Time after time: party organizational strength in new and old democracies

The ‘time factor’ has not been systematically considered in cross-national studies on party organizations. Relying on the largest dataset to date on party organizations, namely the Political Party Database Project (PPDB), the article tests the impact of time as a two-level variable (duration of democracy and age of parties) on parties’ organizational strength in new and established democracies. We add original data from three Latin American countries to the nineteen countries covered by the first PPDB database (132 parties overall). The results suggest that parties in established democracies have less members and more money than those of newer democracies. Among the latter, the greater capacity for mass mobilization produces stronger parties—as in Latin America—compared to the Eastern European countries. The findings challenge the traditional view of the exceptional weakness of Latin American parties and point to the importance of time as a multilevel variable: besides the national context, the “ancestral” party origin in previous regimes have a large impact on organizational strength.

quite surprising, considering that the time and conditions of party existence are essential elements in the processes of party institutionalization (Panebianco, 1988;Casal Bértoa, 2017). While organizational strength and party institutionalization are empirically and conceptually distinguishable, we should expect-relying on the literature on the latter-the time component to be an important factor behind organizational strength (Harmel and Janda, 1994;Dix, 1992;Harmel, Svasand, and Mjelde, 2016).
On the other hand, there is still a gap in comparative and systematic analyses on party organizations in new democracies, with the lack of solid empirical basis (longitudinal and cross-national data), notably about Latin American and Asian parties. In Latin America, contemporary research on parties focuses mostly on the party system level, especially regarding the issue of institutionalization (Carreras, 2012;Mainwaring, 2018). Analyses on political parties are mostly case-oriented, centered on specific countries, parties or dimensions, and party organizations are still "black boxes" in the region (Levitsky, 2001).
In one of the pioneering studies about Latin American parties, Dix (1992) argued that the parties of the new third-wave regimes presented some signs of greater institutionalization compared to the generation prior to the coups d'état of the 1960s and 1970s. Since then, however, little progress has been made in systematically measuring the strength of political parties in the region with broad cross-national perspectives 4 .
Despite that, and despite Dix's optimistic evaluation, the literature has taken for granted that Latin American parties are still organizationally weak (Mainwaring and Shugart, 1997;Levitsky et al., 2016, p. 1-3). However, one question was never clearly answered: to what are they "weak" in comparison?
Following Sartori's (1970, p. 1.035) admonishment that "to compare is to control," this article tests the influence of time on party organizational strength and performs an objective 'proof of the pudding' about party strength in Latin America. Assuming that it takes time to build strong organizations (Panebianco, 1988;Tavits, 2013), we ask: are the parties in newer democracies weaker than the parties in established democracies? Are there substantial differences between the parties in Latin America and in other regions? After controlling for the national level (duration of democracy), would the time variable have a substantial impact on organizational strength also at the party level-hence "time after time"? These research questions are answered by using strategies and indicators that can 'travel' in cross-national and cross-regional comparisons, developed by the Political Party Database Project (PPDB) (Scarrow, Webb, and Poguntke, 2017). We add original data from three Latin American cases (twenty parties) to the nineteen countries covered by the first round of data released by the PPDB (Poguntke, Scarrow, and Webb, 2017). Of course, Brazil, Chile, and Mexico are not representative of all the Latin American party systems. However, they may be the most consolidated systems in the region (Mainwaring, 2018), and this is a first attempt to move beyond the 'usual suspects' and to include Latin America in a comparative cross-regional study on party organizations, which justifies the selection of three cases for which we have more reliable data. In order to expand the coverage towards quite different contexts, we need to rely on less demanding empirical indicators and data-a trade-off faced by all the scholars studying comparative party organizations (Janda, 1980;Katz and Mair, 1992;Scarrow, Webb, and Poguntke, 2017). Therefore, the data on party membership and finance, for the period 2011-2014, are summarized in a simple additive index: the Party Strength Index (PSI) (Webb and Keith, 2017). Overall, we rely on a database with 132 parties in 22 countries, covering third-wave democracies, post-communist countries and advanced democracies. By doing this, we resume the tradition of a "most different" approach to party comparative analyses (Janda, 1980), with both cross-national and cross-regional perspectives.
This article has two components, one rather descriptive and the other more explanatory. The purpose in the descriptive approach is to expand the coverage of comparative research on party organizations towards Latin America. We do that by gathering and presenting original data on the three indicators separately (party membership, party funding, and party strength), with aggregations by country, region, and new and old democracies and by party families in Latin America. With simple exploratory techniques (correlations and tests of mean differences), some inferences are already possible even in this descriptive component. In the multilevel explanatory dimension, contextual factors (region, duration of democracy, and socioeconomic control variables), as well as party-level variables (origins and time of existence of each party), are included in multivariate models to explain the differences between parties in terms of membership, funding, and strength. In both components, we test three specific hypotheses that rely on previous findings about party membership, party finance, and organizational strength.
As emphasized by Bartolini and Mair (2001, p. 327), "any discussion of the role and importance of political parties in new or developing democracies is unlikely to ensue without reference to their role and importance in the more established democracies." More recently, Webb, Poguntke, and Scarrow (2017, p. 317-318) argued that it is time to "move the field beyond its traditional heartland in most-similar-system comparative studies of practices in parliamentary regimes. As the ranks of long-term presidential democracies grow, it seems theoretically indefensible to exclude Latin American, Asian and African regimes from cross-national comparative studies of party organization." Based on systematic empirical indicators, we are better able to test the relative strength of Latin American parties using both cross-national and cross-regional approaches, moving from a perspective rooted in ideal type models (as Duverger's mass party) to one rooted in specific indicators and dimensions.
With those considerations in mind, the article proceeds as follows. The next section briefly discusses the concepts and measures of party strength and the efforts for broad comparative research on party organizations as well as presents the hypotheses. The

TIME AFTER TIME: PARTY ORGANIZATIONAL STRENGTH IN NEW AND OLD DEMOCRACIES
following section introduces the cases, data, and research design. The section 'Findings' presents the data on party membership, party income, and party strength and tests the hypotheses. In the final two sections, we evaluate the hypotheses and discuss the results.
The findings show that parties in established democracies have less members and more money than those of newer democracies. However, much of the variation can be explained by the level of socioeconomic development and the parties' historical origins: traditional democracies are richer (and so are their parties) and have older parties than new democracies. Among the newer democracies, mass mobilization partially compensates for the lack of money in Latin American parties. Overall, time proved to be a strong predictor of party strength, and more significant at the party than at the national (systemic) level.
The "ancestral" party origin in previous regimes has a large impact on the organizational strength, larger than the impact of duration of democracy or the impact of party age in the current system. The findings challenge assumptions about the exceptional weakness of Latin American party organizations and open several avenues for further research.

Party strength in new democracies
Why should we care about party organizational strength? Strong parties are especially important for new democracies (Van Biezen, 2003;Tavits, 2012) and for the institutionalization of party systems (Mainwaring and Torcal, 2006). Strong parties are more resilient and adaptable to environmental challenges than weak parties, and they are better able to generate stability in the structure of competition: they offer information shortcuts to voters and are more efficient at attracting and sustaining their support over time. Stronger parties are also more effective in formulating policy and are more accountable, since they facilitate the clarity of responsibility (Tavits, 2012, p. 84-85;Lupu, 2016).
The most recent survey data show that about a third of Latin American voters have party linkages-a surprising figure given the low levels of institutionalization of party systems in the region. There is great variation across countries, from 14% of party ID in Chile to 61% in the Dominican Republic; Mexico and Brazil have 30% and 28% of party ID respectively (Nadeau et al., 2017, p. 69) 5 . These figures are similar to those found in other newer democracies, such as Poland (27%), the Czech Republic (31%), and Hungary (44%); however, they are below those found in third-wave developed countries, such as Portugal (61%) and Spain (55%) 6 .
Party identification figures as the most significant explanatory variable for voting behavior in several of the recent studies about Latin America (ahead of sociodemographic factors and left-right self-placement), consistent over time and across all countries (Carreras and Castañeda-Angarita, 2014;Carlin, Singer and Zechmeister, 2015;Nadeau 5 Data from the Americas Barometer (LAPOP: Latin American Public Opinion Project), with surveys carried out between 2008 and 2012. See Carlin, Singer and Zechmeister (2015) and Nadeau et al. (2017). 6 European Social Survey, 2014Survey, . et al., 2017. Partisanship in the region is correlated with age, civic engagement, belief in the efficacy of political action, and access to information, which form a profile quite similar to that of advanced democracies (Lupu, 2015).
However, although party identification has been an important measure of citizens' attitudes towards political parties, party organizations are not reducible to the perceptions and behavior of voters and political elites (Panebianco, 1988;Cotter et al., 1989, p. 5-6).
Party membership, party funding, and other objective dimensions of party organizations are still understudied subjects in the region (Levitsky, 2001). One reason for that is the difficulty in gathering data. Data on party membership in Latin America vary widely in terms of reliability, and countries with stronger and more independent electoral bodies (such as the three cases presented here) have better and more systematized information.
These official records show high levels of heterogeneity across the countries in the region, from about 6% of national membership (members/electorate ratio) in Peru and Chile to 60% in Panama (Došek, 2014). Regarding party funding, the predominant model in the region is mixed: parties have access to both public and private resources. Latin American parties are, in general, strictly regulated by the state, and these regulations have expanded to the supervision of intraparty finances. The progressive empowerment of control mechanisms in some countries has made it more feasible to use official records about party membership and finance to assess the party strength in the region (Molenaar, 2017;Londoño and Zovatto, 2014) There is no consensus on the dimensions and indicators to measure party organizational strength (Tavits, 2013, p. 16). The more specific and empirical-oriented literature on the subject took its first steps in the 1980s, evaluating the impact of organizational strength on other variables such as electoral success and legislative behavior (Gibson et al., 1983;Coleman, 1996;Cotter et al., 1989;Tavits, 2013). In a pioneering effort, Gibson et al. (1983) suggest a concept of organizational strength around two dimensions: organizational complexity (bureaucratization) and programmatic capacity (voter support). Subsequent cross-national analyses have listed some alternative dimensions to measure party strength and to test its impact over other political outcomes and processes (see Janda, 1983;Janda and Colman, 1998).
In analyzing the parties in four post-communist countries, Tavits (2012Tavits ( , 2013 employs three dimensions to measure organizational strength: 1) professionalization of the central organization (staff size); 2) territorial extensiveness (local presence); and 3) membership size and activism. She points out that the parties are not homogeneous in terms of the investment made in their organizations, even in the face of similar institutional and social settings (Tavits, 2013, p. 197). The environment is relevant, but not determinant to understand the organizational capabilities-which confirms findings about advanced democracies (Harmel and Janda, 1994;Gauja, 2017;Scarrow, Webb, and Poguntke, 2017). Stronger organizations achieve better electoral results and have a more cohesive legislative behavior (Tavits, 2012;. Challenging the recurring diagnosis about the general weakness of party organizations in new democracies, Tavits (2012Tavits ( , 2013 concludes that this dimension is an important explanatory factor in their respective national contexts. The "time factor" is the most common dimension used to explain differences between parties in terms of party strength and (mainly) party institutionalization-not always in a systematic way though (Janda, 1980;Panebianco, 1988;Dix, 1992;Harmel and Janda, 1994;Mainwaring and Scully, 1995). For the purpose of this analysis, we argue that the "time factor" is a multilevel variable and must be conceptualized and measured systematically, in both contextual (national) and party levels. In theoretical terms, this approach is based on Panebianco's (1988) concept that variations in party organizations usually respond to two major (and non-excluding) factors: (1) party elites make important organizational choices based on a "bounded rationality," limited by the perceptions about the environment; (2) similarly to a biological organism, party organizations usually assume several conformations in each stage of their development (or lifetime), in face of different life-cycle dilemmas (see also Harmel and Janda, 1994, p. 262).
At the contextual level, a long period of stable and competitive democracy encourages party elites and party members to invest resources (money, labor, time etc.) in the party organization building. Therefore, we should expect that duration of democracy and institutional stability to be important factors that strengthen party organizations (Harmel and Janda, 1994). At the party level, the "party age" is the indicator most commonly associated with party institutionalization, "an objectively established survival record" (Harmel, Svasand, and Mjelde, 2016, p. 7). It takes time to stabilize procedures and routines, to build party attachments, to develop autonomous bureaucracies, and to expand the territorial presence of party organizations (Dix, 1992;Van Biezen, 2003;Tavits, 2013, p. 8-9). In Latin America, party age is also associated with higher levels of mass mobilization and stronger societal roots (Mainwaring and Scully, 1995, p. 15).
In short, the party age has been considered a proxy for party institutionalization (Casal Bértoa, 2017, p. 419), chiefly when other indicators and data are not available in large-N cross-national studies. Nevertheless, the empirical connections between time and party strength have remained virtually unexplored until now (Harmel and Janda, 1994;Dix, 1992;Harmel, Svasand, and Mjelde, 2016).

The PPDB Project
The debate on party organizational strength has not always been linked to the broad collaborative efforts that have tried to gather cross-national data on party organizations. Among the pioneering comparative research on the subject, adopting both cross-party and cross-national perspectives, Kenneth Janda's (1980) study-which covered 158 parties in 53 countries-stands out, followed by a few others such as Von Beyme's (1985), Ware's (1987), and Katz and Mair's respective works (1992).
As noted by Janda (1980;1983, p. 171) and Ware (2011), the few efforts in broad comparative research on party organizations have always suffered from the low standardization of concepts and indicators and the tendency for parochialism. According to Janda (1980, p. XIII) The dominant tradition within comparative politics has been overly conservative. Students have been cautioned to limit their comparisons to "things that are similar" rather than to things that are "different. The Political Party Database Project (PPDB) is an important step to overcome these difficulties. Its main purpose is to compare party organizations based on middle-level theorizing and hypotheses, around three dimensions: 1) structures and distribution of internal influence; 2) human and financial resources; 3) representative strategies, linkages, and participation regarding groups and individuals. The project expands the coverage of this type of study towards recent democracies and follows a comparative and empirically oriented approach. Its main concepts recover dimensions of previous studies on party organizations. However, there is no rigid or predefined interaction between those dimensions (Scarrow and Webb, 2017, p. 6-7). In this sense, the dimensions encompass various types of parties, with different origins, rules, resources, internal practices, and environmental contexts. A crucial advantage of the project's framework is moving from ideal types and evolutionary models to dimension-oriented analysis, presenting a common conceptual vocabulary that can "travel" to other regions and institutional settings-for example, to the presidential regimes that prevail in Latin America (Poguntke, Scarrow, and Webb, 2016;Scarrow, Webb, and Poguntke, 2017).
In addition, this cross-national research provides a specific indicator for measuring organizational strength. As is often the case in large-N comparative studies, the use of a relatively simple composite index-the Party Strength Index (Webb and Keith, 2017)inevitably entails some loss of particularities (number of attributes); however, it guarantees a significant improvement in extension, generalization, and explanatory power (external validity).
Relying on previous findings about party membership, party finance, and organizational strength, in the next sections we test three hypotheses: After controlling for the duration of democracy and other contextual variables, we expect that the party age (foundation) and the party origin (both expressed in years) will have a positive impact on the party strength.
With an exploratory approach, we also test if left-wing parties in Latin America are organizationally stronger than conservative/right-wing parties, reproducing the pattern found in other regions, particularly in the Western European democracies (Webb and Keith, 2017).

Cases and data
The first round of the PPDB dataset ( Chile, Brazil, and Mexico are stable democracies with wide variation in terms of democratic trajectory and party system institutionalization since the early 1990s. Chile 8 There are differences between authoritarian and totalitarian legacies. According to Van Biezen (2003, p. 115): "Most parties attribute the lack of large memberships to the legacy of the past, the contention usually being that four decades of communism have given the political party a negative image among the public at large, which is said to discourage people from affiliating to, let alone participating actively in, such a party". 9 We are grateful to Alexander Tan for calling the attention about this point. 10 See <http://www.politicalpartydb.org/>. was, and continues to be, one of the few cases of an institutionalized party system in the region. Brazil has experienced a gradual process of institutionalization since then, while Mexico only completed the regime transition as its party system became more open and institutionalized (Mainwaring, 2018).
The selection also follows pragmatic reasons: in contrast to the increasing availability of data from cross-national public opinion surveys, researchers still face enormous challenges in collecting reliable data on party membership and finance in the region (Došek, 2014;Londoño and Zovatto, 2014). Due to the efficiency of their electoral control bodies 11 , Brazil, Chile, and Mexico are the countries with the highest availability and reliability of data on party organizations in the region.
In order to undertake cross-national and cross-regional comparisons, expanding the coverage to Latin America, we followed the same research design used by the PPDB project (Scarrow, Webb, and Poguntke, 2017). Considering the effective number of parliamentary parties 12 , we selected ten parties in Brazil, five in Mexico and five in Chile: these are the major parties in each country, based on the results in the last two general elections (see the Appendix for a full list of parties by country). According to the six party families employed by the PPDB, and following the classification of the MAPP project 13 , two of the twenty Latin American parties can be classified as liberals, nine as Christian democrats/conservatives, and nine as social democrats, socialists, or greens. Because of the small number of cases in the first category (and considering that liberalism in Latin America is associated with right-wing policies and parties), we recoded the parties into two major blocs: center-left (social democrats/greens) and center-right (liberals and Christian democrats/conservatives).
For the analysis of the "time component," we also build on the database and the criteria of the MAPP project (Van Haute, Paulis, and Sierens, 2017). Two potentially different dates are considered for each party: (i) the date of foundation (or reorganization) in the current democratic regime (party age); and (ii) the date of origin (party origin), which refers to the initial appearance of the party, even with other names and/or in previous (including authoritarian) regimes. For the variable about the duration of current democratic regimes, we adapted the data from Cheibub, Gandhi, and Vreeland (2010).
We gathered official data on national party membership and the finance of national parties' head offices provided by the parties themselves or by the electoral control bodies. The Party Strength Index (PSI) is a quite straightforward composite index. It is calculated by adding the z scores of two indicators: the M/E ratio and the party income / electorate ratio. As the z scores are calculated based upon the mean and standard deviation of the whole sample, we re-calculated the z scores for all 132 parties. Thus, the PSI scores presented in this article are slightly different from those found in Webb and Keith (2017) (see all the party scores on the Appendix, Table B).
Besides the relations hypothesized above at the national (H1), regional (H2), and party levels (H3), we cannot ignore that some contextual variables can also have an impact on the two dimensions that form the PSI (members and money). In the next sections, these dimensions are always controlled for the polity size (electorate), following the procedures of the PPDB (Webb and Keith, 2017). Additionally, it is reasonable to expect that wealthier countries have richer parties, and that party organizations can be more dependent on labor-intensive activities in the less developed countries (Norris, 2002;Tavits, 2013). Therefore, we use the socioeconomic development (GDP, GDP per capita and the Human Development Index) as control variables in the models.
One may still ask how much official data reported by parties to state organs are inflated and whether they can accurately indicate the participation of members within the parties. First, when we work with cross-national comparative research on party organizations, there is no better alternative: "(…) any figures, even if inflated or crudely estimated, are better than none" (Mair and Van Biezen, 2001, p. 8) 15 . In large-N comparative analyses, there are always problems with the data; however, they tend to be randomized, and should not impede one from finding patterns and associations between the variables. Questions can also arise about the importance of parties and party members in Latin America. As exposed in the previous section, the levels of partisanship in the region are significant, and the party attachment has influenced other political outcomes. Besides, the few studies on the subject have indicated that these are not fictitious members: there are active members in all the major parties of the region. The members' profiles are quite similar to those found in advanced democracies: compared to the general population, they are older and wealthier, more informed and more educated, and more sensitive to civic norms; they also have positive attitudes toward institutions and the efficacy of political action (Ponce, 2013;Došek, 2014;Ribeiro, 2014;Lupu, 2015;Speck, Braga and Costa, 2015). 15 "The choice for the analyst is therefore either to accept at more or less face value those figures that are made available by the party organizations (…); or simply to do without, and to accept that little meaningful work can be done on party memberships on a comprehensive cross-national basis. (…) in common with strategies adopted by previous researchers in this field, we have opted for the first alternative" (Mair and Van Biezen, 2001, p. 8).

The Latin American cases
Chilean democracy was restored in 1990, after the defeat of Pinochet's dictatorship in the 1988 plebiscite. The country has a presidential and unitary state, with an asymmetric bicameralism. The Senate has strong legislative powers, but a reduced capacity to oversee the executive when compared to the lower house. As in other multiparty democracies, the distribution of cabinet positions among government parties is decisive in explaining the success of the executive branch. The 1980 Constitution increased presidential powers visà-vis the legislature, and the presidents have controlled the legislative process (Valenzuela, 1995;Angell, 2007).
Chile had a binomial electoral system until 2013: each district elected two senators and two deputies, and each citizen had only one vote. Parties or party coalitions could present lists with two candidates for each position (Valenzuela, 1995;Angell, 2007). This arrangement was replaced by a proportional system in 2015.
Chilean party system before Pinochet's regime was organized around structural cleavages-religion (Catholicism and secularism) and class. After the dictatorship, parties' positions in relation to the authoritarian legacy became an important issue. The major leftwing parties are the PS and the PPD, while the most important centrist party is the PDC.
The three parties constituted the hard core of the center-left coalition that ran the country until 2010 (the Concertación). On the center-right, UDI and RN form the Alianza coalition (replaced by the current Chile Vamos), which won the presidency for the first time in 2010.
In 2014, the new center-left coalition (Nueva Mayoría) regained the control-but lost again in 2017. Compared to other Latin American countries, Chilean party system is usually described as highly institutionalized, with relevant social anchorage (Valenzuela, 1995;Angell, 2007;Carreras, 2012).
Brazil's institutional setting was designed during the democratization process that started in the late 1970s, combining federalism, the presidential system, bicameralism, open-list proportional voting for legislatures, and a multiparty system. Seats in the lower house are distributed between parties or party coalitions, and the most voted candidates in each list are elected. The coexistence of several consensual elements (Lijphart, 1999) was considered problematic: the multiplication of veto players would lead the country to instability and institutional deadlocks (Ames, 2001). Nevertheless, Brazil was governable until the 2015 presidential crisis, and the party system has developed a significant degree of stability over the period (Mainwaring, 2018). As in Chile and Mexico, the president has extensive institutional mechanisms to induce cooperative behavior in the legislature. Parliamentary parties are more cohesive and disciplined than expected, and they have been central actors in Brazil's "coalition presidentialism" (Cheibub, Elkins, and Ginsburg, 2011;Raile, Pereira, and Power, 2011;Palermo, 2016).
Despite the extreme fragmentation of the party system in the legislature, a structure of competition organized around two blocs has been dominant in presidential Congress between elections were structured around this pattern until recently (Limongi and Cortez, 2010;Melo and Câmara, 2012;Braga, Ribeiro and Amaral, 2016;Meneguello and Arquer, 2018). The PT-PSDB dynamic was broken in 2018 when a far-right candidate from a small party, Jair Bolsonaro, won the presidency.
For most of the twentieth century, Mexico was dominated by a single party (PRI): the country was classified as a "hegemonic party system in transition" (Mainwaring and Scully, 1995). Political parties are firmly regulated by state party laws in Chile, Mexico, and Brazil, and they are compelled to formalize their membership structures. In Brazil, only party members can become candidates; in Mexico and Chile, the requirements for party registration and functioning have been linked to a minimum threshold of membership (Molenaar, 2017). The three countries adopted a mixed system for political funding for most of their current democratic periods. In Chile and Brazil (until 2015 for the latter) 16 , parties and candidates can raise private funds from companies and individuals, while in Mexico only individuals can donate. In the three cases, the state guarantees parties and candidates free access to television and radio, through tax exemptions to broadcasting companies. Parties in those countries usually receive direct public funding in proportion to the electoral performance (Molenaar, 2017;Ribeiro, 2013;Londoño and Zovatto, 2014).
This dependency on public funding varies widely across the three countries. The state dependency is extreme for the Mexican parties, while Brazilian parties were close to the average of the countries included in the PPDB in the 2011-2014 period (about 55% of public funding). In Chile, private funding prevails over public subsidies (Casal Bértoa et al., 2014;Londoño and Zovatto, 2014, p. 140;Poguntke, Scarrow, and Webb, 2016) 17 .

Findings: members, money and party strength
In  (2012) argue, Austria is a significant outlier among advanced democracies in terms of party membership. Following their procedures, we excluded Austria from some tests, as in the comparison between old and new democracies presented in Table 1.   Social democratic parties are richer than center-right parties in Latin America. 20 As is true of other analyses based on the PPDB data, it should be noted that this conclusion does not consider the longitudinal variation in the size of party membership. For analysis considering the "curvilinear pattern" of party age and M/E rates over time, see Van Haute, Paulis, and Sierens (2017).  2 also shows that parties in newer democracies are richer than those in more traditional democracies, when controlling for the national GDP. However, when controlling for the size of national electorates, parties in established democracies perform much better (last column). This indicates that the two factors must be evaluated together, through the GDP per capita. Figure 1 shows that this variable is a strong predictor of party income: the richer the country in relation to the size of its population, the wealthier its parties (r = 0.706, p < 0.01, adj. R² = 0.473).

Figure 1 Party income / electorate ratio and GDP per capita 2014, by country
Source: PPDB round 1a data (raw data) and official records for Latin American countries. The newer democracies are highlighted in black.
At the party level, the income is highly influenced by party age and party origin (r = 0.456 and r = 0.451 respectively, p < 0.01). The association between party origin and income is stronger in established democracies than in newer ones (r = 0.451 and r = 0.419, p < 0.01).
Finally, we evaluate the relative organizational strength using the Party Strength Index (PSI), which aggregates the z scores of M/E and party income/electorate ratios (Webb and Keith, 2017). Among the twenty parties in Latin America, eight have a PSI score higher than the median (-0.51), and only two parties are located in the last quartile of distribution (Figure 2 below). The PRI and PRD are the most powerful parties in Mexico, as well as the PMDB and PT in Brazil.

Figure 2
Party strength index in descending order, by party Sources: PPDB round 1a data (raw data, standardized by the authors) and official records for Latin American countries. See all the party scores in the Appendix, Table B. When we look at the average per country (Table 3 below  When we turn to the multilevel analysis, with party-aggregated data controlled for contextual factors (Table 5)  The models on Table 6 include dummy variables to test the differences between Latin America and Eastern Europe (three countries in each). When controlling for region and socioeconomic context, the duration of democracy is the main explanatory variable of party membership: the older the democracy, the lower the quantity of party members.

Discussion
The results presented in the previous section corroborate some of the expectations derived from the literature. Regarding our first hypothesis, the tests of mean differences show that parties in established democracies are stronger than those of newer democracies, primarily because they are much richer. However, when controlling for the socioeconomic context (GDP or HDI) and the party-level variables (party origin), this relationship become less pronounced. This finding suggests that duration of democracy should not be the only factor to be considered when analyzing party strength and party institutionalization in a comparative perspective. The socioeconomic context matters (Tavits, 2013).
Overall, older democracies have less party members. However, as they are the wealthiest countries, the party income partially compensates for the difficulties in mass mobilization-which confirms previous findings (Bartolini and Mair, 2001;Mair and Van Biezen, 2001;Scarrow, Webb, and Poguntke, 2017). Supporting the second hypothesis, the findings show that not all recently democratized countries perform in the same way.
Controlling for national GDP, party-level variables and size of the electorates, the parties in the three Latin American cases are stronger than the Eastern European parties, as the mass mobilization compensates for the lack of money only in the first group.
While differences between countries and regions are relevant, party-level variables also need to be considered. A second key contribution concerns the role played by the historical party origin: the organizational legacy is an important factor in explaining party strength. So far, only the party foundation under the current democratic regime has been included in analyses about party institutionalization or party strength (Casal Bértoa, 2017).
In all the tests, the "ancestral" party origin in previous democratic or authoritarian regimes Time is crucial for the consolidation of party organizations, mainly at the party level, which produces large within-country variations in both advanced and newer democracies (Dix, 1992;Tavits, 2013;Gauja, 2017). In the multilevel analysis, the effects of the socioeconomic context and the duration of democracy are mediated by the differences at the party level. Overall, time proved to be more important at the party level than at the national level. In all but one test, the effects of party origin proved stronger than the impact of duration of democracy (the exception is the M/E ratio in Table 6).
Therefore, the differences between new and old democracies can be explained, at least in part, by the differences in terms of party origin. Established democracies have many more older parties than newer democracies do: in the former, the averages are 70 years old for party origin and 60 years old for party age; in the latter, the averages are 44 and 26 years respectively. The findings suggest that, in comparative perspective, old parties in new democracies can be stronger than new parties in old democracies.
Finally, the data support the view about the bigger capacity (and willingness) of socialist and social-democratic parties to build strong organizations in Latin America (Ames and Power, 2007;Levitsky et al., 2016), which replicates certain findings about established democracies (Webb and Keith, 2017). Particularly about Latin America, the 11 parties in the database that could be classified as conservative/right-wing parties (according to the PPDB criteria) are poorer and mobilize less members than the nine social-democratic and socialist parties 21 . A systematic analysis about the influence of ideology and the patterns related to the party families can be a productive path for further research. 21 While party membership does not significantly oscillate in an interval as short as the period considered in the PPDB project , party funding is more unstable, as it depends on factors such as the electoral cycle (the raising of private and public resources for general elections), and the electoral performance (which influences the distribution of public money). One could note, therefore, that this difference may reflect a snapshot of a period in which left-wing parties were governing Brazil (since 2003) and Mexico (since December 2012). Besides, over almost three decades, the Chilean left-wing Concertacíon/Nueva Mayoría was out of power for only four years (2010)(2011)(2012)(2013)(2014). Definite results about this question would involve the gathering of data from more Latin American countries and over a longer period.

Conclusion
In short, the historical party origin can be considered a crucial dimension for the analysis of party organizational development and institutionalization. Why does time matter? It takes time to develop societal attachments and organizational complexity and autonomy, as well as to adapt to the environment (Panebianco, 1988). The particular conditions under which the parties develop should also matter. During the golden age of parties in the most traditional democracies (1950s and 1960s), parties that experienced the institutionalization process as opposition parties for a long time tended to build stronger organizations, when compared to parties in government (Panebianco, 1988, chapters 4-5). However, access to government and to public resources is more important for organizational strengthening and survival in contemporary democracies, which may have reversed this equation (Katz and Mair, 2009). The conditions of institutionalization and the microfoundations of the time effect on organizational change and strengthening (Pierson, 2004) should be explored in future research, probably in case studies with more qualitative methods.
The same generational dilemmas seem to affect the parties founded in the late 20th century: organizations built by (and for) the party in public office face, with strong state regulation, in a social and technological environment that did not require the construction of extensive communication channels with the electorate. This generational patternmainly at the party level-seems to be more relevant than regional distinctions (Van Biezen, 2003).
In this sense, the article shows the viability and importance of adopting crossregional perspectives in the study of party organizations in Latin America. The major difference between the Latin American cases and the PPDB round 1a parties is about party finance-which is expected given the low per capita income in the region. However, Chilean, Brazilian, and Mexican parties are better able to recruit members than Eastern European parties, due to the totalitarian legacy of post-communist contexts (Van Biezen, 2003, p. 37-38). Party members are still important: they may represent an alternative multidimensional resource to political parties in the poorest and most unequal democracies, where other resources (money) are limited (Norris, 2002;Scarrow, 2015).
Although highly discontinuous, parties in Latin America have historical trajectories that make the difference today. The results for Eastern Europe, on the other side, suggest that the legacy of the past does not always bring positive effects. The differences between the two regions are also connected to the particularities of the democratic transitions. While left-wing parties gradually entered the party systems in several Latin American countries (Levitsky et al., 2016), in Eastern Europe the totalitarian legacy had a negative impact on the consolidation of left-wing popular parties (Van Biezen, 2003). When we consider that socialist and social democratic parties put more effort into building strong organizations everywhere (Webb and Keith, 2017), these particularities also influence the results in the cross-regional analysis. Parties such as the PS in Chile, the PRD in Mexico, and the PT in Brazil made heavy investments in mass mobilization and organizational strengthening since (re)democratization.
Overall, the findings suggest that parties in Brazil, Chile, and Mexico are comparable to the 'usual suspects' normally included in cross-national research on party organizations.
As in the majority of the established democracies, parties in most Latin American countries operate under rigid state regulations and have formal members, formal processes, and executive and deliberative organs with perennial functioning (not only during elections), disposed in hierarchical structures (Molenaar, 2017). The results also challenge the assumptions about the "exceptional weakness" of party organizations in the region. Of course, the three countries analyzed here do not tell the full story of Latin American parties.
Following the PPDB criteria, we selected the electorally strongest parties, from three party systems with relatively high levels of institutionalization. In this sense, a next step would be the inclusion of other countries in cross-regional comparative studies, particularly those with lower levels of party system institutionalization, such as Argentina and Peru (Mainwaring, 2018).
This article shows that time matters. However, other factors should also be considered when analyzing the differences in terms of party strength. The presidential system-predominant in Latin America-does not seem to be a decisive factor in explaining the differences in terms of party strength in a cross-national perspective. This does not mean that the separation of powers is irrelevant to party organizations. As we argued about the time effect, this impact has never been systematically tested to date. As noted by Key (1964) long ago, parties in presidential systems face governing dilemmas when their presidential candidates succeed. The party in public office is split into two groups, the party's executive and legislative branches, which may conflict with each other regarding presidential appointments, policies, and strategies. As the parliamentary group does not have the power to dismiss the president, these conflicts can be quite severe 22 .
Besides the presidential system, other institutional factors (e.g. electoral system, party law, and federal arrangement) should be addressed in future research as possible drivers of variations in party strength across nations. At the party level, the conditions under which the party was founded (authoritarian versus democratic contexts), the participation in the national government, the electoral performance, and strategic choices made by party elites can also be hypothesized in future studies. On the other hand, the party strength may also be regarded as an independent variable, affecting voters' behavior (electoral performance, party ID, trust in political institutions, and support for democracy) and political outcomes, such as the policy-making and the legislative behavior (Scarrow, Webb, and Poguntke, 2017 Europe or in established democracies (Tavits, 2013;Little and Farrell, 2017) but also in Brazil (Ribeiro, Locatelli, and Assis, 2018).