You are on page 1of 28

PA RT Y P O L I T I C S Copyright © 2003 SAGE Publications

V O L 9 . N o . 2 pp. 139–166 London Thousand Oaks New Delhi

An Empirical Measure and an Application to the Americas Mark P. Jones and Scott Mainwaring

Political parties and party systems exhibit widely varying degrees of nationalization, that is the extent to which a party receives similar levels of electoral support throughout the country. The level of party nationalization has a prominent effect on such important factors as the survival of democracy, the types of issues that dominate political competition, legislative behaviour and public policy. In spite of its importance, party nationalization has been neglected in the comparative politics literature. Our article makes two contributions. First, it provides a measure of party and party system nationalization, based on the Gini coefficient, that is superior for comparative analysis to those employed to date. Second, it utilizes these measures to analyse nationalization in 17 democracies in the Americas, the first time nationalization has been examined empirically outside the advanced industrial democracies. The measure underscores the widely varying degrees in nationalization across party systems, within party systems over time, across parties within countries and within parties over time.

KEY WORDS Ⅲ Latin America Ⅲ nationalization Ⅲ political parties

We address an under-analysed issue in the comparative study of parties and party systems: their degree of nationalization. In the massive literature on party systems, considerable attention has been paid to the extent of polarization (Knutsen, 1998; Sani and Sartori, 1983; Sartori, 1976), the number of parties (Lijphart, 1994; Sartori, 1976; Taagepera and Shugart, 1989) and the level of institutionalization (Harmel and Sväsand, 1993; Janda, 1980; Mainwaring and Scully, 1995). With the exception of some work on the United States, very little has been written on the degree of nationalization.

PA RT Y P O L I T I C S 9 ( 2 )

It is time to address this lacuna. Parties and party systems vary significantly in the degree to which they are nationalized. By a highly nationalized party system, we mean one in which the major parties’ respective vote shares do not differ much from one province to the next. In a weakly nationalized party system, the major parties’ vote shares vary widely across provinces. The same logic applies to parties: highly nationalized parties have a relatively even share of the vote across different geographic units, whereas weakly nationalized parties have widely varying shares across geographic units. Party system nationalization reflects an important component of the dynamics of party competition. Two systems may be similar in the number of parties and the degree of polarization, the two dimensions Sartori (1976) used to classify party systems in his classic work. Yet if the level of nationalization diverges sharply between these two systems, then the competitive dynamics are quite different. In a case of high nationalization, electoral competition follows a roughly similar pattern across the country’s sub-national units. In a case of low nationalization, the parties that fare well in some sub-national units are minor electoral competitors elsewhere. Our indicator of party system nationalization allows for precise measurement of such differences across countries. Moreover, two parties (either within the same system or not) may have the same share of the national vote, but their political and electoral strategies likely differ if one party wins a roughly similar share of the vote in all states while the other is dominant in a few states and a secondary force in the others. Our measure of party nationalization captures such differences. In this article, we make three contributions. First and foremost, we propose a new way to measure party system nationalization and party nationalization. For measuring party nationalization, we advocate inverting the Gini coefficient, a measurement of inequality across different units widely used in other contexts. Most previous measures of nationalization have significant shortcomings for comparative research. When clear and careful measurement of concepts is possible, as it is in this case, it forms a fundamental building-block of good social science. Our measures allow for a precise means of assessing nationalization and of comparing across parties and party systems and across time. In addition, they reveal interesting information about change over time within a given political system. If a country moves markedly toward a more nationalized system, or conversely toward greater interstate differences, these phenomena deserve careful attention (Brady, 1985; Claggett et al., 1984; Kawato, 1987; Stokes, 1967). Our measure of party nationalization allows us to compare parties over time, to see if parties’ electoral increases and declines occur relatively evenly throughout the country or are geographically concentrated, and to examine whether most parties in the same country follow a similar pattern or diverge markedly in terms of nationalization. Second, we provide data on the nationalization of party systems and 140

with potentially important consequences for how democratic systems function. Claggett et al. The empirical data in this article for the first time allow for careful and systematic comparisons of party system nationalization across countries outside Western Europe and the United States. Canada and the United States. the swing from one party to others). the vote shift from one election to the next rather than the absolute level was relevant. Rose and Urwin. 1967). Our work indicates that scholars of parties and party systems elsewhere neglect the topic of party and party system nationalization at their peril. Although variance in party nationalization is an important issue. 2000. Stokes (1965. our measures travel across the great majority of countries without difficulty. Finally. the concept of ‘nationalization’ has been used in two different ways.e. which is often not incorporated into comparative analyses of the advanced industrial democracies) variance across party systems has been considerably less pronounced than outside this select group of countries. Schattschneider (1960) and Sundquist (1973) utilized it to assess the extent to which subnational units approximate national vote patterns. in this case.1 We are aware of only two previous efforts to compare party or party system nationalization across different countries (Caramani. notwithstanding great variance in party system nationalization. Conceptual Issues and Measurement A first essential step is clarifying what we mean by ‘nationalization of parties and party systems’. Although this analysis is limited to 17 countries. 1975). Yet. (1999) examined how closely the direction of electoral change in subnational units approximated the national pattern of electoral change (i. In short. Brady (1985) and Bawn et al. In contrast. where (except for Canada. both previous efforts to compare party nationalization across different countries focused on the advanced industrial democracies. we call attention to an important phenomenon that has been badly neglected in the comparative literature on party systems. the absolute level of support for different parties across sub-national units is the relevant measure. with the exception of the United States there have been few attempts to compare across countries or time. One of our principal findings is that party system nationalization varies greatly across cases. the comparative literature on party systems has virtually neglected this issue. 141 . (1984) and Kawato (1987) used both meanings of ‘nationalization’ and clearly distinguished between them.2 Ironically.3 In their work.J O N E S A N D M A I N WA R I N G : T H E N AT I O N A L I Z AT I O N O F P A R T I E S parties in 15 Latin American countries. In the literature on the United States. this article furthers attempts to expand the frontiers of the comparative understanding of party systems with the hope that scholars will begin to pay more systematic attention to differences in party system nationalization and their consequences.

but the term ‘nationalization’ of parties or the party system should be reserved for the former concept. the Gini coefficient is technically superior to most existing alternatives. As used here. Both Stokes and Brady measured not the nationalization of the party system. we subtract the Gini coefficient from 1. but rather the nationalization of electoral trends (or swings). to the absolute levels of electoral support parties win across sub-national units). reflecting very different competitive situations. A party might have pronounced cross-state differences in programmatic character and social base and yet win the same share of the vote in all of the states.6 Our measure of party nationalization is based on variance across states. stated differently.15 in another. notwithstanding the internal differences across states.PA RT Y P O L I T I C S 9 ( 2 ) Both conceptions of nationalization are meaningful. 1998. departments. Two parties may have the same share of the vote but diverging levels of nationalization. however. the most commonly used measure of income inequality. the PNS can be used to compare parties within the same system or across systems. Second. Logically. Shryock et al. not to whether electoral swings are similar across districts. the Gini coefficient assesses the extent to which a party wins equal vote shares across all the sub-national units. The Gini coefficient is a widely used measure of inequalities across units (Creedy. Like any measure of nationalization that merits serious consideration for cross-national analysis. In addition to being widely known and used. and the PNS. it is. A Gini coefficient of 1 means that it received 100 percent of its vote in one subnational unit and 0 percent in all the rest. Of particular interest is verifying the relationship between electoral ascension or decline. Nor does it imply a party that is highly cohesive or disciplined at the national level. It ranges from 0 in cases of perfect equality across all units to 1 in cases of perfect inequality.5 The PNS has two primary purposes.. it is not country-specific. on the other. A nationalized party as 142 . A Gini coefficient of 0 signifies that a party received the same share of the vote in every sub-national unit. On our measure. provinces. for example. The nationalization of electoral swings might bear an empirical relationship to the nationalization of the party system (or. Thus our concept of a nationalized party does not imply one that is programmatically homogeneous across sub-national units. this party would be perfectly nationalized. We call this inverted Gini coefficient the Party Nationalization Score (PNS). it can trace changes in a party’s level of nationalization over time. Let us be clear about what this concept does not measure. First. the concept of party system nationalization should refer to the structure of the party system. administrative regions or parishes in parties’ electoral performance. a Gini coefficient of 0. 1976).15 in one country is comparable to a Gini coefficient of 0. To measure the nationalization of parties. These differences are likely to affect the parties’ electoral and congressional strategies. on the one hand.4 We subtract the Gini coefficient from 1 so that a high score indicates a high level of nationalization.

First. as 143 . a portrait at the national level may be meaningless in understanding sub-national dynamics.8 The effective number of parties is still a meaningful indicator for cross-national comparisons and for understanding some aspects of national-level dynamics. In the remainder of this section. where a party system possesses an intermediate or low level of nationalization. in less nationalized party systems. and to locate cases on an important dimension of party systems. The national-level data.J O N E S A N D M A I N WA R I N G : T H E N AT I O N A L I Z AT I O N O F P A R T I E S defined here could be a loosely knit organization of individual political entrepreneurs who are part of the same national organization. or alternatives such as effective magnitude and effective threshold) and of the national party system (the effective number of electoral parties). we also develop a measure of the nationalization of party systems. national factors may be more important in forging bonds between voters and parties. Schattschneider (1960) linked the nationalization of the party system to voters’ orientations. which are important in most weakly nationalized party systems. the use of these measures is perhaps appropriate. Our measures could also be useful for several other purposes in comparative political inquiry. sub-national factors may be more salient in creating such bonds. These hypotheses are not true by fiat. Conversely. He argued that in highly nationalized party systems.7 The contribution of every party to the PSNS is thus proportionate to its share of the vote. To create this measure we multiplied the nationalization score (PNS) for every party by its share of the national valid vote. In a weakly nationalized system. we note four hypotheses that political scientists have proposed about the effects of different levels of party system or party nationalization that could be more effectively tested through use of the PNS or PSNS. Some Uses of the Measures The PNS and PSNS allow researchers to measure nationalization precisely. the degree of party system nationalization is relevant for the most widely used measure of legislative electoral rules (average district magnitude. can mask major intra-country differences. however. to undertake more informed comparisons across countries and over time. national-level measures such as the average district magnitude and the effective number of electoral parties very likely suffer from validity problems. It is useful to have such a summary expression of the level of nationalization of the party system for the same reason that it is useful to have a summary expression such as the effective number of parties: it enables scholars to trace changes over time within the same system and to compare across countries. the nationalization of parties. and then summed this product for all the parties. For example. We call this weighted PNS the Party System Nationalization Score (PSNS). Building on the nationalization score for individual parties. In contrast. Where a party system is highly nationalized.

Finally. The measures of party nationalization in the United States have also been important in understanding partisan realignment (Brady. Kawato. Diamond (1988). In a weakly nationalized party system in a new democracy with profound ethnic. Stepan. 2000.10 He argues (p. The PSNS could help assess hypotheses 144 . the central party leadership may be less able to speak for the entire party and to deliver its legislative support. national issues are likely to be central in legislators’ careers. Reynolds (1999) and Stepan (2001) argue that in new democracies where pronounced ethnic. This hypothesis also is not true by fiat. Second. Third. national or religious cleavages. These public policy consequences seem particularly likely if a weakly nationalized vote pattern falls along regional lines. we would expect it to base its decisions in part on the degree of support it receives in specific geographic units. as Ames (2001) and Samuels (forthcoming) have convincingly argued. 331) that ‘if the goal is the consolidation of democracy in a multicultural or multinational polity. national or religious cleavages coincide with territory. A good measure of nationalization could help stimulate new research on such subjects and is essential for the empirical testing thereof. Decisions related to national transfers to sub-national units. Stepan (2001) contends this pattern may lead to wedge national and ethnic politics rather than to the ‘politics of accommodation’. Where a party’s base of support is relatively constant across geographic units. it merits more systematic research. if a governing party fares markedly better in some regions than others. sub-national issues are likely to be more important in legislative careers. Brazil illustrates this point. it may be salutary that some parties with sufficient electoral appeal to help form a national government articulate programmatic countrywide concerns. and may have consequences for. and subsidies may be strongly influenced by the degree of party system nationalization (Gibson and Calvo. In highly nationalized party systems. a strong case can be made that the existence of statewide parties is useful’. 2001). In such contexts. that is. and the PNS could serve a similar function in comparative politics.9 In contrast. administrative reform. the degree of nationalization reflects. Samuels. it may be more likely to treat all units equally. Under conditions of weak party nationalization. forthcoming.PA RT Y P O L I T I C S 9 ( 2 ) defined here. the nationalization of some major parties may be a key factor in preserving democracy. 1987). In a ‘patchwork’ or weakly nationalized party system. is analytically distinct from whether voters are responding to national issues. where its support varies widely across geographic units. parties will likely orient their electoral messages toward specific ethnic. Executives might have greater ability to forge legislative coalitions on the basis of national issues and to negotiate with a few key national party leaders. to use Lijphart’s (1977) term. differences in nationalization likely have public policy consequences. national or religious groups. 1985. legislative careers and for executive–legislative relations.

J O N E S A N D M A I N WA R I N G : T H E N AT I O N A L I Z AT I O N O F P A R T I E S about the relationship between party system or party nationalization and democratic stability in multi-ethnic societies. lower house elections form the best starting point. that is since the beginning of the third wave of democratization in the region (Huntington. a handful of countries or elections were excluded due to our inability to gather the necessary electoral data (Dominican Republic 1978–90) or to problems associated with the way the available data are reported (Colombia. our selection criteria were fourfold. Canada) as the geographic unit. presidential elections in multiparty systems have an idiosyncratic logic because of the importance of individual candidacies. although with few exceptions there was little change in these institutions during the time period under analysis. Third. El Salvador. show differences in patterns of electoral competition and pave the way for the testing of these and other hypotheses.e. The utility of our measures does not rest on whether these four hypotheses are correct. yet widely varying. The information reflects the current state of affairs as of the most recent lower house election included in our study. For the six federal republics. Fourth. we included only democratic lower house elections held since 1979. Venezuela) or province (Argentina.11 For Bolivia. and because there are greater. the measures are useful because they allow for systematic comparison. Although it would be possible to examine presidential and senate elections. many senates are renewed via partial renovation. A party could be quite nationalized in presidential elections because of the drawing power of a particular candidate and/or because of interparty coalitions. In terms of party nationalization. Furthermore. yet be weakly nationalized in lower house elections (the Brazilian PSDB [Partido da Social Democracia Brasileira] was an example in 1994 and 1998). we included only medium to large countries (i. Second. Ecuador. we included only countries that held at least three consecutive democratic lower house elections. Guatemala. First. Honduras. We used geographic units of roughly comparable political status. Venezuela 2000). Case Selection: Countries. those with populations greater than 2 million). Table 1 provides summary information about the political institutions in these 17 countries. this article empirically is limited to the United States. Within the Americas. and some districts do not vote in every senate election year. United States. Canada and 15 Latin American countries. because of party coalitions. We use lower house rather than senate elections because several countries have a unicameral legislature or a senate that is not popularly elected. Rather. Mexico. Nicaragua. Ecuador 1998. we utilized the state (Brazil. 1991). incentives for strategic voting. Geographic Units and Elections Although our measure has broad applicability. Peru and 145 .

. Abbreviations: PR = Proportional Representation. 22 and 20 additional members are elected using a separate ballot from a national district. of members in the lower house 257 130 513 301 120 57 82 84 113 128 60 500 90 180 435 99 207 Elections included 1983–2001 1985–1997 1986–1998 1980–2000 1989–2001 1982–1998 1979–1996 1994–2000 1990–1999 1981–2001 1980–1997 1994–2000 1990–2001 1980–1990 1980–2000 1984–1999 1983–1998 Country Argentina Bolivia Brazil Canada Chile Costa Rica Ecuador El Salvador Guatemala Honduras Jamaica Mexico Nicaragua Peru United States Uruguay Venezuela Electoral system Multi-member PR Mixed-member PR-C (national threshold) Multi-member PR Single-member plurality districts Binomial PR Multi-member PR Multi-member PR (national tier) Multi-member PR (national tier) Multi-member PR (national tier) Multi-member PR Single-member plurality districts Mixed-member PR-NC (national threshold) Multi-member PR (national tier) Multi-member PR Single-member plurality districts Multi-member PR (national allocation) Mixed-member PR-C (national tier) PA RT Y P O L I T I C S 9 ( 2 ) 146 Note: The federal district in Argentina. NC = Non-Compensatory. Mexico and Venezuela is included since residents of these districts elect voting members of the lower house (unlike their counterparts in the United States). Political institutions in 17 countries (as of the most recent election) Federal or unitary Federal Unitary Federal Federal Unitary Unitary Unitary Unitary Unitary Unitary Unitary Federal Unitary Unitary Federal Unitary Federal Name of unit of analysis Province Department State Province/Territory Region Province Department Department Department Department Parish State Department Department State Department State/Territory No. of units 24 9 27 13 13 7 21 14 23 18 14 32 17 26 50 19 24 No. Brazil. respectively. C = Compensatory. One US state employs Single-Member Majority Runoff Districts. 12 (every 4 years).Table 1. In Ecuador. Guatemala and Nicaragua.

with a median of 19. Brazil. use of these districts (for Canada. Furthermore. which given the nature of the Gini coefficient is not desirable. The data thus confirm previous scholarly work that emphasized the importance of provincial-level politics in Argentina. from a low of 7 (Costa Rica) to a high of 50 (United States). In the tables on party nationalization (Tables 4. Honduras. Combined with reporting PNSs at two decimal places. Because of the strategic behaviour by elites and voters that takes place in single-member plurality districts. provinces and parishes. Nicaragua 147 . Costa Rica. In contrast. 6 and 7). Costa Rica and Jamaica we employed the administrative regions. while for Chile. Although the Gini coefficient is superior to the alternatives. Jamaica and the United States) as our unit of analysis would provide less valid and reliable measures of party support than use of the larger territorial units. the effect on our interpretation caused by these differences in the number of units is modest. Jamaica. Mexico. because of spatial constraints. The level of nationalization varies markedly across party systems. Although the literature on the nationalization of the United States’ party system has emphasized its federalized nature. respectively. Bolivia and Guatemala also have weakly nationalized party systems. Peru and especially Brazil and Ecuador are much less nationalized than those of the six most nationalized countries.J O N E S A N D M A I N WA R I N G : T H E N AT I O N A L I Z AT I O N O F P A R T I E S Uruguay the departments were our unit of analysis. the United States and El Salvador are intermediate cases with PSNSs far higher than those of Ecuador and Brazil but lower than those of most unitary Latin American countries. the sensitivity to the number of units is modest given the relatively limited range in the number of units we use in this analysis. it might not work well for comparisons of countries with a wide range in the number of geographic units. the data on party system nationalization are based on all the parties as well as independent candidates. Canada and Venezuela and of regional-level politics in Ecuador. The party systems of Chile. The use of these administrative units is less arbitrary and suffers from less measurement error than the use of alternative measures such as deciles or quintiles of the rank-ordered sub-national administrative units. according to our measure. The Nationalization of Party Systems The bottom row of Table 2 gives the average PSNS for the 17 countries. we limit the presentation to parties that received a minimum of 5 percent of the national vote in the respective election. Canada. Fortunately. employment of these single-member districts would result in a much wider variance in the number of units per country in the analysis. during this period it ranked as an intermediate (not low) case of nationalization among our 17 cases. The systems of Argentina.

86 (1997) 0.92 (1993) 0.86 (1996) 0.92 (1986) 0.60 (1996) 0.83 (1984) 0.85 (2000) 0.85 (1982) 0.79 (1983) 0.72 (1997) 0.76 0.87 (1992) 0.94 (1980) 0. 1979–2001: PSNS.67 (1997) 0.80 (1989) 0.85 0.95 (1989) 0.68 (1993) 0.74 (1987) 0.85 (1988) 0.82 (2000) 0.90 (2001) Costa Uruguay Rica 0.72 (1997) 0.74 (1990) 0.92 (1997) 0.93 (1993) 0.87 0.86 (1998) Honduras Jamaica 0.80 (1985) 0.72 0.51 (1994) 0. Election Year and Period Average (in boldface) Ecuador Brazil 0.93 (1997) PA RT Y P O L I T I C S 9 ( 2 ) 148 0.84 0.73 (1999) 0.47 (1986) 0.63 (2000) 0.87 (1994) 0.78 (1988) 0.63 (1988) 0.93 (1985) 0.74 (1995) 0.81 (1985) 0.82 0. Party system nationalization in 17 countries.91 (2001) 0.76 0.82 (1980) 0.88 (2001) 0.70 (1980) 0.79 (1985) 0.86 0.Table 2.92 (1993) 0.79 (1997) 0.59 (1992) 0.80 (1998) 0.62 (1993) 0.86 (1983) 0.89 (1994) 0.77 0.62 (1979) 0.85 (1996) 0.91 (1990) 0.90 (1982) 0.58 (1994) 0.88 (1999) 0.85 (1990) 0.76 (1989) 0.72 (1993) 0.62 (1990) United El Argentina Canada Guatemala Venezuela Bolivia Mexico States Salvador Nicaragua Chile 0.54 (1984) 0.83 (2000) 0.88 (1984) 0.86 (1994) 0.93 .61 (1998) Peru 0.86 (1989) 0.83 (1984) 0.58 0.87 (1994) 0.92 0.83 (1997) 0.62 (1991) 0.78 (1993) 0.59 (2001) 0.86 (1990) 0.77 (1989) 0.51 (1990) 0.91 (1981) 0.80 (1999) 0.90 0.79 (1995) 0.87 0.70 0.65 (1986) 0.72 0.94 (1989) 0.57 0.85 (1988) 0.84 (1994) 0.83 (1986) 0.61 (1998) 0.78 (1980) 0.55 (1990) 0.

03 0.17 Country Jamaica Uruguay Honduras Nicaragua Costa Rica Unites States El Salvador Guatemala Bolivia Mexico Chile Argentina Ecuador Canada Venezuela Brazil Peru 149 . All six countries have unitary systems which. 1990 and. 1999) and rose (vis-à-vis the previous non-concurrent house election) in the presidential election year.02 0. the non-concurrent elections of 1986.04 0. in Argentina and Ecuador the level of nationalization has been higher in years of concurrent presidential/congressional elections.22 0. Ecuador (1984–96) and the United States (1980–2000) employed a mixed electoral cycle (i.03 0. the third measures the greatest change from one election to the next.14 0.03 0.04 0. the PSNS dropped following every presidential contest (1983.00 to 0.12 0.00 0.13 Greatest inter-election PSNS change 0.06 0.12 0.01 0.06 0.03 0. with the exception of the 1985 election.02 0. 1994 represent clear valleys Table 3. to a lesser extent.05 0.13 0.16 0.04 0.06 0. other things being equal. The second column measures mean change from one election to the next.07 0.05 0.08 0. In the United States there is no relationship between the electoral cycle and party system nationalization (perhaps because of the country’s decentralized two-party system).01 0.02 0.03 0.16 0.06 0. Table 3 provides information on change and stability in nationalization by country over time.08 0.e.05 0.18 0.07 0. every other lower house election is concurrent/non-concurrent with the presidential contest) for several cycles.25 0.08 0. Argentina (1983–2001). Measures of stability in party system nationalization Mean inter-election PSNS change 0.21 0. In Ecuador.03 0. and the fourth column measures aggregate change.02 0.03 0.14 0.03) in their level of party system nationalization.07 0.07 0.01 0. In contrast. offer stronger incentives to organize parties across departmental lines.01 0.06 0. In Argentina.J O N E S A N D M A I N WA R I N G : T H E N AT I O N A L I Z AT I O N O F P A R T I E S and Uruguay are quite nationalized. The PSNS also enables us to trace change and stability within a given country.02 0.06 0. 1989.17 Highest–lowest PSNS 0. Nine of the 17 countries experienced considerable stability (a mean interelection PSNS change of 0.14 0. 1995.

Ecuador’s smaller parties are even less nationalized. Low Nationalization Countries12 We now shift from party systems to parties as the unit of analysis. 1988.79 in 1985 to 0. Ecuador’s major parties possess low average PNSs. from 0. 1975). Argentina. with several among the least nationalized of our entire party population. in none of the eight elections did the parties winning 20 percent or more of the vote account for more than half of the national vote. Although we have not systematically collected data for other regions. Canada. the PJ (Partido Justicialista) and UCR/ALIANZA (Unión Cívica Radical/Alianza). Table 4 presents data on party nationalization in Ecuador. with Fujimori’s C90 (Cambio 90) registering an extremely low (especially for a governing party) 0. Three of the 11 least nationalized major parties are Brazilian (Table 5). In 1990 all parties possessed relatively low PNSs. when President Alberto Fujimori staged a coup in 1992. which normally compete in only one province and hence 150 . as the traditional Peruvian party system disintegrated. 1996). Brazil’s parties have won widely diverging vote shares across the 27 units of the federation. Brazil. some elections in developing countries such as India and Malawi have yielded less nationalized party systems than have elections in any country in our population. These smaller parties account for a substantial portion of the vote.45. with three among the 11 least nationalized major parties (45 total) in our population (Table 5). In our population. The Rose and Urwin indicator we consider reliable (see below) indicated that the United States had a less nationalized party system than any Western European country. Peru. Guatemala. 2000.PA RT Y P O L I T I C S 9 ( 2 ) with respect to the peaks provided by the presidential years (1984.62 in 1990. Furthermore. the United States’ party system is more nationalized than that of a majority of the other countries. This finding indicates the nationalizing effect that presidential elections can have on congressional elections. Brazil is another outlier. both of which have intermediate to high PNSs. and regional and state differences in the party system have long been salient. coexisted with several powerful provincial parties. Peru provides the largest inter-election PSNS change. With the partial exception of the PMDB (Partido do Movimento Democrático Brasileiro) in 1986. Peru’s nascent democratic experience ended shortly thereafter. the national party system is a pastiche of state party systems. Ecuadoran and Peruvian. 1992. by contrast. Canadian. but its pattern of party federalization diverges from the Brazilian. For much of this period. State-level politics is very important. the norm in Western Europe is a high degree of nationalization (Caramani. Venezuela and Bolivia. two main parties. Argentina is another case of low nationalization. Rose and Urwin.

The RP drew many of its votes from the historic base of the PC (Progressive Conservatives). In contrast. in our population). Most of Bolivia’s major parties possess intermediate levels of nationalization.78 in 1988 to 0. 151 . the Canadian party system had a level of nationalization only slightly lower than that of the United States. CONDEPA possesses the lowest PNS of the major parties in our analysis (Table 5). In the 1980s. the other Guatemalan parties had low levels of nationalization. CONDEPA (Conciencia de Patria) is a weakly nationalized party. its support is heavily concentrated in the Department of La Paz. became increasingly nationalized over time. Since 1993. CONDEPA helps drag the country’s party system from the intermediate to low classification. resulted in a massive PSNS decrease (the largest aggregate shift. This was especially the case for the FRG with a PNS of 0. also has a very low PNS. pushing Venezuela into the low nationalization category. In the 1990s. the PAN (Partido de Avanzada Nacional) and FRG (Frente Repúblicano Guatamalteco). only the LP remotely approached a significant nationwide presence. In the 1990s.25. the decline of the two parties (AD [Acción Democrática] and COPEI [Comité de Organización Política Electoral Independiente]) that had dominated Venezuelan politics for the previous quarter century.5 percent of the national vote. two new forces emerged at the national level and dramatically lowered Canada’s PSNS. of the five major parties. Guatemala’s two largest parties during this period. which saw its support dissipate in the West. The Reform Party (RP).08). reflecting its concentration in the West. With one minor exception. generally averaging between 0. In contrast to the 1980s. The remarkable changes in the Canadian party system in 1993 were associated with the second largest single inter-election PSNS change (tied with Ecuador) in the 68 inter-electoral periods covered in this study. however. from 0. Until 1993. cumulatively these parties have won a meaningful share (between 5 percent and 8 percent) of the national vote in most elections. the Venezuelan party system ranked in the high nationalization category. Canada has approximated Brazil’s PSNSs. when two of the country’s three major parties (PC and LP [Liberal Party]) were relatively nationalized.72 in 1990). which first contested national elections in 1988 (garnering a meager 2 percent of the vote). 0. is the sole single-province party in the 17 countries analysed here that obtained 5 percent or more of the national vote. in the 1990s.91 in 1999 (up from 0. After Canada’s BQ.J O N E S A N D M A I N WA R I N G : T H E N AT I O N A L I Z AT I O N O F P A R T I E S have extremely low PNSs. but remained competitive in the Maritime provinces. along with constitutional reforms that decentralized power to the states. The BQ (Bloc Quebecois) (PNS 0.62 in 1993. however. Although each provincial party accounts for limited weight in the PSNSs presented in Table 2.5 percent and 1. which competes only in Quebec.

69 PC 0.82 0.64 C90 Other2b 0.55 0.71 0.61 0.50 0.82 0.68 Othere 0.34 0.08 0.61 Political parties Other1a 0.61 0.08 0.69 IU 0.68 0.48 0.56 MPD 0.49 0.76 0.45 0.59 0.52 PLRE 0.79 PFL 0.53 0.54 0.68 APd 0.67 0.58 0.72 0.77 0.92 0.84 0.57/0.88 0.82 PJ 0.53 0.37 0.50 PDT 0.75 0.32 0.60 PSC 0.40 0.62 UCR 0.53 IS DP 0.58 0.61 0.52 0.89 0.63 0.83 0.84 0.79 0.66 0.75 LP 0.62 0.91 0.47 0.73 0.88 0.82 0.52 0.74 0.82 0.67 0.87 0.53 0.63 0.65 PRN 0.46 0.71 0.38 0.69 0.37 0.46 0.Table 4.48 0.57 0.50 PT 0.65 0.64 0.62 0.54 PRE 0.82 0.64 0.86 ID 0.51 0.51 0.45 BQ RP 0.70 CD 0.46 0.52/0.86 0.88 152 0.60 0.59 0.70 0.49 PSDB 0.77 0.43 0.41 0.67 PA RT Y P O L I T I C S 9 ( 2 ) PMDB 0.61 0.59 0.72 0.74 APRA 0.37 0.67 0.61 NDP 0.91 0.59 0.69 0.08 ALIANZA 0.86 0.55 .71 0. Party nationalization (PNS) in the low nationalization countries Country Ecuador 1979 1984 1986 1988 1990 1992 1994 1996 Brazil 1986 1990 1994 1998 Peru 1980 1985 1990 Canada 1980 1984 1988 1993 1997 2000 Argentina 1983 1985 1987 1989 Election year CFP 0.67 0.55 0.81 0.81 0.47 0.31 0.55 0.73 0.69 PPBc 0.54 0.87 0.34 PTB 0.73 0.

1995: FREPASO. 1989: IU.84 0.86 0.88 AD 0.86 0.84 0.85 0.77 0.68 MVR PRVZL ALIANZA Guatemala LCR 0.76 UCN 0. . (continued) Country Argentina (continued) Election year 1991 1993 1995 1997 1999 2001 1990 1995 1999 Venezuela 1983 1988 1993 1998 Bolivia 1985 1989 1993 1997 a b J O N E S A N D M A I N WA R I N G : T H E N AT I O N A L I Z AT I O N O F P A R T I E S Political parties PJ 0.82 0.69 0.87 0.66 0.57 1979: PCE.64 Otherj 0.76 MAS 0.86 0.76 ADN 0.86 0.33 0. g The DCG. 2001: ARI. 1986: FADI.83 0. c PDS in 1986 and 1990. 1992: PUR.83 PAN 0.74 153 0.77 Othere 0.24 0.77 0.81 0.88 0.77 0.83 0. 1999: AR. 1999: DIA-URNG. 1991: UCEDE.88 MAS 0.62 MNR 0. 1989: ADC. 1995: FDNG.80 0. 1993: MBL.58 MIRi 0.72 0. h 1990: MLN.75 0. PP/PPR in 1994. e 1985: PI.72 0.51 0. f Presented in alliance with the PID and FUN in 1990.88 0.82 UCR 0.66 CONVERG 0.60 0.82 DCGg 0.90 0. 1994: APRE.88 0. 1996: MUPP-NP.37 0.70 CONDEPA 0. 1979: CID. UCN. 1987: UCEDE.81 0. d FREDEMO alliance in 1990.36 0. j 1985: MNRI. and PSD presented in alliance in 1995.92 0. 1992: PCE. 1993: MODIN.73 Otherh 0. 1984: FADI.69 0. 1990: PSE.Table 4.72 0.88 0. 1986: FRA.45 UCS 0.82 0.62 0.80 0.88 0.91 COPEI 0. 1994: PCE.42 0.93 0. i Presented in alliance with the ADN in 1993.82 0. 1984: FRA/PD.46 FRGf 0.80 0.

Mexico’s PSNS thus conceals important cross-party differences (Table 6). is tied as the tenth most nationalized of our major parties (Table 5).95 0. For additional information.92 0.67 0.69 0.PA RT Y P O L I T I C S 9 ( 2 ) Table 5.95 0. Mayhew.92 0. the PRD (Partido de la Revolución Democrática) is the tenth least nationalized. the United States’ party system is more nationalized than half of the 154 .93 0.40 0.63 0. the PRI (Partido Revolucionario Institucional).70 0. In contrast.96 0. Intermediate Nationalization Countries Despite formal federalism. 1986).65 0. the Mexican political system was highly centralized until the 1990s.93 0. Brazil and Canada. 1987. Although a substantial literature has correctly emphasized the federalized character of the United States’ party system (Kawato. In contrast to the situation in the latter countries. Ranking of the 11 most nationalized and least nationalized major parties Ranking Top eleven Party PUSC PLN ARENA CONCERTACION PLH PNH PNP JLP PC ALIANZA PRI ID PRD NDP PSC PT PFL PSDB PRE RP CONDEPA BQ Average PNS 0.54 0.93 0.93 0.94 0.59 0.65 0. The most nationalized party.94 0.55 0. It is thus unsurprising that Mexico has a more nationalized party system than Argentina.08 Average vote share (%) 40 45 39 53 48 45 54 45 34 37 42 14 20 14 18 11 15 14 14 21 14 12 Country Costa Rica Costa Rica El Salvador Chile Honduras Honduras Jamaica Jamaica Uruguay Chile Mexico Ecuador Mexico Canada Ecuador Brazil Brazil Brazil Ecuador Canada Bolivia Canada Bottom Eleven Note: Includes only parties (45 total) that averaged 10% or more of the vote and competed in at least three elections during the analysis period. Mexico is the country with the sharpest differences in nationalization from one major party to the next. see Tables 4. Mexican state governments are comparatively bereft of resources. 6 and 7.

85 0.83 ARENA 0.88 in 2000.88 0.78 0.86 0.85 0.71 PCN 0.84 0.84 0.83 0.69 FMLN 0.72 CD/CDUa 0.83 0.91 DP 0.75 0.67 0.94 PAN 0.86 0. The FMLN’s growth was more even.87 0.83 0. had become El Salvador’s second largest party. the United States’ parties are indeed quite federalized.90 0. merely 2 percent behind ARENA (Alianza Republicana Nacionalista). The Salvadoran party system underwent a striking transformation during this period.68 CD in 1994 and CDU in 2000.88 PDC 0. Honduras and Jamaica. its PNS rose only slightly. the PDC (Partido Demócrata Cristiano).74 0. High Nationalization Countries The two parties that have dominated Nicaraguan politics since 1990. By 1997 the FMLN (Frente Farabundo Martí para la Liberación Nacional).96 0. Party nationalization (PNS) in the intermediate nationalization countries Country Mexico 1994 1997 2000 United States 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 El Salvador 1994 1997 2000 a Election year PRI 0.94 0. suffered an electoral haemorrhage.84 RP 0.75 0.47 0.82 0. The PDC collapse occurred very unevenly across the country: the party moved from a relatively high PNS to a low one between 1994 and 1997.91 0. a revolutionary group in the 1980s. the FSLN (Frente Sandinista de Liberación Nacional) and PLC (Partido Liberal 155 .90 0. from 0. But if we compare the PNSs of the Democratic (DP) and Republican (RP) parties with those of the major parties in Costa Rica.85 Political parties PRD 0.91 0. systems in this population.84 0. The largest party in the 1980s.90 0.J O N E S A N D M A I N WA R I N G : T H E N AT I O N A L I Z AT I O N O F P A R T I E S Table 6.87 0.84 0.96 0.87 0.71 0.84 in 1994 to 0.84 0.79 0.87 0.83 0.

81 UNO alliance in 1990.94 0. Party nationalization (PNS) in the high nationalization countries Country Nicaragua 1990 1996 2001 Chile 1989 1993 1997 2001 Uruguay 1984 1989 1994 1999 Costa Rica 1982 1986 1990 1994 1998 Honduras 1981 1985 1989 1993 1997 2001 Jamaica 1980 1989 1993 1997 a b Election Year FSLN 0.93 0.85 PU 0.94 0.96 0.PA RT Y P O L I T I C S 9 ( 2 ) Table 7.90 0.96 0.92 0.92 0.94 0.95 JLP 0. In response to the two-member district legislative electoral system employed in Chile.96 0.94 0. the Nicaraguan minor parties are much less nationalized than the dominant parties.93 0.94 0.87 0.75 0.91 0.92 0.73 NEb 0. Constitucionalista).92 PN 0. however.94 PC 0.92 0.72 0.95 0.72 0.94 0.96 0.93 0.97 0.87 0.71 0.93 0.94 PNH 0.94 0.83 FD 0.95 0. As is the case in the other Central American countries.97 0.88 0.95 0.94 PLH 0.68 0.87 PUSC 0.92 0.63 0.95 0.91 0.93 PAIS 0.93 0.76 0.89 CONCERT 0.95 0.94 PLN 0.94 PNP 0.86 0.97 0.93 0.94 0.47 PC 0.89 ALIANZA 0.79 0. PGP in 1989.95 0.67 FA 0. the major Chilean parties formed two large and relatively stable national coalitions. The Concertación (CONCERT) includes 156 . have relatively high PNSs.96 Political parties PLCa 0.87 0.

13 Sundquist (1973: 332–7) and Kawato (1987) devised measures of nationalization based on the DP vote share. but they are appropriate only for two-party systems. They represent six of our nine most nationalized parties (Table 5). Costa Rica. Claggett et al. In contrast. The Alianza por Chile (ALIANZA). garnering relatively equal shares of the vote in all districts. But in a multiparty system this measure greatly overstates the nationalization of small parties compared to large ones.J O N E S A N D M A I N WA R I N G : T H E N AT I O N A L I Z AT I O N O F P A R T I E S four centre and centre–left parties. They tend to attribute higher values to large parties and lower values to small ones because there is likely to be less variance in absolute terms with a lower share of the vote. measured the standard deviation (SD) of the Democratic Party’s (DP) share of the vote at the county level. With the partial exception of the ALIANZA in 1997. Sundquist (1973) and Kawato (1987) devised good measures of party nationalization for the United States. Schattschneider (1960). the leftist FA (Frente Amplio) had a markedly less nationalized pattern until 1994. (1984). as it is currently named. Schattschneider (1960: 78–96) measured the DP’s deviation from 50 percent. includes two centre–right parties along with a substantial number of independents. We therefore used coalition (rather than party) results. but focusing on one party is suitable only in two-party systems. Previous Measures of Nationalization Claggett et al. The FA’s ability to win more votes across a broader range of departments was a key factor in its electoral growth from 21 percent of the national vote in both 1984 and 1989 to 31 percent in 1994 and 40 percent in 1999. the PN’s electoral decline after 1989 (39 percent). In contrast to what has occurred in most cases. SDs perform poorly for comparing parties that win very different mean shares of the vote because they are not comparable across parties with different vote shares. to 31 percent of the vote in 1994 and 22 percent in 1999. the SD is a poor option for comparative research. in every state and summed this value for all states. Because most democracies (those with two-party systems are rare exceptions) have relevant parties that win widely different vote shares. Honduras and Jamaica have the most nationalized party systems. has yet to produce sharply lower PNSs. It fared much better in Montevideo than in the interior. these two major coalitions have consistently had very high PNSs. have consistently had relatively even vote patterns across the country’s 19 departments. expressed as an absolute value. It is possible to follow this procedure for other countries by taking the deviation (as an absolute value) from a given party’s mean share of the vote in all states or provinces. The respective two dominant parties in these countries are highly nationalized. this is one of the measures that Rose and Urwin (1975) used. 157 . Uruguay’s two traditional parties. the PC (Partido Colorado) and PN (Partido Nacional).

as noted. also used by Caramani. Canada. Rose and Urwin’s Index of Cumulative Regional Inequality (ICRI). which reproduces the modification of Schattschneider discussed in the previous paragraph. In addition to the SD and ICRI. it has serious shortcomings for comparative research. is their only solid indicator. which with a few exceptions have fairly nationalized party systems. the works by Rose and Urwin (1975) and Caramani (2000) are the main contributions. They both devised three different measures of nationalization. which. India and Russia without paying attention to the widely divergent vote shares that parties win in different states or provinces. taking the absolute value. and dividing by two. Rose and Urwin also used an Index of Variation (IV). This is in part because the mainstream theoretical literature on parties and party systems has focused on the advanced industrial democracies. Such profound discrepancies reflect the serious problems with the SD and IV as measures of party nationalization. Brazil. But when we turn our attention to Africa. It is constructed by calculating the percentage of a party’s national vote won in a given territorial unit. Conclusion The extent of party and party system nationalization is an important topic that has been neglected by the scholarly literature. As noted. subtracting that territorial unit’s share of the national vote. the United States ranks as the least nationalized party system among 20 countries on 2 of them but as the most nationalized on the third. It is impossible to understand many party systems in the world. Caramani employed the CV. Their three measures perform very differently. in comparative politics little has been written on this subject. the debate about nationalization in the United States usefully highlighted the importance of the subject. Asia and Latin America. 158 . We have proposed a means of measuring nationalization that travels well across parties. Hence.PA RT Y P O L I T I C S 9 ( 2 ) Although the measures of nationalization used for the United States do not work for most other countries. the importance of analysing variance in nationalization is greater for these regions than for the advanced industrial democracies. including those of such large and important countries as Argentina. This article is the first to provide empirical information about party system nationalization outside of the advanced industrial democracies. performs very poorly for comparing parties of unequal sizes. This measure performs well but has the disadvantage of being less well known and used than the Gini coefficient. a much greater proportion of the party systems are weakly nationalized. the shortcomings of which were discussed earlier. summing these absolute values for all units. Both used the SD. Curiously. countries and time. except for the United States. Our work makes clear that party system nationalization varies markedly across countries.

in sharp contrast. falling to 0. The correlation between a binary federal–unitary variable and the PSNS is –0. the federal countries tend to have lower nationalization scores. (b) it is measurable. In Mexico. They should be useful for scholars studying various aspects of the party system and its relationship to factors such as voters’ orientations. with the PNS of the governing party. the more fragmented party systems tend to be the least nationalized. executive–legislative relations. But beyond this important similarity the nationalization data reflect a difference. the larger parties tend to be more nationalized and the smaller parties less nationalized. Consider the transitions from hegemonic party systems to democratic politics in Brazil (1974–85) and Mexico (1988–2000). The PNS calls attention to these differences between otherwise similar cases and captures their magnitude. reflecting greater variance in parties’ electoral performances across the sub-national units.87.36. In both countries the pro-regime parties held remarkably similar positions: each had barely lost its majority in the lower house but retained a solid majority in the senate.41 in 1986.J O N E S A N D M A I N WA R I N G : T H E N AT I O N A L I Z AT I O N O F P A R T I E S Issues of measurement provide a key underpinning for examining important substantive issues in the social sciences. These transitions were exceptional in the degree to which elections dictated the transition’s pace. three potential explanatory factors leap out in this set of countries. The legislatures in the low nationalization countries are not only 159 . It is worth constructing new measures if: (a) the subject of measurement is important enough to merit sustained substantive discussion. The PRI’s PNS remained high (0. measure stability and change in nationalization within a given party system. Second. public policy and democratic stability in multi-ethnic societies. In Brazil. Our measures could also be useful for tracking and comparing different political processes. and assess the relationship between realignment or dealignment and nationalization. Our measures enable us to compare across party systems. legislative careers. and (c) social scientists gain a useful tool by being able to measure the phenomenon precisely. compare across parties in one or more countries. Federalism gives incentives for parties to organize and compete at the state level and tends to foster more differences than unitary systems in interstate patterns of electoral competition. small parties win most of the vote.91) even in 2000. First. We believe that all three conditions obtain. Although for reasons of space a statistical analysis of the sources of variance in party system nationalization must be left to a future study. The correlation between the PSNS and the effective number of electoral parties is a remarkable –0. the PRI remained competitive in all 32 states. hence the powerful tendency towards low PSNSs.14 In this universe of countries. analyse the relationship between electoral growth or demise and party nationalization. the PDS (Partido Democrático Social). notwithstanding the fact that the two issues are obviously conceptually and operationally discrete. the wealthiest states tilted strongly to the opposition by 1974 and became overwhelmingly dominated by it in 1982. In fragmented party systems.

the party cannot afford to lack a presence in any administrative unit and thus invests resources in the development and maintenance of a nationwide party structure. and have a chance to win seats. More remains to be done on these issues. Third. This party structure in turn presents candidates in the legislative elections (especially when the presidential and legislative elections are concurrent) regardless of their chances of success. the presidency. Large parties (those that win over 30 percent of the vote) almost always compete. 160 . A low level of nationalization does not necessarily result from a deliberate effort of parties to cast state or regional appeals. the driving force behind this relationship is most likely the previously noted correlation between fragmentation and nationalization. However. a deliberate focus on state or regional appeals characterizes very few parties. One finding stands out in the variance across individual parties: small parties tend to have substantially lower nationalization scores. In contrast. adding one more potential complication to coordination efforts in these assemblies. Because the president is elected by a national vote. pave the way for future research. the PNS and the PSNS. Large parties can realistically compete for the country’s most important electoral prize. in virtually all of the sub-national administrative units. Individual parties range from almost perfectly nationalized (the Costa Rican PUSC [Partido Unidad Social Cristiana] having three times registered a nationalization score of 0. significant electoral growth was accompanied by a PNS increase and significant decline by a decrease. This difference likely stems in part from the distinct incentives for nationalization facing large and small parties in presidential democracies. but we believe that our measures. present candidates and provide these candidates with campaign resources in districts where they have little chance of success. Almost invariably.97) to perfectly provincialized (the Canadian BQ has won 100 percent of its votes in Quebec). It would be unusual for a party to win 30 percent or more of the vote if it were a minor party in a substantial number of provinces. in addition. and our empirical information on 17 countries. Party nationalization varies even more dramatically than party system nationalization. Conversely. these different partisan forces tend to have distinct geographic constituencies. most small parties in our population won wildly divergent shares of the vote in different provinces. there is a noteworthy inverse relationship between electoral volatility and party nationalization. small parties lack a realistic chance of winning the presidency and thus have a weaker incentive to develop and maintain a party structure. A related pattern that emerged consistently is the relationship between party growth or decline and nationalization. In these countries.PA RT Y P O L I T I C S 9 ( 2 ) generally fragmented along partisan lines.

161 . The Gini coefficient is obtained by subtracting the sum of column 7 from the sum of column 8. Gi = (iYi+1) – (i+1Yi) where Xi is the cumulative proportion represented by the percentage of the valid vote won by Party X in the ith state divided by the sum of percentages won by Party X in all states.) The sum of these vote shares is 80 percent (column 2). The final row in column 7 is the sum of all row entries for column 7. Column 8. from the state where Party X won the smallest share of the vote to the state where it won the largest.50. Party X wins 0 percent. i. and so on. Column 7. province or department equally – a choice that we would vigorously defend.15 In this case. The rows should be arranged by the ascending or descending vote shares that Party X won in the different states. provinces or departments. The final row in column 8 is the sum of all row entries for the column.25 * 0. which represents an example of uneven electoral performance across states. not in some random order. we divided Party X’s vote share in that state by 80 percent to determine that state’s contribution to the party’s unweighted (by population) aggregated vote percentages. or 0. i. 0. The PNS = 1 – Gini. Column 7. for each state. row A is the product of Xi+1 (row B in column 6) times Y1 (row A in column 5). row A represents the product of Xi (row A in column 6) times Yi+1 (row B in column 5). which in this case plots the cumulative proportion of states that the ith state represents of the total number of states (column 5 in the example below) against the cumulative percentage of vote shares (not of votes) won by Party X in the ith state relative to the sum of percentages won by Party X in all states (column 6 in the example below). and Yi is the cumulative proportion that the ith state represents of the total number of states or other political sub-units. Column 6 gives the cumulative totals for column 4.e. row B represents the product of row B in column 6 times row C in column 5.e. and so forth. 10 percent. In the hypothetical example below.J O N E S A N D M A I N WA R I N G : T H E N AT I O N A L I Z AT I O N O F P A R T I E S Appendix: Using the Gini Coefficient to Measure Party Nationalization The following explanation of the Gini coefficient is based on Shryock et al. 30 percent and 40 percent of the vote in the four states of a given country.5625. The same method is used for determining the level of income inequality across states and many other purposes. i.4375.e. the Gini coefficient is therefore 1. 2001). It is based on the Lorenz curve.0313 – 0.125. 0 * 0. In column 4. 0. We calculated the Gini coefficient for each party using the STATA command ‘ineqdec0’ (Jenkins. (1976: 98–100). Using the Gini coefficient to calculate nationalization scores entails weighting every state.5938. The Gini coefficient as used in this article measures the inequality in a party’s vote share across different states. (The example given below uses an ascending order.

50 0.00 0.5000 – 0.10 0.25 Column 2 divided by sum of column 2 in all states (4) 0.40 State represents what share of all states (i.50 Cumulative share of states Yi (5) 0.5938 Xi+1*Yi (8) 0.25 0.30 0.0313 162 Sum = 0.25 0.125 0.00 Cumulative share of states for column 4 Xi (6) 0.00 0.e.7500 – 1. 1 divided by no.2500 0.125 0.25 0.00 0.00 0.0938 0..25 0.State (1) A B C D Party’s share of vote in state (2) 0. of states) (3) 0.375 0.80 .50 1.75 1.00 PA RT Y P O L I T I C S 9 ( 2 ) Xi*Yi+1 (7) 0.0313 0.

In those analyses where the Gini coefficient is most commonly employed (i. it works only in democracies with personal voting. Cox (1997) analysed the ways in which district-level political competition leads to varying party system patterns at the national level (see also Chhibber and Kollman. 2000). 1967) and Sundquist (1973) measured the nationalization of parties or electoral trends in the United States. Daniel Brinks. Michael Coppedge. Schattschneider (1960). 2 Although scant comparative work exists on party system nationalization. The RM compares the percentage of the vote won by a party in the top quartile of states to that won in the bottom quartile. some scholars have addressed related issues. but rather at the district level. Observing that electoral systems create incentives not at the national level. The CV has the disadvantage of being less easily interpretable than the Gini coefficient because its values have no intrinsic meaning (i.11 (Caramani. where Vi is the vote share of the ith party. 300). In our analysis this small sample property is most notable when a party wins 100 percent of its vote in a single electoral district. 5 For additional information on calculating the Gini coefficient. William Reed. Ashutosh Varshney. small sample analyses) the limiting result is not obtained. Although we do not attempt to explain how districtlevel competition is linked to the national party system. Claggett et al. 3 Bawn et al. but it did so in a less consistent manner. Their contributions are valuable. see the Appendix. They defined electoral cohesiveness as ‘the extent to which the electoral fates of incumbent candidates of the same party are tied together’ (p. Stokes (1965.J O N E S A N D M A I N WA R I N G : T H E N AT I O N A L I Z AT I O N O F P A R T I E S Notes We are grateful for comments provided by Ana María Bejarano. Matthew Cleary. and it ranges from 0 to 1. Aseema Sinha. The CV weights the standard deviation through the mean.e. no upper limit). it is technically inferior because it arbitrarily selects two data points and ignores the rest. The RM provided results similar to the Gini coefficient. 4 This is the asymptotic result. but their measures do not work well except in two-party systems. But the Gini coefficient has two advantages over the CV. Because their measure is based partly on an incumbent’s previous margin of victory. While this measure is easier to interpret. Herbert Kitschelt.e. It provided results substantively indistinct from those obtained using the Gini coefficient. two anonymous reviewers. 8 This is a particular problem for econometric analysis that employs the effective 163 .52 to 3. (1999) used the term ‘electoral cohesiveness’ rather than ‘nationalization’. Support for this research was provided by the National Science Foundation (SES 9911140). 1 Brady (1985). Our endeavour is related. Kawato (1987). 1998). Fran Hagopian. differences in nationalization reflect how closely results from one district mirror those of others. Pradeep Chhibber. whereas the Gini coefficient ranges from 0 to 1. 7 PSNS = PNSi * Vi). It is better known and equally good. we calculated two alternative measures of nationalization: a ratio measure (RM) and the coefficient of variation (CV). the mean CV for 17 Western European countries from 1918 to the 1990s ranged from 0. (1984). 6 As a diagnostic. For example. Wonjae Hwang. and panel participants at the 2000 Meeting of the Midwest Political Science Association and the 2001 Congress of the Latin American Studies Association. Jim Granato.

This correlation would most likely be lower in Western Europe. territorial and religious lines. 164 . Barry (2001) The Deadlock of Democracy in Brazil. it would be impossible for PB to be less nationalized. the highest possible deviation would be 50 percent ϫ 20 = 1000 percent. Cox and Frances Rosenbluth (1999) ‘Measuring the Ties that Bind: Electoral Cohesiveness in Four Democracies’. Ann Arbor.5 percent of the vote in 10 states and 59. Yet PA would be obviously and profoundly more nationalized than PB. they inflamed ethnic tensions rather than seeking accommodation. (1985) ‘A Reevaluation of Realignments in American Politics: Evidence from the House of Representatives’.PA RT Y P O L I T I C S 9 ( 2 ) 9 10 11 12 13 14 15 number of parties as a dependent variable and average district magnitude as an explanatory variable and pools countries/elections with varying levels of party nationalization. MI: University of Michigan Press. With a different set of cases. party leaders cultivated their own ethnic and territorial bases to the exclusion of others. National territories where citizens vote in national legislative elections are referred to as provinces/states. In Tables 4. PA would have the same nationalization score as PB’s theoretical maximum (based on winning 100 percent of the vote in one state and 0 percent in 19) if PA won 40. 1988) and to a lesser degree in 1983. then the Gini coefficent is the absolute value of this process of subtraction. Gary W. Rather than trying to build bridges across ethnic. Assume a country has 20 states. ethnic and religious identities. and reinforced. If the descending vote shares are used. 6 and 7 we group the 17 countries into cases of low. Ann Arbor. If Party A (PA) wins a mean vote share of 50 percent. Bawn. in Bernard Grofman et al. where despite considerable variance in the effective number of parties variance in nationalization is lower than in the Americas. In the 1960s. Also relevant here are factors such as the percentage of a party’s national-level vote accounted for by a province and the percentage of a party’s legislators who come from a province. This would result if PB won 100 percent in one state and 0 percent in 19. the highest possible deviation would be (95 percent ϫ 1) + (5 percent ϫ 19) = 190 percent. If Party B (PB) wins a mean share of 5 percent of the vote. David W. (eds) Elections in Japan under the Single Non-Transferable Vote. Brady.5 percent in the other 10. Kathleen. intermediate and high nationalization based on the PSNSs. MI: University of Michigan Press. References Ames. American Political Science Review 79: 28–49. In fact. some regrouping would result. The resulting conflicts contributed to the breakdown of democracy in Nigeria in 1966 (Diamond. This would result if PA won 0 percent of the vote in some states and 100 percent in the others. Nigeria exemplified the perils of a party system with highly regionalized parties that simultaneously arose from. a fact that is captured by the Gini coefficient. These groupings are purely inductive. as would have been necessary for parties that aspired to have a nationalized electoral base.

Norman. Scully (1995) ‘Introduction: Party Systems in Latin America’. Gary W. Stanford. and 165 . Mainwaring. (Forthcoming) Careerism and its Consequences: Federalism. Creedy. Lijphart. Stanford. Harmel. Robert and Lars Sväsand (1993) ‘Party Leadership and Party Institutionalisation: Three Phases of Development’. New York: Free Press. and Ernesto Calvo (2000) ‘Federalism and Low-Maintenance Constituencies: Territorial Dimensions of Economic Reform in Argentina’. Jenkins. West European Politics 16: 67–88. 1945–1990. Edward L. Mayhew. University of Essex. New Haven. OK: University of Oklahoma Press. Cambridge: Cambridge University Press. Elgar. Cox. and Democracy in Nigeria: The Failure of the First Republic. CA: Stanford University Press. Beverly Hills. David R. Janda. CT: Yale University Press. Chhibber. Gibson. Oxford: Macmillan. Andrew (1999) Electoral Systems and Democratization in Southern Africa. Richard and Derek W. Samuels. Northhampton. William Flanigan and Nancy Zingale (1984) ‘Nationalization of the American Electorate’.J O N E S A N D M A I N WA R I N G : T H E N AT I O N A L I Z AT I O N O F P A R T I E S Caramani. Daniele (2000) Elections in Western Europe since 1815: Electoral Results by Constituencies. Syracuse: Syracuse University Press. Electoral Settings. MA: E. David J. Mainwaring. (1991) The Third Wave: Democratization in the Late Twentieth Century. American Political Science Review 78: 77–91. New York: Oxford University Press. Arend (1977) Democracy in Plural Societies: A Comparative Exploration. Party Politics 4: 5–31. Samuel P. Sadafumi (1987) ‘Nationalization and Partisan Realignment in Congressional Elections’. Kollman (1998) ‘Party Aggregation and the Number of Parties in India and the United States’. Sage Series Number 06–007. Kawato. Urwin (1975) ‘Regional Differentiation and Political Unity in Western Nations’. Scully (eds) Building Democratic Institutions: Party Systems in Latin America. CA: Stanford University Press. Giacomo and Giovanni Sartori (1983) ‘Polarization. NJ: Princeton University Press. Knutsen. John (1998) The Dynamics of Inequality and Poverty: Comparing Income Distributions. Larry (1988) Class. Cambridge: Cambridge University Press. and Government Activity in the Twentieth Century. (1986) Placing Parties in American Politics: Organization. Oxford: Oxford University Press. Stephen P. Arend (1994) Electoral Systems and Party Systems: A Study of TwentySeven Democracies. American Political Science Review 81: 1235–50. Ethnicity. Princeton. Studies in Comparative International Development 35: 32–55. American Political Science Review 92: 329–42. William. Sani. Rose. Reynolds. in Scott Mainwaring and Timothy R. (1997) Making Votes Count: Strategic Coordination in the World’s Electoral Systems. Huntington. Lijphart. Pradeep and Ken W. Scott and Timothy R. Oddbjørn (1998) ‘The Strength of the Partisan Component of Left–Right Identity: A Comparative Longitudinal Study of Left–Right Party Polarization in Eight West European Countries’. Elections. CA: Sage. and Policy Making in Brazil. Scott P. Fragmentation. (2001) ‘ineqdec0. Kenneth (1980) Political Parties: A Cross-National Survey. (1999) Rethinking Party Systems in the Third Wave of Democratization: The Case of Brazil. STATA Command. Claggett. Diamond.ado’.

elite and mass political behaviour. Siegel and Associates (1976) The Methods and Materials of Demography. and representation. Schattschneider. E. and Building Democratic Institutions: Party Systems in Latin America (Stanford University Press. (1967) ‘Parties and the Nationalization of Electoral Forces’. Sartori. 1997). Indiana 46556-5677. Giovanni (1976) Parties and Party Systems: A Framework for Analysis. ADDRESS: Kellogg Institute. Notre Dame. TX: Arnold Foundation. CT: Yale University Press. East Lansing. New Haven. [email: smainwar@nd. He is currently conducting an NSF funded study (with David Samuels) of split-ticket voting in presidential democracies. Stepan. (1960) The Semi-Sovereign People: A Realist’s View of Democracy in America. Shryock. JONES is Associate Professor in the Department of Political Science at Michigan State University. New York: Academic Press. Sundquist. (Multi)Nationalism.: Brookings Institution. in Alfred Stepan. Jacob S. Stokes. 166 . Dallas. D.. and Democracy: Beyond Rikerian Federalism’. Washington. Oxford: Oxford University Press. New York:] Paper submitted 10 August 2001. Rinehart and] SCOTT MAINWARING is Director of the Kellogg Institute for International Studies and Eugene Conley Professor of Government at the University of Notre Dame. (1973) Dynamics of the Party System: Alignment and Realignment of Political Parties in the United States.) Mathematical Applications in Political Science. (1965) ‘A Variance Components Model of Political Effects’. in William Nisbet Chambers and Walter Dean Burnham (eds) The American Party Systems: Stages of Political Development. Donald E. Claunch (ed. Stokes. His latest books are Rethinking Party Systems in the Third Wave of Democratization: The Case of Brazil (Stanford University Press. Michigan 48824-1032.C. New York: Oxford University Press. New York: Cambridge University Press. Comparative Political Studies and the Journal of Politics. 1995).PA RT Y P O L I T I C S 9 ( 2 ) Competition in Western Democracies’. Rein and Matthew Soberg Shugart (1989) Seats and Votes: The Effects and Determinants of Electoral Systems. Henry S. CA: Sage. Alfred C. ADDRESS: Department of Political Science. co-edited. 1999). MARK P. James L.msu. accepted for publication 25 January 2002. Presidentialism and Democracy in Latin America (Cambridge University Press. USA.Jones@ssc. His research focuses on the manner in which electoral laws and other institutions influence party systems. University of Notre Dame. His recent publications have appeared in American Journal of Political Science. in Hans Daalder and Peter Mair (eds) Western European Party Systems. Donald E. co-edited. Arguing Comparative Politics. [email: Mark. in John M. Beverly Hills. E. Michigan State University. Taagepera. USA. (2001) ‘Toward a New Comparative Politics of Federalism.