Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Sage Publications, Inc. University of Utah

Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

University of Utah

Democratic Progress and Regress: The Effect of Parties on the Transitions of States to and Away from Democracy Author(s): Brian Lai and Ruth Melkonian-Hoover Reviewed work(s): Source: Political Research Quarterly, Vol. 58, No. 4 (Dec., 2005), pp. 551-564 Published by: Sage Publications, Inc. on behalf of the University of Utah Stable URL: http://www.jstor.org/stable/3595641 . Accessed: 20/05/2012 21:03
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Sage Publications, Inc. and University of Utah are collaborating with JSTOR to digitize, preserve and extend access to Political Research Quarterly.

http://www.jstor.org

Democratic Progress and Regress: The Effect of Parties on the Transitions of States to and Away from Democracy
BRIAN LAI, UNIVERSITY IOWA OF

RUTH MELKONIAN-HOOVER, GORDON COLLEGE

Thisarticleexamineshow politicalpartiesand partycompetition affectthe likelihoodof nationsbecomingand While manyscholarshave long assumedthatthis is the case, the rolesof partiesand democracies. remaining in are such a likelihoodhas rarely been examinedrigorously cross-national partycompetition indeedaffected, In evaluations. additionto examiningthe links betweenpartiesand politicaltransitions, analysiscontrols our for other factorspurportedto have a significant To effecton democratization. test the effectsof partiesand on the transition and survivalof democracies autocracies, articleutilizesevent this to and partycompetition on all countriesin politicaltransition betweenthe yearsof 1950-1992, using threedifferent historyanalysis
measures of democracy. Through this multifaceted and unique approach, we are able to demonstrate that

states acrossall threemeasures democracy, of role partiesdo indeedplayan important in causingauthoritarian to transition democracy helpingdemocratic and to nationsremaindemocracies.

Political parties are not always perceived as the most desirableof political institutions by governmentsand politicians. Some prefer democracy without parties, fearingpolitical partiesa source of factionalismand a threat to stability.ThomasJeffersonhimself wrote, "IfI could not go to heaven but with a party,I would not go there at all." Indeed, in the early decades of U.S. history other critical structuresof democracywere establishedpriorto the formal institutionalization political parties. of In spite of such concerns and prejudices, and while acknowledging that numerous factors do play a part in this democratization, articlemaintainsthat the development and formationof politicalparties,and the politicalspace for such parties to participate,are primary factors in a states chances of becoming and remaining a democracy While many scholars have long argued this to be the case, it has examinedand empirically tested.Tobe rarelybeen rigorously some types of democracycan exist without strong, certain, coherentparties (Schmitterand Karl 1991), but we present evidencethat the inclusionand competitionof partiesis crucial to democracy's occurrenceand its survival;nations are to become or remaindemocraticwithout it. unlikely Voluminous literature in political science has been devoted to studying democratization,analyzing the factors that lead to democracy and its alternatives,as well as the consolidation thereof. We maintain, however, that political parties have been insufficientlyanalyzed in prior research. While many scholars have assumed democracies cannot operate without parties, cross-national evaluations have often overlooked the role that these fundamentalinstitutions play in politicaltransitionsthemselves.In those analyNOTE: The authors wish to thank Dan Reiter, RichardJoseph, Dennis Hoover, and Emily Lai for their help with this article. Political Research Vol. Quarterly, 58, No. 4 (December2005): pp. 551-564

ses that have included parties,the focus has often been limited to the negative, destabilizingimpacts of party fragmentation and extremistparties on democratization.While this is one way partiescan influence a democracy,there are also crucial positive linkages between parties and democratization that warrantsystematicevaluation. this artiTo test the effectsof partieson democratization, cle utilizes event history analysison all countriesin political transitionbetween the years of 1950-1992, using three differentmeasuresof democracy. examine how partiesand We party competition, along with control variables for other institutional, economic, international,and cultural factors, affect the process of a nation becoming a democracy and We preventit from revertingback to authoritarianism. evaluate the possible reciprocalrelationshipbetween democracy and our party measures as well. Through this multifaceted approach,we are able to show that acrossall three measures of democracy, parties play an important role in causing authoritarian states to transitionto democracyand causing democraticstates to remain democracies. This researchmakes several distinctive contributions to the literature.In a fresh way, it empirically examines the function of parties in democratic transitionsand survival. This articlepresents one of the first quantitativeanalysesof the role of parties and goes beyond simply the effects of multipartyismand party extremism to actual party institutions and competition. In addition, this article contributes to the debate about a wide range of other relevant factors purportedto affectdemocratization explicitly examining by proceduraland structuralfactorsempirically. The articleproceeds in five parts. The firstbrieflyidentifies gaps in the existing literature vis-a-vis parties. The second part offers a discussion of the importance of competitive political parties in democratization,culminatingin testable hypotheses. In the third section we review the researchdesign used to test the effects of parties and other 551

552 552 control factors on the transition from authoritarianism to and on democraticsurvival,explicatingdifferent democracy measures of democracy and of our independent variables. The fourth section presents the results of the empirical analysis, and the final section highlights the implications and conclusions of this research.
THE PUZZLE OF PARTIES

RESEARCH POLITICAL POLITICAL RESEARCH QUARTERLY QUARTERLY

In the extant literature on democratization, findings regarding the role of parties are often either missing or mixed. Some empiricistshave focused on structuralsocioeconomic factors in democratization(Bollen and Jackman 1985; Muellerand Seligson 1994; Gasiorowski1995; Przeworski et al. 1996; Gasiorowskiand Power 1998), while others have focused more on social factors such as ethnolinguistic fractionalizationand/or religion (Feng and Zak 1999; Gasiorowskiand Power 1998; Muller and Seligson 1994; Gasiorowski 1995; Bernhard, Nordstrom, and Reenock2001). While these findings help us to understand the process of democratization,they overlook the institutional effects of inclusive and competitiveparty systems. Those quantitativestudies that do focus on parties are most often designed to test their potentiallynegative effects on democraticsurvival,specificallythe effects of party fractionalizationand/or extremism.Furthermore,the results of these studies point in conflicting directions. Some have found that extrememultipartyism leads to the breakdownof democraticstates (Linz 1978), while others have found that and extremistpartiesare associatedwith govmultipartyism ernment instability,but not necessarilydeadly violence and breakdown(Powell 1982). Others do not find any relationship between parties and democratic survival (Power and Gasiorowski 1997; Gasiorowski and Power 1998). Bernhard, Nordstrom, and Reenock (2001) determined that political institutions like parties do not matter to the survival of democraciesby themselves, but ratherthat it is the interactionof political institutions with economic performance that affectsthe survivalof a democracy. While we concur that partiesmay have an interveningas well as primaryrole in relationshipto democracy, believe we much of these negative or non-existent findings are due to the ways in which partieswere incorporatedin these studies. The analysisof partiesin many of these studies is based in part on presuppositionsthat parties (whether too many or too narrow) affect democracy negatively by creating instabilityand distrustand by reinforcingcleavages.Undeniably, in post-communist countries, parties have been found to be the politicalinstitutionsthat inspire the greatest levels of distrust amongst the public (Mishler and Rose 1997: 422). Still, we suggest that the crucial roles of party legitimacyand competition have been overlooked.
THE ROLE OF COMPETITIVE POLITICAL PARTIES

mediarybetween the state and civil society.Partiescomprise a legitimate, stable proceduralmechanism for the alternation of power and the representation majorityand minorof interests.Opposition partiescan act as necessarychecks ity on the ruling party, assure the representationof citizens' policy preferencesand by so doing, increasecivic participation. Party competition can reduce state limits on citizen dissent and lead to increased parity and access to government resourcesfor parties. It can also provide a democratic means of redress for the citizenry,and stimulate increased responsivenesson the part of the ruling party. PartyInclusion As Juan Linz and AlfredStepannote, political partiesare vital in "habituation the norms and proceduresof demoto cratic conflict-regulation" (Pye 1990). The ability of political partiesto legitimatelycontest the ruling party reinforces norms of peaceful alternation of power (Linz and Stepan 1996). Political parties thus transmit and solidify democratic norms, helping a state democratize. In order for democracyto stabilizeit must be known as "theonly game in town," and parties help in this process of political normalization. As Nancy Bermeo (1990: 369) emphasizes, "electorallaws are really only important as institutional assets for political parties. Partiesare ultimatelythe institutions that influence the 'democraticcompromise'most." While some nations have tried to democratizewithout opposition parties, democracyin such cases has been tenuous at best. In the case of Uganda, afternearly two decades in power PresidentMuseveni,infamousfor supportingwhat he labels a "no-party democracy,"is finally steering the nation towardmultipartyism.Leadersof a majoropposition movement threatena guerrillawar if they remain shut out of the system; a lack of viable political alternativesto regulate conflict and produce compromisehas led them to consider a path of violence (TheEconomist, 3/1/2003). President Museveni has resisted political pluralism thus far because, like many other leadersin the region, he fearsthreatsto his alliance'spower and party fragmentationbased on tribal cleavages (Dahl 1971). While successfully avoiding fragmentation, Uganda has clearly sufferedfrom an undemocratic concentrationof power (Bogaards 2000). Partiesare not only a means for the peaceful transferof power, but provide a peacefulmeans to facilitateadaptation and compromise.Politicalparties,opposition groupsin particular,can respond and adaptto changesin society because they tend to have less of a vested interest in preservingthe currentregime order (Coppedge 1993). In addition, parties are useful in providing a sense of longevity and institutionalization. By offering a means for future regularity,they increase an individual candidate'schance of winning successive elections and of achieving policy changes while in office (Aldrich 1995). Partiescan also play a crucial role in channeling citizen participationin appropriateand productive ways. Groups with a history of authoritarian behavior (e.g., the military)

The institution of political parties plays a fundamental role in democratization,in part by functioning as an inter-

DEMOCRATIC PROGRESS AND REGRESS

553 553

may be less likely to intervene in the political system because of the assumption that radicalvoices will be moderatedby working within the institution of political parties. A longstandingconcern in democratization literatureis that narrowpartiescan function to exacerbatesocietal factionalism. For instance, while Samuel Huntington (1968) has argued that parties are important to establish and consolidate democracy, has warned that the process of elections, he and partiesthemselves,can deepen societal divisions if parties appeal to ethnic and communal conflicts or are antidemocratic(Huntington1996). However,this concern ignores the fact that, over time, working within a system of party inclusion and competition often tempers the agendas of political parties-even parties based in particular social divisions (religious, ethnic, linguistic, etc.). A prime example is that of the BharatiyaJanata Party (BJP) in India. Although it grew in part due to distinctive religious identificationand tactics, when in power, under the leadershipof Prime Minister Vajpayee, the BJP dropped much of its Hindu nationalistagendain orderto form coalitions, attract more voters, and furtherits opportunities to retain power 2004). (Rudolphand Rudolph 2002; Lancaster Democracyrequiresa balancebetween conflict and consensus, arguedAlexis de Tocqueville(Lipset 1960), and parties (as well as otherprivateassociations)often help maintain this balance.Even partieswith ambitiousor radicalagendas arenot necessarilydisruptiveto this balance.Forexample,in contrastto KarlMarx,who presumedworkers'partieswould be a basis for revolution,SeymourMartinLipset(1960: 22) has shown that, "it is precisely in those countries where workers have been able to form strong unions and obtain representationin politics that the disintegrativeforms of politicalcleavageare least likely to be found." By providing a balance, a means to address conflict and consensus in a legitimate fashion, parties can thus help nations avoid the illiberal authoritarianand revolutionary tendencies of the right and the left. Accordingly, hypothwe esize the following:
Hla:

Thegreatertheinclusion political of partiesin authoritarianstates,themorelikelythosestateswillexperience a democratic transition. Hlb: Thegreaterthe inclusionof politicalpartiesin democratic states, the more likely those states will remain democracies.

PartyCompetition For democracy to be initiated and consolidated, states need more than the mere presence of partiesin the system; they also need those partiesto be meaningfullycompetitive. By party competition, we refer to the real possibility of changeand influenceon the politicalsystem. This is fulfilled when partieshave the space to campaign,to rally,and to be active players.To be clear,we are not making claims about electoralsystems. As has been observed of late, appropriate Africannations undergoingrecent political transitionshave

experimented with multiple forms of electoral systems; of dominantpartysystemshave regardless the configuration, been the end resultin most cases (Bogaards 2000). Without competition, resources can too easily be monopolized by the party in power. For much of Mexico's recent history, due to its hegemonic party system, opposition parties have been considerablydisadvantagedbecause of the ruling party's privilegedaccess to media coverageand state coffers (Sartori1976). The InstitutionalRevolutionary Party (PRI) ensured that other parties functioned only as second-classinstitutionsthat legitimized,not threatened,its power (Ware 1996). By contrast, in nations with open access to public funds and governmentpositions there is a greaterlikelihood of opposition party support for democracy,even by non-liberal parties, as evidenced by multiple non-liberal parties in the interwar period in European democracies (Ertman 1993). With real competition and improved access to political resources, opposition parties can move from repression and co-optation to become strong, independent institutions. Another advantageof party competition is that political parties operatingin politically pluralisticsystems are more likely to have distinctivepartyplatformsthan those in dominant partysystems (Huntington 1968). Politiciansmay feel a greaterobligationto be accountableto campaignpromises consistent with the party line than when operating in an arena in which no real competition exists. In the U.S., in states that historicallyhave been dominated by one party, numerous candidates of different ideological persuasions have aligned themselveswith the dominant party to ensure greaterpersonal success. Accordingly,state parties, seeking to incorporatebroaderbases, have moved to the left or right of their respectivenational parties (Lipset 1960). Likewise, historically dominant, aggregate parties like the PRI in Mexico, the CongressPartyin India, and the LDPin Japan have not been known for being ideologically distinctive as they have sought to incorporateor co-opt groupswith varying agendas (Powell 1982). Further,political pluralismprovides for a much needed check on the party in power; without competition it is far too easy for the ruling partyto limit dissent and force cooperation. Clearly, party competition can provide a viable threat that political power may be lost if expectations are not met; it is politically advantageousto placate the populace with good governance.For example, in his comparison of rural politics in the mid-20th century in Mexico and Venezuela, Michael Coppedge (1993) found that richer competition in the latter country resulted in decreased repression of and greater respect shown to the citizenry. Likewise,as BinghamPowell has found, political pluralism is associated with low levels of non-democratic political engagement and provides strong incentives for democratic involvementwithin the system (1982). In short, competitive parties provide a check on ruling power, help ensure greaterequity of resources for parties, and createmore incentivesfor good governanceand support for democracyThis leads to our second set of hypotheses:

554 554
H2a: The greaterthe competitiveness politicalparties in of

POLITICAL RESEARCH QUARTERLY

authoritarian statesthe morelikelytheyare to become democracies. H2b: The greaterthe competitiveness politicalparties in of democratic states the more likely they are to avoid a to reversion authoritarianism. To be sure, our positive hypotheses about the significance of parties are not entirely novel. Other scholarshave noted the beneficial roles of parties (Randalland Svasand 2002). But what is missing is a large-scale comparative empiricalfoundationfor these positive suppositions.
RESEARCH DESIGN AND DATA

To test our hypotheses, we conduct two sets of tests using event history analysis of the effects of parties on the and H2b).For the firstset of tests, we examine whether parties hasten the transitionof authoritarian states to democracy.The second set of tests examine whether parties help prevent democracies from revertingback to authoritarianism. This is an appropriate method given that event history analysishas been widely used to study the factorsthat either lead an authoritarian state to transitionto democracyor lead a democracy to revert back to authoritarianism(Reiter Nord2001; Pevehouse2002; Gasiorwoski1995; Bernhard, strom, and Reenock 2001; Feng and Zak 1999; Przeworski et al. 2000). Simultaneously,while numerous studies on democratization have used event history models, there has been little consensus on how to measure democracy in such models. Accordingly,this articleexamines three of the more prominent measuresof democracy,therebytesting for the robustness of the effects of party across different conceptualizations of democracy First,we use the Political Regime Change Dataset (Gasiorowski 1996), which categorizes states into four types of regimes: authoritarian, semi-democratic, democratic, and transitional. This data covers 97 Third Worldcountries from their independence to 1992, drawing on definitions of regime type from Diamond, Linz, and Lipset (1989). The next measureof democracywe utilize is that of The ACLP Political and EconomicDatabase (Przeworski et al. 2000). This data categorizesstates as either democraciesor dictatorships,examining 135 countriesfrom 1950 to 1990.1 The authorsof this projectdefine democracyas "aregimein which those who govern are selected through contested elections"(Przeworskiet al. 2000). Finally,we use the Polity III data for the last measure of and democracy(Jaggers Gurr1995). The Polity IIIdatahave been widely used to study democratization(Reiter 2001; Pevehouse 2002) as well as the effects of democracy on
1The startdate is 1950, independence (if after 1950), or when economic growth data is first available.The codebook for this data is availableat http://www.ssc.upenn.edu/-cheibub/data/ACLP_Codebook.PDE

survival of autocracies (Hla and H2a) and democracies (Hlb

international conflict (Russett and O'Neal 2001).2 The Polity data rank states on a variety of measures of democracy and autocracy.Summing across these differentmeasures leads to a democracy and autocracy score, both of which rangefrom 0-10, with 0 correspondingto least democratic (for the democracyscore) or least autocratic(for the autocracyscore) and 10 correspondingto most democratic or most autocratic. Generally, researchers subtract the autocracyscore from the democracyscore to get a -10 to 10 measure.Previouswork on democratization used 7 as a has threshold for considering a state as a democracy (Reiter 2001; Pevehouse 2002). This articlestays with that convention, consideringstates with a 7 or higher on the -10 to 10 scale as democraciesand all others as non-democracies.The Polity data comprisea broad dataset,including all countries with a population of at least 500,000 from the 19th century to the present. When we evaluate these three measures of democracy side by side, we find correlationcoefficientsyielding values between .67 and .77.3 While there is some expected overwhen a state is a democlap, each datasetdefines differently racy and when it is no longer a democracy.For example, in the case of Argentina,the ACLPdata show it to be a democracy from 1951-1953, 1958-1961, 1963-1965, 1973-1975, and 1983-1990 (endpoint for the ACLPdata). The Political Regime Change data indicate that Argentinawas a semidemocracy from 1958-1965 and 1973-1975. It was a democracy from 1983-1992 (the endpoint for these data). Finally,the Polity III data code Argentinaas a democracy from 1983-1992 (the endpoint for our use of these data). So while there are similarities (they all identify a democratic Argentinain 1983), there are also differences(only ACLP considers Argentina a democracy prior to 1975, and the Polity III data only code Argentinaas failingfrom autocracy once).4 Another point of difference between these data is that the PoliticalRegime Change dataset only covers Third World countries. Altogether,these three datasetsoffermuch-utilizedmeasures of democracy;the inclusion and examinationof each provides robustnessto our empiricalfindings.And to elicit the most appropriate comparisons,we utilize the set of cases in available each datasetfrom 1950-1992 (1990 for the ACLP data), conductingtwo empiricaltests (the survivalof autocraciesand the survivalof democracies)on each datasource. The dependent variablefor the first set of empiricaltests (the survivalof autocracies)is the durationof an autocratic government until it becomes a democracy or until 1992 (1990 for the ACLPdata), at which point the data are right censored. For the second set of empiricaltests (the survival
2

The literatureon democracy and internationalconflict is vast. Russett and Oneal 2001 provides a helpful overview of the literature. 3 Each is recoded into two values of 0 for non-democracy and 1 for democracyfor this correlation. 4 A full discussion of all the regime transitiondates is beyond the scope and purpose of this article.Since all three are widely accepted measures of democracy, simply use them to test the effects of parties. we

DEMOCRATIC PROGRESS REGRESS AND DEMOCRATIC PROGRESS AND REGRESS

555 tion in partycompetition and legitimacyacross states.9One potential limitation with the Polity measure of party competition is that it is one component used in the construction of the Polity measure of authoritarianism. This creates the that independent variablex is explaining potential problem a dependent variablebased in part on independent variable x. Fortunately,the Polity measure of authoritarianism utilizes numerous other components, reducing the likelihood that the use of party competition significantly biases the findings. As well, the use of the Political Regime Change Data and the ACLP data helps reduce bias, as these two measuresare not based on the Polity data. In addition to the hypothesizedvariables,we control for some factorsthat have been found to affectthe transitionto and survival of democracies. First, we include economic developmentand growth,measuringthem respectivelywith the log of GDP/capita and the annual change in real GDP/capita (i.e., the real annual growth in GDP/capita) (Gasiorowski 1995; Przeworski et al. 1996; Gasiorowski and Power 1998; Bernhard, Nordstrom,and Reenock2001; Reiter 2001). Economic development has been found by many to be conducive to democracy (Gasiorowski and Power 1998; Bernhard, Nordstrom, and Reenock 2001; Reiter2001). Data for both of these variablescome from an expanded dataset produced by KristianGleditsch to overcome some of the missing data inherent in national economic measures. Gleditsch (2002) uses the Penn World Tablesas a startingpoint and uses other sources to supplement missing data. In addition to economic variables,we include controls for culturalfactorssuch as variousreligionscited frequently in the literatureas being compatible or incompatible with democracy.Catholicismappearsto be increasinglysupportive of democracy (Philpott 2004), while there remains a democracy gap between Muslim states and the rest of the world, particularlyArab states (Karatnycky2002; Stepan and Robertson2003).0 We include two dichotomous variables, one for Catholicismand one for Islam, each coded as 1 if the largest religious subpopulation is Catholic or Islamic.Data are from Ellingson (2000) and the COW Cultural Dataset(Henderson 1998). A second culturalvariable

of democracies),the dependent variableis the durationof a democratic state until it becomes an autocracy or until 1992/1990.5 For the PoliticalRegimechange data, we consider a state to be an autocracyif it is coded as authoritarian. Thus it can fail when it becomes a semi-democratic democratic.Simor ilarly,we considera stateto be a democracyif coded as democratic;it can fail either becoming semidemocraticor autocratic.6The ACLPdata only have two values for regimes, dictatorshipsor democracy:a state is coded as authoritarian if a dictatorship,and failswhen it becomes a democracy. The reverseis true for democracies.Finally,for the Polity data, a state is considered an authoritarian governmentif it scores lower than a 7 on the combined -10 to 10 scale and it fails when it becomes a democracy (7 or greater).Again, the reverse is true for democracies.For all of these measures, states that were either authoritarian democraticprior to or 1950 areleft censored,which is a common, albeitimperfect, (Reiter2001; Pevepracticein the study of democratization house 2002; Gasirowowski1995; Bernhard, Nordstrom,and Reenock2001; Feng and Zak 1999). 7 To test Hla and Hlb, the inclusion of political parties,we use a measureof political inclusion from the CrossNational Time SeriesArchive(1997). The variableis coded ordinally: 0 = No Parties; 1 = Significant Exclusion of Parties;2 = Minor/Extreme Parties Excluded; and 3 = No Parties Excluded.We test hypotheses H2a and H2b, the competitiveness of political parties,by using a variablefrom the Polity Data. It is an ordinalvariablewith the following values: 0 = Unregulated-No enduringnationalpolitical organizations; 1 = Repressed-no party activity is allowed; 2 = Suppressed-some limited party activity;3 = Factional-party activity is primarilythat of parochialor ethnic parties that reinforcesocial cleavages(Jaggers Gurr1995); 4 = Tranand sitional-increased party activity with minor limitations; and 5 = Competitive-significant partyactivity(Jaggers and Gurr 1995).8 These variablesare useful in examiningvaria-

The model actuallyestimatesthe effects of the independentvariableson the hazardrate for the failureof autocracyor democracy, this hazard but rate is based on the durationof democraciesand autocracies.Thus, the actualvariables used to estimatethe hazardratearethe durationvariables. 6 For the transitionstates, we follow Gasiorowski's (1995 suggestion and code them as the regime they were under the previous year. 7 Left censoring refers to the counting of time at the start of a series. For example, some statesbecome democraciesin 1950 and are thus credited as being a democracy for one year in 1950. However, a state like the United States has been a democracy since the late 1700s but is also counted as being a democracyfor one year in 1950. The problem of left censoring is common in studies of democratizationbut is better than ignoring temporaldynamics altogether. 8 While this measure potentially conflates party competition with party unity, we believe that this scale best captures the theoreticalaspects of competition(e.g., partysystems that are factionalare likely to reduce the positive benefits of competition). We also test the robustness of this measureby dichotomizingit two ways. Firstwe createa variablethat is 1 if a state scores 3 or higher and 0 if otherwise. Second, we createthree variables:one that is 1 if a state scores 0, 1, or 2, and 0 if otherwise;a second that is 1 if a state scores a 3 and 0 if otherwise. The results are

essentially the same. The only differenceis that the dichotomous version of party competition is not significant in the Political Regime Change Survivalof Democracymodel. However, when we utilize the three variableswith the first as a referencecategory,the third variables with the first as a referencecategory,the third variableis significantin all models. 9 Even for democratic states, there is variationin both party measures, though the variationis much smallercomparedto that of authoritarian states. 10 While some researchhas also focused on the effects of Protestantism (Bollen and Jackman 1985; Woodberryand Shah 2004), in our study there were too few cases where there was variationin the dependent variablefor Protestantcountries, leading to exclusion from a few of the models. In addition, overall,the effects of Protestantism were generally insignificant and thus, for the sake of model comparison, were not included.

556 556 evaluated is ethno-linguistic fractionalization,which has been associated negatively with democracy (Muller and Seligson, 1994). We use data from Roeder(2001). We also consider several internationalcontrol variables, including a measure of the percentageof democraciesin a state'sregion, a factorshown to affectthe democraticstatus of a nation positively (Gasiorowski1995; Przeworskiet al 1996; Gasiorowskiand Power 1998; Bernhard,Nordstrom, and Reenock2001; Reiter2001). Regionsare divided based on the Correlatesof War coding: Americas,Europe,Africa, the Middle East, and Asia/Oceania.Based on each different datasetof democracy,we construct a measurebased on the percentage of states in a country'sregion that are democratic. A measure of a state's democratic history is also included, which has received mixed empirical support in the literature(Feng and Zak 1999; Gasiorowski1995; Bernhard, Nordstrom, and Reenock 2001; Przeworski et al. 1996). This measureis the number of years a state has been democratic (since 1950 or its independence) and is based on each of the measuresof democracy. Finally,in our test of the survivalof democraticstates,we include anotherinstitutionalvariable,a measureof the type of democracy, presidential (coded 1) or parliamentary (coded 0) (Gasiorowski 1995; Przeworski et al. 1996; Gasiorowskiand Power 1998; Bernhard,Nordstrom, and Reenock 2001). Some have found that parliamentary systems are significantly associated with democracy (Gasiorowski and Power 1998; Przeworski et al. 1996), while others have not (Gasiorowski1995; Bernhard,Nordstrom, and Reenock 2001). Data come from the Cross NationalTime SeriesArchive(1997) for the PoliticalRegime Changeand Polity Data. For the ACLPdata, we use a measure of presidentialismalreadydelineatedtherein. To examine how these independent control variables affect the duration of autocraciesand democracies,we use two event historymodels, a Cox semi-parametric model and a Weibull model. In addition, all the independentvariables arelaggedone year. 11Thedifference between the two models is whether or not they assume an underlying distribution about the failureevent (i.e., transitionto democracyor failure of democracy).The Cox model is semi-parametric and does not assume a functionalform of the durationdependence of the data, making it a more flexible event history model. The Weibull model, while not as flexible as a Cox model due to its assumptionsabout the shape of the failure distribution,has been used to study whether democracyis duration dependent or not (Bernhard, Nordstrom, and Reenock2001). Given that both have been used in the literature and we have no a priori reason to favor one over the other we use both to ensure the robustnessof our findings. A final methodologicalissue is the direction of the relationship between the dependent variableand independent variables (democracy and parties). Muller and Seligson

RESEARCH POLITICAL QUARTERLY

(1994) addressthis concern in their examinationof the relationship between civic culture and democracy.The issue of reciprocitycreates distinct problems for our two different sets of analyses. In our examinationof whether or not parties help end non-democraticregimes, the issue is not one of democracypromotingparties,as the cases we observeare all non-democracies.However, a potential problem is that alternativefactorscould simultaneouslylead to more inclusive and competitive party systems as well as the end of autocraticgovernments.Thus, while it may appearthat parties are ending non-democracies,it may be a thirdset of factors influencingboth partiesand the regime type of a state. This is not a major concern as we are able to control for importantfactorsshown to influence whether or not a nondemocracy transitions to a democracy.Also, our use of lagged independent variables should help alleviate problems of simultaneity. Finally,as discussed below, we employ additionalstatisticaltests to address this problem. In our second set of analyses on the influence of party inclusion and competition on the survival of democracies, our problem of potential reciprocityis similar to that identified by Muller and Seligson (1994). While democratic institutions such as regular elections, legislatures, and checks on executive power can create a system that allows political partiesto flourish,we believe that partiesare foundational in establishingand fleshing out democraticstructures in the first place. In addition, our use of lagged independent variableshelps address this problem. If the causal arrow is the reverse of our hypotheses-the collapse of democracy leads to the collapse of party competition and inclusion-then looking at the effects of party competition and inclusion in the year prior (time t-1) on whether or not a state is a democracyin the year of focus (time t) should avoid results based on this alternativecausal relationship. For example,if a democraticcountrystartsto become undemocraticin a particular year,the influence on partiesshould be in that year or in futureyears, not the year prior. Thus, in our models where we examine the survivalof democracy as a function of variablesincluding party competition and inclusion, if the causal arrow is truly the reverse, namely that the collapse of democracyleads to the collapse of parties, then we should see no results or opposite results for our two independent variables.Additionally,we conduct a series of bivariateprobit models to test whether the potential endogenous relationshipbetween democracyand party competition/inclusion significantly influences our analysis.12These models also allow us to address the potential
12

1I We also run the model without lagging the independent variablesand

find no differencein the results.

A bivariate probit model allows us to estimate two probit models whose error terms are correlated through the use of the bivariate normal distribution (Greene 1997). This method allows us to determine if a model predictingdemocracyand partiesare relatedand if this relationship influences the effect of party competition/inclusion on whether a state is a democracyor not. The firstprobit model is whether a state is a democracy or not using the various definitions previously discussed. These models also incorporate the independent variables and some controls for temporal correlation.The second probit model has a dependent variable which is a dichotomous measure of either

DEMOCRATIC PROGRESS AND REGRESS

557

TABLE1 COX AND WEIBULL ANALYSIS TRANSITION DEMOCRACY OF TO

Model 1 Polit. Regime Change Cox LogGDPC


Growth Islam Catholicism ELFI Past Dem Experience Regional Democracy Party Inclusion Party Competition Constant Rho

Model 2 Polit Regime Change Weibull .438* (.238)


-.513 -.032 -.040 1.15* .042 2.77* -.245 .865*** -10.14*** 1.03 (2.02) (.390) (.431) (.690) (.029) (1.18) (.243) (.172) (1.95) (.107)

Model 3 ACLP Cox .236 (.359)


-4.71* .519 .634 -.256 .055* 4.83*** -.201 .586**

Model 4 ACLP Weibull .157 (.362)


(2.19) (.535) (.710) (.880) (.026) (.962) (.314) (.214) (2.74) (.157)

Model 5 Polity Cox -.339 (.388)


-.117(2.76) -.323 (.534) .910 (.635) .727 (1.03) .094 (.114) 3.45* (1.70) .190 (.308) 1.38*** (.338) N = 3347 Subjects = 126 Fail = 28 LL = -70.18 Chi2= 64.27

Model 6 Polity Weibull -.338 (.417)


-.840 -.422 .968 .435 .105 3.80** .245 1.48*** -10.60 1.77* (2.96) (.600) (.716) (1.03) (.106) (1.46) (.301) (.270) (3.05) (.394)

.588** (.225)
.285 -.017 -.041 1.23* .051** 1.83 -.213 .923*** (2.04) (.349) (.426) (.704) (.026) (1.33) (.269) (.188) -

-4.20* (2.57) .548 (.504) .590 (.727) -.012 (.853) .050* (.024) 4.79*** (.941) -.425 (.320) .654*** (.236) -9.16*** 1.10

N = 2365 Subjects = 109 Fail = 45 LL = -155.26 Chi2 = 56.81

N = 2365 Subjects = 109 Fail = 45 LL = -83.80 Chi2 = 95.70

N = 2277 Subjects = 118 Fail = 35 LL = -114.38 Chi2 = 92.56

N = 2277 Subjects = 118 Fail = 35 LL = -64.07 Chi2 = 102.18

N = 3347 Subjects = 126 Fail = 28 LL = -10.78 Chi2 = 186.77

Note: p < .05, ** p < .01, ***p < .001. All Significancetests are one tailed. Robust StandardErrorsare reportedin parentheses.DV is durationof AuthoritarianGovernment.Failureevent is democracy.Coefficientsare interpretedas affectingthe hazard rate. So a negative sign decreasesthe hazard rate, thus increasingthe survivaltime.

problem of whether the relationshipbetween party competition/inclusionand democracyis spuriousand based on the effects of other indicators.
EMPIRICAL RESULTS

We begin by reviewingthe effectsof partieson transitions to and survivalof democracies.We then turn to the examination of other independentfactorspurportedto play a role in the process. Table 1 displays the results of the Cox and Weibullmodels for the durationof authoritarian states. The failureevents for the models in this tableare the transitionto The democracy. coefficientsarenot hazardratios,ratherthey assess the impact of the independentvariableson the hazard rate.A positive sign indicatesthat increasingthe value of the independentvariableleads to an increasein the hazardrate and a decrease in the survival time, while a negative sign indicatesa decreasein the hazardrateand an increasein the survivaltime. Thus positive coefficientsrevealthat increased values for the independent variables lead non-democratic states to become democraciesquicker,while negativecoefficients show that highervalues for the independentvariables increasethe durationof non-democraticregimes.
partycompetition or inclusion coded as 1 when the originalpartycompetition has a value of 4 or 5, and 0 if otherwise. For the party inclusion dependent variable,it is coded as 1 when the originalis 3 and 0 if otherwise. The independent variables are the economic and social measures.

Acrossall the models, the partycompetitioncoefficientis significantand positive as hypothesized:the more competitive the party system within an autocraticstate, the greater the hazard rate and the shorter the survival time of the autocracy, providingempiricalevidence for hypothesis 2a. It that competitive parties do indeed hasten the appears decline of autocraticregimes, most likely by reducing the monopoly of power held by the ruling regime and by limiting the ability to suppress attemptsat democratization. Chile in the late 1980s and Uruguay at the end of the 1970s provide appropriateexamples of this experience. In both cases, military governments allowed parties to compete in plebiscites,with the prospects of furtherdemocratization or military rule at stake. Parties took advantageof increasedpolitical space to mobilize the public against the in militaryregimes. Even with irregularities Chile, in those two nations, the military government lost the plebiscite, a shock in both countries (Scully 1995; Gonzalez 1995). These exampleshighlight the abilitiesof competitiveparties to mobilize the public against imposing military governments and furtherthe democraticprocess in spite of overwhelming odds. While party competition appears to help authoritarian regimes come to an end, there is no empiricalsupport that inclusion of partieswill help authoritarian statestransitionto The coefficientis only correctlysigned in democracy(Hla). two of the six models and is never statisticallysignificant. The inclusivenessof political partieshas less effect than the competitivenessof parties.One possible reasonfor this result

558
TABLE 2 COX AND WEIBULL ANALYSIS SURVIVAL DEMOCRACIES OF OF

POLITICAL RESEARCH QUARTERLY

Model 1 Polit. Regime Change Cox LogGDPC


Growth Islam Catholicism ELFI President Past Dem Experience

Model 2 Polit Regime Change Weibull -.431 (.498)


-.787 .908 .950* 1.40 -.545 -.091* -11.52*** -.689*** -.204 1.14 2.34*

Model 3 ACLP Cox -.827** (.270)

Model 4 ACLP Weibull -.928** (.335)

Model 5 Polity Cox .422 (.339)

Model 6 Polity Weibull .287 (.273)

-.001 (.589)
(2.79) (1.04) (.574) (1.13) (.927) (.042) (3.53) (.283) (.220) -

-2.53 1.26 1.41* 2.27* -.500 -.045 -12.35*** Regional Democracy -.749** Party Inclusion -.413* Party Competition Constant Rho

-6.57* (2.89) -6.30* (2.78) 2.89 (2.45) -.780 (.531) 1.89* -.818 (.573) (1.02) -.916 (.625) -1.05* (.603) .764 (.545) -.754 (1.01) -.909 -.733 (.871) (1.03) 1.38*** (.413) -.783* 1.25*** (.409) (1.04) -.033 (.035) .054 (.035) -.099* (.047) 1.51 (3.38) -3.95*** (1.08) -4.16*** (1.12) (.219) -.699** (.305) -.788** (.315) -1.62*** -.253+ (.158) -.904*** -.260* (.144) (.203) 7.60** (2.99) (3.47) 1.62* (.266) (.422) N = 1500 Subjects = 90 Fail = 36 N = 1500 Subjects = 90 Fail = 36

2.29 (1.69) (2.37) .920 (.711) (.811) .533 (.580) (.760) -1.15* (.697) (.886) -.512 (.522) (.463) (.047) -.136*** (.039) 1.56 (1.86) (1.29) (.286) -1.71*** (.348) (.219) -.936*** (.283) 4.29** (1.72) 2.66* (.321) N = 1557 Subjects = 87 Fail = 19

N = 600 Subjects = 53 Fail = 24

N = 600 Subjects = 53 Fail = 24

N = 1557 Subjects = 87 Fail = 19

LL= -51.79
Chi2 = 53.67

LL= -23.66
Chi2 = 73.17

LL= -110.59
Chi2 = 112.39

LL=-53.81
Chi2 = 140.36

LL=-29.99
Chi2 = 170.74

LL=-5.67
Chi2 = 158.02

Note: *p < .05, ** p < .01, ***p < .001, + p < .1. All Significancetests are one tailed. RobustStandardErrorsare reportedin parentheses.DV is durationof DemocraticGovernment.Failureevent is non-democracy.Coefficientsare interpretedas affectingthe hazardrate. So a negativesign decreasesthe hazardrate, thus increasingthe survivaltime.

is that the two variablesare highly correlated.Dependingon which datais used, the correlation between the two variables ranges from .53 to .71. To furtherexaminationof this relationship, we test the models without the PartyCompetition variable.The PartyInclusivenessvariableis only significant for the Polity data, though the change in the log likelihood indicates that Party Competition cannot be excluded from the model. One explanationfor this finding is that authoritarianregimesthat do not exclude partiesmay not allow parties to fully compete, limitingtheirabilityto influencea transition to democracy When parties are actually able to compete, they can influence a transitionto democracy Table 2 presents the results of the Cox and Weibull models on the survival of democracies. The failure events for the models in this table are reversionsto authoritarianism. Similar to Table 1, there is empirical, though not as consistent, support for our hypothesis that party competition helps nations avoid reversions to authoritarianism (H2b). Competition between political parties significantly decreasesthe hazardrate and increasesthe survivaltime of democraticgovernmentsin five of the six models. Competitive politicalpartiesin democraciescan provide a check on dominant political partiesensuring a functionaldemocratic process. The effect of competitive political parties is just slightly less robust for the survivalof democraciesthan for the transitionto democracy as the party competition variable is not significantfor one model (the Weibull based on the PoliticalRegimechange data).

Unlike earlier findings (Table 1), the coefficient for the party inclusion variableis significant and in the hypothesized direction.The more inclusive a democracyis towards politicalparties,the greaterthe survivaltime of that democracy (Hlb). This variableis statisticallysignificantacross all the models. Inclusionof politicalpartiesallows democracies to legitimize their political systems and facilitateadaptation and compromisebetween competinggroups.The resultsfor the PoliticalRegime Change data indicate that these results arerobustacrossdemocraciesoutside of Europeand the U.S. The role of partiesin Argentinain the 20th century provides a good example of the importance of inclusive and competitive parties. Prior to 1983, Argentina flirted with democracyseveraltimes afterWorldWarII, with each nascent democracy falling to military rule. One difference between the transition in 1983 and past democratic experiences is the role of the respective political parties. Earlier attempts at democracy were often marked by the banning of parties either by Peron and his party or against Peron and his party (McGuire 1995). While other factors contributed to the downfall of democracyin Argentinaduring this period, the lack of inclusive and competitive parties helped ensure that social discontent and state and military reactions to it were manifested in less than democratic ways. However, following the collapse of the militaryjunta in 1983, the majorpartiesdid not follow the historicalpattern of party exclusion but rather attempted to create an inclusive and competitive party system. While Argentina's

DEMOCRATIC PROGRESS AND REGRESS DEMOCRATIC PROGRESS AND REGRESS

559 559

democracyhas sufferedsome difficulties,in 1989 it experienced the first executive transferof power by election from one party to another. Altogether, across both tables, we find that parties do contribute to the breakdown of authoritarian governments and the survival of democracies. These results differ from previous studies that found no independent or a negative effect for parties (Linz 1978; Power and Gasiorowski1997; Bernhard, Nordstrom,and Reenock2001). Again,we maintain that studies on the negative effect of parties tended to focus on multipartyism extremistparties,while those that or found no effect focused more on the numbers of parties, while we concentrateon the legitimacyof parties and their ability to actually compete. While some previous studies found that multipartyismcan have a detrimentaleffect, our work demonstratesthat parties can actually have a beneficial effect on democratization. While Tables1 and 2 demonstratethat partycompetition and inclusion play importantroles in the transitionto and survivalof democracies,it is possible that these findings are spurious and drivenby other variableslike economic development and/or that the causal relationship is the reverse (stable democracieslead to competitive and inclusive parties). As discussed in the research design, we believe that lagging the independent variablesshould help alleviateany threatsfrom the latterproblem. We also conduct additional statisticaltests to check for this possibility. The results of our bivariateprobits indicate that we do not have the problemof anothervariablecreatinga spurious relationship between party competition and democratic transitionand survival.In all the models, the party competition variableis consistentlysignificant,even aftercontrolling for correlationsin the effect of an explanatoryvariable on whether a state is a non-democracy or democracy and party competition.13 These results also reveal that while there is a significantcorrelationacross a few of the models, the effect of the party competition variableis always significant. Had the causal arrow simply been reversed (with strongerdemocraciesleading to more competitive parties), the results for party competition variablewould not have been statisticallysignificantin these bivariateprobits.14 We do not infer from these results that party competition does not benefit from stable democracies.Still, partycompetition is not solely a byproduct of a stable democracy and party competition itself influences the transitionto a democracy and the survivalof a democracy
Thereare six models (three datasets,and two differentpartymeasures) available at http://myweb.uiowa.edu/bhlai/data. we are interested As first and foremostin the effects of party competition and the inclusion on democratictransitionand survival,we do not discuss all results.The models are analyzedsimply to confirm the robustnessof our findings. 14 Similar to a two stage least squares, a bivariateprobit in which the dependent variable of one of the models (party competition or party inclusion) is used as an independent variable in the other model allows us to examine the potential endo(democracy/non-democracy) geneity of the two variablesby correctingfor co-variationin the error terms of the two probit models.
13

While the results of our additionalanalysisindicate that party competition is statisticallysignificant even after we control for a possible reciprocal relationship, the party inclusion variable is only significant in the ACLP model when we control for a possible reciprocalrelationshipwith a state'slevel of democracy.This finding comports with earlier results (Table 1): party inclusion does not significantly influence the transition to democracies. It does, however, weaken support for the effect of party inclusion on democraticsurvival(Hlb). Because we do not expect party inclusion to influence autocraticgovernments,we re-examineour models, looking at only cases of democracyand their initial failure(e.g., if a countryis a democracyfrom 1950-1960 and a non-democracy from 1961-1992, only the 1950-1960 period and the 1961 year are included in this analysis)." In so doing, we find that for the Polity data,thereis a significantrelationship between partyinclusion and democraticsurvival,even after controllingfor possible correlationsacross the two models. However,the partyinclusion variableis still insignificantfor the PoliticalRegime Change data. In short, we find mixed support for the effect of party inclusion on the survival of democracies.We are confidentthat it significantly influences democraticsurvivalas defined by the ACLPdata and Polity data.With the PoliticalRegimeChangedata,it is less clearif it directlyinfluences the survivalof a democracy Table3 displays the predictedmedian survivaltimes for authoritariangovernments (Table 1) and democratic governments (Table 2). The median survival times are presented to demonstratethe substantiveeffect of the statistically significant party variables on the longevity of authoritarian democraticgovernments.The effectof the and percent of states that are democraciesin the region is also presented for comparison as this variableis almost consistently significantacrossall the models. The median survival times are calculated by holding all the variables at their mean (except Islam, Catholicismand President,which are held at 0), while varyingthe variablesof interest.Lookingat Table 3, party competition has a substantively significant impact on the predicted survivaltime of authoritarian governments. Going from Restricted (1) to Suppressed (2) politicalpartiesreduces the survivaltime of an authoritarian government by more than half (86 years to 38 and 327 years to 146) for the Political Regime Change and Polity data and a little under 50 percent for the ACLPdata. Compared to the effects of being in a region comprised of democracies, parties have a substantialeffect. An authoritarian government that suppresses political parties has roughly the same survival time as if it were located in a region that is 25-50 percent democratic. While party competition has a dramatic effect on the survivaltime of authoritarian governments,its effect is less

15

The first set of tests was designed to predict the regime in any given year. This analysisis similar to the hazardmodels in that it is trying to model failure(Beck, Katz,and Tucker1998).

560 560
TABLE 3

POLITICAL RESEARCH QUARTERLY

EFFECTS SIGNIFICANT OF INDEPENDENT VARIABLES PREDICTED MEDIANSURVIVAL TIMEFORAUTHORITARIAN ON AND DEMOCRATIC GOVERNMENTS

Authoritarian Governments Independent Variables RegionalDemocracy Independent Variable Values 0 .25 .5 .75 1 0
1 2 3 4 5 Party Inclusion 0 1 2 3

DemocraticGovernments Political Regime Change 8 24 77 240 754 *


* * * * * 15 27 36

Political Regime Change 80 46 26 15 9 195


86 38 17 7 3 * * * *

ACLP 398 131 43 14 5 335


178 95 51 27 14 * * * *

Polity 336 203 123 74 45 734


327 146 65 29 13 * *20 *

ACLP 17 30 53 93 164 32
38 45 53 64 76 24 36 53 78

Polity * * * * * 16
23 35 51 75 111 15 30 59 114

PartyCompetition

Note: Valuesare PredictedMedian SurvivalTime, holding all other independent variablesat their mean, except for Islam and President(which were both held at 0). Estimatesare based on the Weibull models in Tables2 and 3 * Variable was not statisticallysignificantin the model. PredictionEstimatesare calculatedusing STATA 7.0

dramatic for the survival of democratic governments. It requiresgoing from Restricted(1) political parties to Competitive (5) political parties to double the survival time for democraticgovernmentsin the ACLPdata. The effect of the percentageof democraciesin the region is much greater,as going from 0 to 50 percent democracies in a region increasesthe democraticsurvivaltime by over 200 percent in the ACLPdata. Partycompetitionhas a substantiveeffect on democraticsurvival,althoughnot as greatas its effect on the survival time of autocracies.Partyinclusion also has a substantive effect on the survival time of democracies. Going fromno partiesallowed (0) to no exclusion of parties (3) yields a 140 percent increase in the survival time of democracies for the Political Regime Change data, a 225 percent increase for the ACLP data, and a 660 percent increasefor the Polity data. In all these examinations, we measure democracy in a dichotomous fashion.16 To confirm the robustness of our findings, we also evaluate the effects of party competition and inclusion on continuous measures of democracy We use two common continuous measures,Polity and VanHannen'sPolyarchydata (Casperand Tufis2002). For Polity,we use as a dependent variable, the continuous -10 (highly
16

autocratic) to 10 (highly democratic) scale and for Polyarchywe use the provided index of democracywhich theoretically runs from 0 (highly autocratic) to 100 (highly democratic).We use the same independentvariables,but do not include past democratic experiences and measures of Presidential/Parliamentary systems. We exclude the latteras it mostly measuresdemocraticstates and is missing for severalnon-democraticstates. Insteadof past democraticexperiences, we include a lag of the dependent variableto deal with problems of autocorrelation. For both models, the partyinclusion and party competiare tion variables positiveand statistically significant,indicating that each contributesto higher levels of democracyfor states. These results are robust across a varietyof specifications and provideus with confidenceaboutthe importanceof These findingsprovideus partyinclusion and competition.l7

17 These resultsare availableat http://myweb.uiowa.edu/bhlai/data. do We

Collier and Adcock (1999) provide a good overview of the debate on the usage of continuous and dichotomous measuresof democracy

not include them for the sake of space and because they do not change our primaryfindings that partyinclusion and competition significantly influence the transitionto and survival of democracies.The results of other independent variables are relatively consistent with already reported findings. In addition, we tested these models against a seemingly unrelatedregressionwhich included two models or three models (Measure of Democracy, Party Inclusion, and Party Competition as Dependent variables)and a three stage least squareswhere partyinclusion and party competition were instrumentedvariables.While we did

AND REGRESS DEMOCRATIC PROGRESS DEMOCRATIC PROGRESS AND REGRESS

561 The results for the economic variableson the survivalof democraciesmodels (Table 2) are similar. The GDP/capita variableis significantfor the ACLPdata, comparableto the findings of Przeworskiet al (2000). The GDP/capitavariable is not significant for other measures of democracy (similar to Reiter 2001 and distinct from Gasiorowski 1995), but when the models for the Political Regime Change and Polity data are run without the partyvariables, the GDP/capitavariablesare statisticallysignificant and in the right direction. However, the log likelihood values reveal that the party variableshave a significantimpact on the model. In addition, the correlationbetween party variables and the GDP/capitais low (less than .1), indicating that parties play an important role in sustaining democracies. The growth variable is also significant only with the ACLPdata. As for the social indicators, looking at both tables, the Islam variableis significantin only one model (Polity dataCox Model). The results are mixed for prior studies using the PoliticalRegimeChangedata:one study found it significant in preventing a democratictransition (Feng and Zak 1999) while another found no effect (Gasiorowski 1995). While the Catholicvariableis not significantin Table 1, it is significantin Table 2 for the Political Regime Change data and one model for the ACLPdata, though its sign varies based on the measureof democracy.Looking at model 3 of Table 2, the Catholic variable is significant and negative indicating that democracieswith major Catholic subpopulations have longer durations. However, the effect is opposite for the PoliticalRegime Change Dataset (models 1 and 2). In these models, the Catholic variable is positive and significant, indicating that Catholic democracies have shorter durations. These results vary from that of Gasiorowski (1995) who finds no significant effect for Catholicism, though they provide further evidence for Bollen (1979). The change in sign is likely to be due to differencesin the ways in which states are classifiedas democracies across the three datasets. Finally,the ethno-linguistic fractionalizationvariable is only significant in both tables for the Political Regime Change data. However, in both tables, it is positive and significant for the Political Regime change data, indicating that higher fractionalizationreduces the duration of nondemocracies and democracies, a finding analogous to that of Gasiorowski (1995), although he does not find this variable to be significant. We also find this variable to be significant but negative for the Polity data. Once again, this difference in sign is likely due to the difference in states covered. The regional democracyvariableis significant across all the models in Table 1 except one, consistent with most studies on the transitionto democracy(Gasiorwoski 1995; Reiter2001). In Table2, the regionaldemocracyvariableis significantfor all models except the Polity data:this is consistent with other studies revealing mixed results with some finding support (Gasiorowski1995; Przeworskiet al 2000; Bernhard, Nordstrom, and Reenock 2001) and

with confidencethat partycompetitionand inclusion signifito cantlyinfluencethe transition and survivalof democracies. As for the controlvariables,theireffectsvaryacrossdifferent measuresof democracyWe begin with a thoroughreview of the effectsof economic factors,given theiremphasisin the variableis sigliterature. Lookingat Table1, the GDP/capita is consistentwith nificant for the Gasioriowskidata, which other studies (Gasiorowski1995). However,the GDP/capita variableis not significantfor the ACLPor Polity data,which differsfrompreviousstudies (Reiter2001; Przeworski 2000). One explanationfor this differenceis the inclusion of the partyvariablesand differentmethods of analysis.18 One explanation(as discussed earlier)may be that ecoacts nomic developmentas measuredthroughGDP/capita to promotecompetitivepoliticalpartieswhile also leadingstates found to transition Thus,the relationship awayfromautocracy. to betweencompetitive politicalpartiesand transitions democbasedon ecoracy(Table1) maybe spuriousand alternatively nomic development.Thereare a few reasonswhy we do not think this is the case. First,the impactof partysystemsis sigand for nificantevenwhile controlling GDP/capita, the log likelihoods of the models indicatethatthe partyvariables provide a statistically to contribution the model fit. Second, significant it evenif GDP/capita to supportpartycompetition, is likely acts The to have two effectson democratization. firstis the direct relationshipthrough economic development.The second is indirectthrough the fosteringof competitiveparty systems, in which procedurally help to bring about democracy a state. Economicdevelopment influenceboth partycompetition may and democratization this does not necessarily but negate the betweenpartycompetitionand democratization. relationship Finally,we examine the relationshipbetween partycompetition, GDP/capita and democratization by modeling party competition and democracy simultaneously using bivariateprobit models. We find that the choice of party competition and democracy are interrelated (as we expected) and that GDP/capita significantlyinfluences party but does not significantlyinfluence democraticompetition zation. Party competition is still significant even when we control for the possible relationshipacross the two models, leading us to conclude that while GDP/capita may act through party competition, party competition also has an independenteffectas indicatedby the significantcorrelation between the two equations.19

find significantcorrelationsacross the models, the effects of the party inclusion and competition variables on the level of democracy were unchanged. 18 Reanalyzing the Polity data and the ACLPdata without the party variables results in the GDP/capita variableachievingstatisticalsignificance in all the models (though for the Polity model only at the .1 level). The log likelihoods of the models indicate that the partyvariablesprovide a statisticallysignificantcontributionto the model fit. For the ACLPdata, the GDP/capitavariable is significant when analyzed using a probit model similarto the ACLPstudy. 19 Becauseour main resultsdo not change,the bivariate probitsarenot presented here but are availableat the following website: http://myweb. uiowa.edu/bhlai/data.html.

562 562 others none (Reiter2001; Pevehouse 2002). The past democratic experience coefficient is significant and in the correctly signed direction in Table 1 except for the Polity models and one of the Political Regime Change data models. In the Polity models, this variable does become significant when the party variables are excluded, though the log likelihoods indicate that party variableshave a significant effect on the model. In Table 2, the past attempts at democracyvariableis only significantfor the Polity data and one of the Political Regime Change data models, demonstrating weak empirical support for this variable, consistent with other studies (Bernhard,Nordstrom, and Reenock2001). Finally,in Table2, the PresidentialDemocracy variableis positive and significantin the ACLPmodels and one Polity model, similar to the findings of Przeworski et al. (1996) and distinct from Gasiorowski (1995), who finds this variable to be insignificant for the Political Regime Change data. Overall, the results of the control variables are similar to that of other studies, providing some confidence in our evaluation. Differencesin results across the three differentmeasures of democracyare likely to be due to two factors.First, the Polity and ACLPdata cover a differentset of states than the Political Regime change data, which focuses primarilyon developing countries. Second, as discussed previously,all three measures differ on when a state is a democracy and when states transitionto and away frombeing a democracy. Finally,to be certain of our findings, we reranour analysis of the ACLPand Polity data using the set of countries analyzed by the Political Regime Change data and the results did not change significantly.
DISCUSSION AND CONCLUSION

RESEARCH POLITICAL RESEARCH QUARTERLY POLITICAL QUARTERLY

The results of this article provide empirical evidence for the argument that political parties play an important role in facilitating the transition to and the survival of democracy,evidenced across all three measures of democracy. Conventional wisdom is correct; parties do matter. The inclusion of political parties within the political system and the existence of competitive parties can help authoritarianstates to transition to democracy and facilitate the survival of democracies. Similar to other scholarship, we do find that democratic neighbors increase the likelihood that authoritarian states will transition to democracy and that democracies will remain democracies. The economic growth variable demonstrated mixed results and tended to indicate that economic growth benefits whatever regime is in power. Similarly,control variables such as Islamic or Catholic states, presidential versus parliamentarysystems and past democratic experience had only weak and inconsistent support, varying considerably across models. Our results are robust across a variety of measures and methods. Altogether,based on the findings, it is clear that competitive parties contributeto the demise of autocraticregimes, as they chip away at the power of the ruling regime and

broaden political space. Party competition plays a significant and crucial role in the decline of authoritarianstates and the transitionto democracy Only regionaldemocracyis as consistently significant as party competition in the process. Partyinclusion by itself does not lead to a transition to democracy.While the inclusion of parties is obviously a requirementfor the competitionthereof(with which inclusion is highly correlated),inclusion, while necessary,is not sufficient:restrictedand/or hollow partiesare unable to and facilitatea transisufficientlyweaken authoritarianism tion to democracy.When parties are able to actually compete in authoritarian regimes,however,and even when they are partiallysuppressed, they can dramaticallyreduce the survivalof an authoritarian regime and influence its transition to democracy. For democraticregimes to survive, however, the simple inclusion of political partiesis essential.For democraciesto endure as such, parties must be allowed to continue to flourish. Having a democratic neighbor is not something easily manipulatedby governments.Allowing political parties to exist and freelycompete, however, are measuresthat governments can more readily undertake to move away and toward fuller and more stable from authoritarianism democracy.Advocates of democracy can help facilitateits realizationand survivalby ensuring the inclusion and competition of political parties. Clearly some nations have chosen to avoid the establishmentof political partiesdue to fears of fractionalization,distrust, and instability.Yet-as the data so clearlyshows-when parties are excluded risks to the qualityand sustainabilityof democracyare significant and substantial. Our importantfindings reinforcesome of the prevailing conclusions regarding international, economic, institutional, and culturalfactorsand their functions in democratization, but counter others by highlighting the essential role that parties play in the process of transition. This empirical examination utilizes fine-grained, appropriate methods and data to evaluatequestions of democratictransitions and survivalin a mannernot yet undertaken.While admittedly some details are lost in the use of large crossnationaldata, the quantitativefindings of this study are significantly robust, generalizable,and fundamentallynecessary to the study of parties and democratization. This researchcan help inform the debate about democratization by focusing discussion on the importance of studying a range of cases, and the independent effects of partiesin the democratization process.
REFERENCES Aldrich, John H. 1995. Why Parties? The Origin and Transformation of Political Parties in America. Chicago: University of Chicago

Press. Beck,Nathaniel,JonathanKatz,and RichardTucker.1998. "Taking Time Seriously: Time-Series Cross-Section Analysis with a
Binary Dependent Variable."AmericanJournal of Political Science

42 (4): 1260-88.

AND REGRESS DEMOCRATIC PROGRESS

563 563 Democracy's Jaggers,Keith,and Ted RobertGurr.1995. "Tracking ThirdWave with the Polity III Data." Journalof PeaceResearch. 32: 469-82. Adrian. 2002. "The 2001 Freedom House Survey: Karatnycky, Muslim Countriesand the DemocracyGap." Journalof Democracy 13: 99-112 Lancaster, John. 2004. "FeelGood Politics BolsterIndia'sVajpayee Post Ahead of Vote."Washington (April 19): A13. In Linz, Juan J. 1978. "Crisis,Breakdown,and Reequalibration." of Juan Linz and Alfred Stepan, eds., The Breakdown Democratic Regimes. Baltimore, MD: Johns Hopkins University Press. Linz, Juan, and Alfred Stepan. 1996. "Toward Consolidated 7: Democracies." Journalof Democracy 14-33. Man:SocialBasesof Politics. Lipset,SeymourMartin.1960. Political GardenCity,NY: McGuire,James W 1995. "PoliticalParties and Democracy in Argentina."In Scott Mainwaring and Timothy Scully, eds., Institutions. Democratic Stanford,CA: StanfordUniverBuilding sity Press. Mishler, William, and Richard Rose. 1997. "Trust,Distrust and Skepticism:Popular Evaluationsof Civil and Political Institutions in Post-Communist Societies."Journal of Politics 59.2 (May):418-51. Muller, Edward N., and Mitchell A. Seligson. 1994. "Civic Culture and Democracy: The Question of Causal Relationships." AmericanPolitical Science Review 88.3 (September):
635-52.

Bermeo, Nancy. 1990. "Review Article: Rethinking Regime


Change." Comparative Politics April: 359-77.

Bernhard, Michael, Timothy Nordstrom, and Christopher InstitutionalIntermeReenock.2001. "Economic Performance, and Democratic Survival." diation, Journalof Politics63: 775803. Bogaards,Matthijs. 2000. "CraftingCompetitive Party Systems: ElectoralLaws and the Opposition in Africa."Democratization 7.4 (Winter):163-90. Bollen, KennethA. 1979. "Political Democracyand the Timing of
Development." American Sociological Review 44: 572-87.

Bollen, Kenneth A., and Robert W Jackman, 1985. "Political Democracy and the Size Distribution of Income." American
Sociological Review 50.4 (August): 438-57.

Versus Casper, Gretchen, and Claudiu Tufis. 2002. "Correlation The Interchangeability: LimitedRobustnessof EmpiricalFindPolitical ings on DemocracyUsing Highly CorrelatedDatasets."
Analysis 11: 1-11.

Collier, David, and Robert Adcock. 1999. "Democracy and Dichotomies: A PragmaticApproach to Choices about Concepts." Annual Review of Political Science. Palo Alto, CA: Annual

ReviewsPress:53765. Coppedge, Michael. 1993. "Partiesand Society in Mexico and Politics Venezuela: Why Competition Matters."Comparative 25.3 (April):253-74. Cross National Time Series Data Archive. 1997. www.databanks.sitehosting.net
Dahl, Robert A. 1971. Polyarchy: Participation and Opposition. New

Haven, CT:YaleUniversityPress. Diamond, Larry, Juan J. Linz, and Seymour Martin Lipset, eds. LatinAmerica.Boul1989. Democracy Developing in Countries: der, CO: LynneRienner. Economist. 2003. "Party time?"366.8313: 45 (March1). Ellingsen, Tanja.2000. "ColorfulCommunity or Ethnic Witches' 44.2: 228-49. Brew?" Resolution Journalof Conflict and Dictatorshipin Interwar Ertman,Thomas. 1993. "Democracy World Politics50 (3): 475-505. EuropeRevisited." Feng, Yi, and Paul Zak. 1999. "The Determinantsof Democratic 43: Transitions." Resolution 162-77. Journalof Conflict Gasiorowski,MarkJ. 1995. "EconomicCrisisand PoliticalRegime American PoliticalScience Change:An Event History Analysis." Review.89:882-897.
_

Pevehouse, Jon C. 2002. "With a Little Help from My Friends? RegionalOrganizationsand the Consolidationof Democracy." Science.46: 611-26. American Journalof Political Daniel. 2004. "TheCatholicWave." Journalof Democracy Philpott, 15: 32-46. ParticiDemocracies: Powell Jr., G. Bingham. 1982. Contemporary MA:HarvardUniverand Violence. Cambridge, pation,Stability sity Press. Power, Timothy, and Mark J. Gasiorowski. 1997. "Institutional Design and Democratic Consolidation in the Third World." Political Studies30: 123-55. Comparative Przeworski,Adam, Michael Alzarez,Jose Antonio Cheibub, and FernandoLimongi. 1996. "WhatMakesDemocraciesEndure?" 7: Journalof Democracy 39-55.
. 2000. Democracy and Development: Political Institutions and

. 1996. "An Overview of the Political Regime Change

Dataset."Comparative Political Studies29: 469-83. Gasiorowski,MarkJ.,and TimothyJ.Power. 1998. "TheStructural PolitDeterminantsof DemocraticConsolidation." Comparative ical Studies31: 740-71. Tradeand GDP Data." JourGleditsch, Kristian.2002. "Expanded nal of Conflict Resolution 712-24. 46: and Gonzalez,LuisE. 1995. "Continuity Changein the Uruguayan PartySystem."In Scott Mainwaringand Timothy Scully, eds., Institution. Democratic Stanford,CA: StanfordUniverBuilding Press. sity Greene, William H. 1997. Economic Analysis.New York. Prentice Hall. Henderson, ErrolA. 1998. "The Democratic Peace Through the 42: Lens of Culture, 1820-1989." International StudiesQuarterly 461-84. Societies. Huntington, Samuel. 1968. PoliticalOrderin Changing New Haven, CT:YaleUniversityPress.
_

. 1996. "Democracy for the Long Haul."Journal of Democracy

7: 3-13.

1950-1990. Cambridge,MA:Cambridge WellBeingin theWorld, UniversityPress. Science and the Crisis of AuthoritariPye, Lucien. 1990. "Political Political anism."American ScienceReview84: 3-19. Reiter, Dan. 2001. "Does Peace Nurture Democracy?" Journalof 63: Resolution. 935-48. Conflict Fractionalization(ELF) Roeder, Philip G. 2001. "Ethnolinguistic Indices, 1961 and 1985." February 16. <http//:weber.ucsd. edu\-proeder\elf.htm>. Rudolph, Susanne Hoeber, and Lloyd I. Rudolph. 2002. "New 13: Dimensionsof IndianDemocracy."Journal Democracy 52of 66. Peace:DemocRussett,Bruce,andJohn O'Neal.2001. Triangulating New York: ad International Organization. racy,Interdependence, Norton. Sartori,Giovanni. 1976. Partiesand PartySystems: Aframeworkfor analysis.Cambridge,MA:CambridgeUniversityPress. Schmitter,Philippe,and TerryLynnKarl.1991. "WhatDemocracy 2: Is ... and Is Not."Journalof Democracy 75-88.

564 Scully,TimothyR. 1995. "Reconstituting PartyPoliticsin Chile."In Scott Mainwaring and Timothy Scully,eds., Building Democratic Institutions. Stanford,CA:StanfordUniversityPress. Stepan, Alfred C., and Graeme B. Robertson. 2003. "An 'Arab' Morethan a MuslimDemocracyGap." 14: Journalof Democracy 30-44.
Ware, Alan. Political Parties and Party Systems. Oxford: Oxford Uni-

POLITICAL RESEARCH QUARTERLY RobertDudley,and TimothyS. Shah. 2004. "ThePioWoodberry, 15: Journalof Democracy 47-61. neering Protestants." Received: July 22, 2004 Accepted for Publication: January 3, 2005 brian-lai@uiowa.edu ruth.melkonian@gordon. edu

versity Press, 1996.

You might also like