- dependent courses of development.

One can also see the advantages of a combii~ed focus on institutions a ~ i d temporal processes in the development of extensive historical institutionalist work on policy feedbacks (Heclo 1974; Weir and Skocpol 1985; Pierson 1993). Marly of the works cited here emphasize that government policies establish some of the most important structuring rules of modern societies. 'The specific design of policies can have an enormous effect on the resources and strategies subsequently available to political actors- indeed, in inany contexts, policies can be as important as formal political institutions in shaping political processes and outcomes (Esping-Andersen 1990; Skocpol 1992; Weir 1992a; Pierson 1994; Huber and Stephens 2001; Carnl>bell forthcomirlg; Hacker 2002). 'l'hc success of this line of argument de~no~lstrates very clearly how the particular vantage points encour- aged by different analytical strategies can either highlight or obscure important aspects of social reality. Because rational-choice scholarship is so focused on the pursuit of broad generalizations, it has concentrated almost elltirely on recurrent, carefully delimited formal institutions such as legislatures or central banks.

Only a few rational-choice scholars (such as Bates 1988b) focus on the state overall and highlight the ulterior effects of state policies. In forinal in- stitutioilal accot~nts, policies are usually treated as dependent variables, with their further political consequences through feedbacks falling beyond the scope of analysis, either because these consequences are seen as too idiosyncratic or because analysis of policy feedbacks ~vould require attentiveness to processes unfolding over time. But this omission will not do, because the expanding activities of states-and the policy feedbacks flowing from them-have been a central fact of modern life. Fortunately, state expansion and policy feedbacks have been at the center of historical- institutionalist analyses, even as these phenomena have largely vanished off the radar screen of the more universalist studies of formal institutions. The distinctive foci and contributions of rational-choice and historical- institutionalist work on institutions tliils provide a clear demonstration of the advantages of pluralist political science, employing multiple strategies of inquiry. Without coinpetitive pl~ralism in social science theorizing and research, important issues can too readily fall off the agenda of scholarship.

The strong emphasis on interaction effects in historical institutionalism re- flects some core working assumptions about how most sociopolitical processes operate. Analysts are strongly affected by their basic presump-

tions about how social processes work; and they tend to frame their problems, generate hypotheses, and employ nlethods of analysis alid rescarch designs that fit such presumptions (ilbbott 1988; Hall 2002). Behavioral- ists, for example, are happy to use statistical techniques to analjzc data from as many "cases" as possible-often data from surveys of tliousi~nds of individuals-because they are prepared to assume that very g e ~ ~ e r a l variables operating independently of one another come together to accotult for the patterns of behavior they are trying to explain. Historical ii~stitutionalists, by contrast, assume that operative variables may not be indepe~~dent of each other at all. When it comes to analyzii~g the origins and iinpact of institutions, causally i~llportant variables are often bundled together in the real world; and there may be alternative causal patl~s to siinilar outcomes (Ragin 1987; Shalev 1998). Historical institutionalists tend to suspect from the get-go that causal variables of interest will be strongly influenced by overarching cultural, institutional, or epochal contcxts. This is not a matter of getting mired in thick descril>tion. As Andrew Abbott puts it: "Context has t\?.o senses. . . . 'I'hc strict sense . . . denotes those things that enviroil and thereby define a thing of interest. l'hc loose s e ~ ~ s e siinply denotes detail. The acute reader will notc that these correspond nicely to tlie two judg~neiits of the scientific worth of contextual information. If decontextu- alization is merely thc reinoval of excess detail, then it's a fiiie thing, scie~l- tifically. If, on the other hand, it is the rc~iioval of defining locational information. it is a scientific disaster" (1997, 11. 10). Because historical institutionalists share thc latter set of expectations, researchers in this tradition tend to nlove LIP from single institutions to broader contexts-they look at forests as well as trees. Arid they almost always seek to discover and explicate the impact of configl~rations of organizations and institutions on outcomes of interest (Katznelson 1997).

In addition to presuming-and concentrating attelltion on-causal configurations conceptualized at the organizational and institutional level, there is another way in wl~ich historical institutionalists highlight interaction effects. They often do this by pointing to overarching contexts-types of regimes, eras, regions, and cultures-that place bounds around the theorizing being done in any given study. Historical institutionalists rarely aim to write about all of humankind through all of global history. Scan the ti- tles of major historical-institutionalist works, for example, and you will often find the proper names of regions of the world or the beginning and ending dates of the specific period that the argument is to cover. This is not done because historical institutionalists in political science are trying to be historians; they do not aspire to just tell stories about a time and place. Nor are dates given only to describe critical junctures when a change occurs that the analyst seeks to explairi (e.g., Clemens 1997; Skowronek 1982). Beyond such an obvious use of dates and place, historical institutionalists often set limits to the applicability of their causal arguments, arguing in

theoretically explicit ways why variables appear and combine in characteristic m,ays in one era but might not exist or combine in the same way in other eras.

For example, in his book about the (or not) and success or fail-

ure of guerrilla-led revolutionary moveinents, Timothy Wickhaln-Crowley (1992) develops a rigoro~is causal analysis and suggests [in his conclusion) that his model links up with theoretical explanations of re\olutions in many regions a i d eras. But Wickham-Crowley carefully boullds his own argument, restricting it to Latin An~erica since 1956. In analytical terms, he explains that specific d e ~ e l o ~ ~ n e n t s and events created background conditions to generate many guerrilla moveinents with similar aiins and inetll- ods, thus setting the stage for the variables he explores to account for movement growth and success or failure. In the end, Wickham-Crowley's careful delineatioil-and explanation-of the overarching context within which his variable-based analysis applies reinforces its theoretical power. As we move to other continents and periods, we call ask about changes that might influence tlie \;ariables in play and tlie likely relationships ainong theni.

Another way in which historical-institl~tiollalist works highlight overarching coiltexts is by deliberately juxtaposing two or more coritexts, to show how variable configuratiolls already analyzed (across n i r~ l t i~ le cases) niay plav out in new ways u~llen tlle overarching context changes. 111 the revolutions literature, Good\viu (2001 ) does this by, first, developing explanations for revolutionary successes and failures within Central America and East Asia, and then kiigl~li~l~ting the somewhat different actors and conditions that come into play within each region. Addressillg a very different problem-Pierson (1994) identifies variables about institutions and established policy characteristics that can explaiil why British and U.S. conservative politicians succeeded or failed with attempted cutbacks in a ilurnber of social policy areas. Yet he also steps back and stresses that the key causal relationships governing policy determination are quite different in the recent period of austerity than they \Yere when welfare states were greatly expanded.

Scholars in other major political science traditions are often much less attentive to overarchillg contexts than historical institutionalists, in part because they prefer to focus on individual-level behavior or inicroprocesses but also because they are reluctant to write like historians. Ironically, the result can be less theoretically work. Behaviorialists happily rely

on types of data only available at one point in time or just for short time spans b ~ i t do not necessarily think through what that means. As a result, significant overarching contexts can go unnoticed, ~inconce~t~la l ized. For example, behavioralists sometiines fail to notice that very similar individual-level patterns-joining illore or fewer voluntary organizations, voting at different rates-can end up having very different meanirlgs, depending on the kinds of organizations or institutions that predominate in a

given nation or era. 'I'heories invoked in behavioralist studies can elid up b e i ~ ~ g profollndly ~mderspecified.

Rational-clloice scholars, meanwhile, often write as if the models the!; present were infinitely generalizeable, even as they slnuggle all kinds of institutional, cultural, and epochal specifics into their enipirical operational- izations. This may seein an optimal way of arguing; isn't social science supposed to generalize? Biit, in practice, this kind of approach means that we leave importarit variables i~nplicit and we fail to see how c h a ~ i ~ i n g background conditions iriight cause the same variables to play out ill very different ways. Again, the preference of rational-choice theorists for exalii- ining microsettiilgs strongly reiiiforces this tendency.

1 Research Strategies and the Accumulation of Knowledge

Tackling big, real-world qnestioils; tracing processes thro~~gl i tiine; and all- alyzing iiistitt~tiolial configl~ratio~~s and coiitexts-these are tlie features that define llistorical institutio~ialism as a nlajor stratcgy of rescarch in contemporary political scieiice. We have stressed the coiilparative advalitages these approaclles afford to all who would bctter t~nderstaild goveriinlellt and politics. But wc must also ackilowledgc the claims of critics who dis- miss historical institutio~ialisin as a valid approach to doing cumulative social science. Critics of historical institutionalisln sometimes pen manifestos announcing that casc studies and small-n comparisons cannot generate valid knowledge, bccause cases are not randoinly selected and there may not be enough statistical degrees of freedom to test all conceivable hypotheses rigorously icf. Geddes 1990; Goldthorpe 1991; Lieberson 1991). Statistical methodologists (such as King, Keohane, and Verba 1994) also worry that the llistorical-ir~stitutio~lalist proclivity for tackling significant issues predisposes them toward "selection on the dependent variable," that is, choosing cases where a phellomenon of interest has occurred, while ig- noring the instances where it has ilot occurred.

Historical institutionalist books and articles are sometimes denounced in ways such as these one at a time, in isolation from one another-an approach that surely ~ ~ o u l d be laughed out of court if applied to isolated works in any other tradition. In all research traditions, individual works build on orie another, often extending lines of analysis, retesting arguments, and correcting for earlier limitations. The real issue is whether historical institutionalists in general have headed up blind alleys, because their studies, considered collectirely as well as individual13 are improperly designed. This is not the place for an exhaustive review of the profusion of recent n~ethodological reflections in this genre, but the major signposts on the road can be mentioned.

Methodological challenges have, ill our view, been good for historical institutionalism. Not only have these critiques been testimony to the visi- bility and intellectual impact of the studies they have dissected. Challenges have had a bracing impact, prompting historical institutionalists to spell out their metatheoretical presumptions and sharpen rationales and tools for doing valid macroscopic and historical studies. A number of scholars, for exarnple, have asked how studies can best be designed to take intuitions about configurational causation and temporal process seriously. In other words, if historical institutionalists doubt that single, highly general variables have uniforin effects across contexts, and regardless of iilteractions with other factors, how can research designs allow for adequate einpirical exploratioli of hypotheses about contexts and configurations (see Abbott 2001; Mahoney 1999; Hall 2002; Ragin 2000)? If historical institutionalists stress mechanisms rather than silnple associatioli in causal arguments, how can process tracing be done in a rigorous way? And \?,hat criteria nlust be met to demonstrate valid arguments about path depeildence, critical conjunctures, or sequence effccts? Although Picrson (2000a, 2002), 'l'helen (1999, 2000a), a i d others have primarily focused on the tlleoretical cliar- acteristics of argriments about temporal causality, their reflections also set empirical standards to be met in case aiialyses and comparative studies that aim to establish the presence of causally relevant processes or events.

Reflecting on the stre~lgtlis and limits of hundretls of studies in various important literatures, metl~odologically oriented historical institutionalists have also nude the case for circumstances under which in-depth case studies a i d small- to mediuln-n comparisons are an optimal research strategy (Collier 199 3; Mahoney 20002; Munck 1998; Ragin 1987; Rueschemeyer and Stephens 1997; Tilly 1984). Random selection of cases-say, of particular regimes or nations froill hundreds across the globe-is often not appropriate; and it is far from the only way to test hypotheses rigorously. Whatever the risks or drawbacks, hypotheses can be rigorously tested even when scholars cannot sainple from large numbers of truly independent cases (for an excellent overview of strategies of inference in small-n research, see Mahoney 20002). Alternative strategies of causal inference have been developed and applied, because there are important intellectual and practical advantages to focusiilg agendas of research on a small to interlne- diate number of cases, including instances of substantively compelling outcomes and arrangements we want to understand.

For example, a scholar who wants to understand revolutions is u11- likely to be willing to start with a purely random set of times and places; she will want to make sure that clear-cut revolutions are iilcluded in her research. But in the best past studies of revolutions-as in many other historical-institutionalist literatures-it has long been standard to test hypotheses with comparisons between positive cases, where phenomena of interest occur, and closely matched negative cases, where the phenomena do not occur (Skocpol 1979; Goodwin 2001). Speaking about comparative studies

in many topic literatures-analyses have juxtaposed time periods, regions, and policy sectors, turning what appear to be one or a few national instances into settii~gs for many carefully compared cases. And even within what appear to be single case studies, empirical observations have often been nlultiplied by formulatiilg and testing hypotheses about thc inecha- nisrns that connect causes to effects (Bennett and George 1997). All of these strategies of analysis have been used to good effect by scientifically sound historical-institutional studies, and the rapidly developing methodological literatures will make it easier for future scholars to recognize and apply the standards and rationales for best practices ill case-based and small-n research.

No matter what theories or research methods are deployed, individual studies in isolation never do inore than move the scholarly enterprise a step or two forward. Scholarship is an inllerently commuilal eilterprise, and so it is appropriate to reflect on how well clusters of scholars do at uccun~ulut- ing valid and ilnportant findings. Here is where historical institutionalists do quite well, in our view, because substantively compellillg, problem- driven research facilitates exactly the sort of iiitellectllal cu~iiulation tliat allows a community of researcliers to make clear progress over tilne.

Because historical institutionalists tend to tackle big, Iiumanly i~npor- tant questiolls, scholars are likely to come back to the issues again and again over several acadeiilic generations. Within each generation as well, smart people jump in, willing to argue about how best to frame the questions and describe patterns wort11 explaining, as well as about how to for- mulate and test hypotheses. Researchers reexamine cases and llypotl~eses and extend established l~ypotl~eses to new sets of cases. A good example is spelled out in James Mahoney's review of several decades of llistorical- institutionalist research 011 the origins and dynamics of democratic and authoritarian regimes. In this literature, research took off from pioneer- ing agenda-setting works (i.e, Moore 1966; O'Donnell 1973; Linz 1978), with waves of succeeding scholars executing dozens of historical a i d comparative-historical studies that served to refute some argu~nents, refine others, discover new lines of causal argument, and extend findings across eras and continents. Cumulative research now offers a convincing picture of when, why, and how different types of regimes have emerged in various continents over much of modern world history. "Most of our knowledge about the relationship between democracy and various other factors (social classes, international conditions, elite behavior) has been derived through such step-by-step accumulation . . ." Mahoney concludes. "Plainly, if one were to strike all comparative-historical research from the record, most of what we currently know about the causes of democracy and authoritarianism would be lost" (Mahoney 2002). What Mahoney demonstrates abotlt the literature on democracy and authoritarianism holds as well for literatures on revolutions, state building, and varieties of public policymaking, to name just some areas of clear-cut progress in recent decades. In each of

i

A 2 z = 5 2 9 . E 5 1 Y r m + * O h _ C "i

k . - 2 O C - l a 9s r ' 2 - r j .- ., Y -= 0 2 - - 4-

v, d LC 5 " 0 + ;

3 " $ 2 7 .5 5s 3 y . ; : Q ) r j o

$ - 2 "i, arJ

LC-

i L C - L Q ) o - m 2343 i , m C 2 g g 2

a Y i)

o c , r J . 5 Y .- s , Q ) + - =

2 5 2 2 Q ) + 0 -

a m .o 0 a, + z g s -j m ;.. $ 3

. - a"' % r' gz 2 X - *. ' 2 2 sg 2 E F , t ; "

: g 2 & o ; m 2

PIERSON A N D SKOCPOI , 8 HISIORICAL ~ N S ' T I T L ~ ' I . I O N A L I S M ) 721

four time period5 (eacll time period lasting a decade or more). The result is the most serious effort yet to test the strength of a numbcr of prominent theories of institutional choice over an extended time period, a11d [he develop~llent of some strikingly original insights into the sources of change a ~ ~ d continuity in legislative rules.

Across the globe, new democracies are emergir~g from en-

trenched authoritarian polities. Explaining variations in new institutional rules and ill the processes by which thcv take shape reprcse~lts a ~najor challenge for students of comparative politics. In her new book OII Institu- tiolaal Change and Political Coiatii~uity in Post-Soviet Ceratrul Asiu, Paulinc Luong (2002) compares the establishment of electoral systcn~s in Kaz21kl1- stan, Kyrgyzstan, and Uzbekistan and, in the process, develops of ~ l c w and potentially generalizeahle theory of "institutions designed under transitional circu~nstances," blending insights fro111 both rational-cl~oice institutio~~alisiil and historical insti t~~tionalis~~l. Altlio~~gll historical institli- tionalists have directed our attention to persistent legacies fron~ t l ~ c past.

C I I I ~ S alliollg new rules of the gallle can and do cmeqe from strategic Ijarg . elites, especially in a period of crisis and l~ilcertai~lty. Yet rational-choicc aplxoxlies too easily fall into the trap of assutni~~g that clite bargaining over new arra~ige~ne~its occi~rs O I I a tabula rasa, \vithoi~t regard to ell- trcnched understandi~lgs and power relatio~lships. 111 Jones's model, actors cllallgc goals a ~ l d perceptions in rcsponse to unccrt~linty and bargain in a

dynamic way, producing different otltconles ill three ccntral I\sian polities with many prior structural si~nilarities. But elites work fro111 power posi- tions and understandings cmbcdded in inherited arrangements-i~~tleetl they try to encode those older meanings and powcr relationships into seenringly new structures. 1.11ong's work would not have been possiblc witllout prior breakthrouglls in game theory. At the sa111e time, she is in- spired by the theoretical agendas set out b), such historical institutionalists as Kathleen Thelen. calling for careful analyses not just of institutional outco~nes but of the temporally and structurally ernbedded processes by which actors entrenched in previous institutions maneuver to create modi- fied arrangements that retain many continuities from the past. Rilethod- ologically, moreover, Luong nlakes a pon!erft~l case for close compariso~l of kindred polities within a geocultural region, as a valuable laboratory for working out expla~iations of institutional change that may have nlucll - broader theoretical application.

In different ways, both Schickler and Luong work at the intersection of rational-choice theorizing and historical-institutional analysis. Our third instance of creative synergy, Andrea Campbell's Shaping Policy, Shapii~g Citizens: Senior Citizen Activism (2nd the American LYe1f;lre State (forth- coming) exelnplifies possibilities for coll~bining strong points of survey- based behavioral ailaly5is with analytic stratcgies that have been developed by historical institutionalists. Campbell's subject is a traditional focus of behavioralis~n: citizen participation. Specifically, she is iuterested in thc 2

changing patterns of political beha~ior of the U.S. elderly, and she ably uscs survcy results to docunleilt aggregate behaviors and attitudes.

:It the same t i~ne, Can~pbcll drabvs two major insights fro111 historical institutionalisn~. First, she stresses the in~portance of studying participatio~~ over time, both to increase variation in her outcome and to identify and as- sess the contrib~~tion of rclativcly slow moving factors to changing patterns of participation. Second, she liighligl~ts the contributions of policy feedbacks-specifically fro111 Social Security and 1CIedicare-to both the lcvel and style of citizen participatio~~ among the elderly. 'I'llrol~gh the paiustak- ing assenlbly of longitudinal iiiicrodata on participation among the elderly and thc thoughtful con~bination of multiple traditions, Campbell demonstrates how major federal programs both raised t l ~ e level of eltlerly participation and prodllced a n uniusual pattern of participation. Aino~ig working-aged 1J.S. adults, participation is heavily skewed toward better- ed~~catetl slid Ilighcr-i~lco~ile groIips, L L I ~ this "~~pwilrd" bias is Iargcly ab- sent amolig thc eldcrly, where concerns about Social SecuriQ and Medicarc l~avc L-~~couragctl ~l~itldle- and lower-strata elders to participate much inore fillly. 'l'hrough its crcative combinatio~~ of behavioral and historical-institutio~~al analysis, Ca~n~~be l l ' s study speaks to major issues of our timc. Its findings are highly rcleva~~t to ongoing public as well as scholarly debatcs about eu t i t l e~~ ie~ l t r c f o r ~ ~ ~ s and a b o ~ ~ t civic engagelllent in U.S. democracy.

h o n e of the exemplary new studies we have just described could have bcen protluced without illlagillative cfforts to draw on and combine the best of ~nultiple analytical stratcgies. By featuring thesc studies, we mean to underline that it has not bee11 our purpose to pit major approaches against one another in any kind of zero-sum game. We have simply sought to clarify the distinctive contributio~rs and advantages of historical-institutionalist scholarship Indeed, a good reasoil to clarify distinctions among approaches and highlight potential comparative advantages of each is to fa- cilitate exchanges that corllbi~le the strengths of different approaches to maximunl effect. We have made the case for historical institutionalism as one of the three research pillars in contemporary political science, and our task has been made easy by the outpouring of so much excellent work in recent years. Historical institutionalism has become, of late, more self- conscious both theoretically and n~ethodologically, in ways we have indi- cated and illustrated throughout this review essay. Readers who wonder what political sciei~ce would be like without self-conscious historical insti- tutionalisrn need only ask what we would have lost if all the works in our bibliography, and many others like them, had not been published. We think the answer is obvious: without historical institutionalism, our disci- pline would be shorn of inoch of its ability to tackle major agendas of concern to all political scientists. And without historical institutionalists, political science would have much less to say about questions of great im- port to pcoplc beyond as well as within the ivory tower.

Pierson Skocpol 2002

Documents

Transcript of Pierson Skocpol 2002