It’s Been Mostly About Money! A Multi-method Research Approach to the Sources of Institutionalization

PSI: Conceptualization and Operationalization

Summarizing a discussion sketched out elsewhere (Casal Bértoa 2015), there is little agreement in the literature on how PSI should be defined. This is so because, with very few exceptions (e.g., Meleshevich 2007), most authors pay little attention to the notion itself and simply assume its multidimensional character. Still, and despite the ongoing discussion on what are its main elements, most conceptualizations of the notion clearly refer to one dimension, namely, stability in the nature of interparty competition (Lindberg 2007). For this reason,

and bearing in mind that the core of a party system is to be found in the patterns of interaction among its subunits (i.e. political parties) […], I consider PSI to be the process by which the patterns of interaction among political parties become routine, predictable and stable over time […]. (Casal Bértoa 2012:453)

An overview of the level of PSI between 1990 and 2010 in 13 Eastern European democracies ³ is displayed in Figure 1. The most evident conclusion derived from these summary data is that party systems in postcommunist Europe have institutionalized at different rates and in different ways (Casal Bértoa and Mair 2012). It is to explaining why this has been so that I will devote the rest of the article.

OPEN IN VIEWER

Figure 1.

Party system institutionalization (PSI) in postcommunist Europe (1990–2010). Note: The year of the “founding” elections is in brackets.

Sources of PSI: A “Comprehensive” Approach

When looking at the current literature on the topic, it is possible to identify up to 17 different factors which, either alone or in combination, have been considered essential when trying to explain PSI (Casal Bértoa 2012). However, the quasi-natural experiment produced by the dissolution of the Soviet Bloc in 1989 followed by the birth of newly independent and centralized states in the Baltics, the Balkans, as well as in former Czechoslovakia, allows me already at this early stage to exclude from the analysis two of them, namely, nature of state and time of transition. ⁴ Moreover, and in a similar vein, years of authoritarianism can also be left out as it clearly overlaps with previous democracy (PDEM)—especially when dichotomized. ⁵ From a methodological perspective, it seems obvious that the variance in the outcome cannot be explained by constant conditions (Przeworski and Teune 1970). For all these reasons, the following paragraphs will focus only on the remaining 14 conditions, presenting each of them in turn.

Relevant Factors (What?): MSDO/MDSO

As we have already seen, comparative political theory offers different possible (co-)explanations for the distinct levels of PSI observed in new and old democracies. In this article, where the number of possible combinations of conditions (214 = 16,384) clearly dwarfs the number of cases available for analysis (13), I will make use of De Meur and Berg-Schlosser’s (1994) MSDO/MDSO procedure, a technique particularly well suited as a prior step before using csQCA and, on the whole, extremely useful for systemic analyses which, like this one, present the so-called limited diversity problem. ⁶ It is in the name of parsimony and in order to avoid a simple description of cases—with one individual explanation per case—that a solution to this problem needs to be found before proceeding with any QCA-type analysis (Berg-Schlosser and De Meur 2009:27).

The idea is, thereby, that by carefully matching all the cases (i.e., party systems) under study across the different (potential) explanatory factors ⁷ found in the institutionalization literature, using a stepwise elaboration of distance matrices and dis-/similarity graphs (see Online Appendix available at http://whogoverns.eu/biography/publications/), I can identify the most similar pairs of cases with a different outcome as well as the most different pairs of cases displaying a similar outcome (Rihoux 2006:688). This will allow me to reduce the number of conditions to the minimum and, therefore, to be able to achieve a less complex comparison which, without any preconceived ideas, focuses on those relevant factors that might account for the different degrees of systemic institutionalization observed (De Meur and Gottcheiner 2009:215).

Bearing in mind that we have 14 possible explanatory factors, and following the logic of the MDSO/MSDO procedure (De Meur and Berg-Schlosser 1994; De Meur and Gottcheiner 2009; De Meur, Bursens, and Gottcheiner 2006), I have first clustered the different conditions into three rather homogeneous categories, namely, socioeconomic (A), historic-structural (B), and systemic-institutional (C; see also Casal Bértoa (2012:455, 472)). Second, and because the criteria used to calculate the distance between (two) factors are based on Boolean algebra, all conditions need to be dichotomized (De Meur et al. 2006:69). This is done according to the criteria established in Table A1, ⁸ which displays not only the threshold for the dichotomization of both the conditions and the outcome but also the sources according to which such thresholds are established. The result is a data matrix (Table A2) featuring 13 cases (seven positive/institutionalized and six negative/noninstitutionalized), and 14 Boolean conditions where 1 indicates presence and 0 stands for absence.

Once these operations have taken place, and before proceeding with any further comparison, it is essential to identify which pairs of cases are the most similar and which the most dissimilar. For that, it is necessary to build and synthesize distance matrices within and across categories (De Meur et al. 2006:75), as cases can be similar in one category (e.g., socioeconomic) but dissimilar for another (e.g., systemic-institutional). ⁹ In order to do so, I make use of the software (beta version 8/7/2006) developed by De Meur (available at http://www.jchr.be/01/beta.htm), which helps to select which cases share the smallest number of same-valued conditions and identical outcomes (MDSO pairs) and the smallest number of different-valued conditions and different outcomes (MSDO pairs).

Table A3 summarizes the levels of dis-/similarity for each pair of cases at different levels of requirement (more or less demanding) within each outcome, ¹⁰ namely, most different with a positive outcome (MDSO+ in zone 1, in blue); most different with a negative outcome (MDSO—in zone 2, in purple); and MSDO (zone 3, in yellow). The higher the added value (integer), the more dissimilar (blue and purple zones) and/or similar (yellow zone) the cases are and, therefore, the more valuable the comparison. ¹¹

The pairs of cases selected by this process are then aggregated in three dis-/similarity graphs (Figures A1 to A3), with different levels of dis-/similarity illustrated by continuous (higher) and dotted (lower) lines. On the basis of these three graphs/figures, I then proceed to compare ¹² the most dissimilar but institutionalized party systems, namely, Romania and the Czech Republic (integer = 22,222). Of the initial 14 conditions, only PCON and PFUND are present in both cases and, therefore, can be considered to explain the (presence of the) outcome. Adding to the comparison the countries with the second and third highest integer (Slovenia and Hungary, respectively) reduces the number of similar—hence relevant—factors to just one (PFUND), although PCON is still present in three of the four cases (just as POLAR). At a lower level of dissimilarity (integer = 12,222), the comparison of Ukraine with Hungary yields two (similar/relevant) conditions: CCUM and again PCON, although the latter becomes irrelevant once Slovenia is added to the comparison. A third comparison (Slovakia vs. Croatia) highlights WEALTH, POLAR, and, again, PFUND.

A comparison between the two most dissimilar noninstitutionalized party systems (i.e., Latvia and Bulgaria; see note 14.) yields three (similar/relevant) conditions absent in both cases, namely, WEALTH, PI, and PFUND. However, the inclusion of Serbia (a less dissimilar country) in the comparison allows for the exclusion of the last two (present in the Serbian case).

Serbia and Romania are the MSDO cases (integer = 13,333). Indeed, they are similar in every condition but two: ERCON and PCON. When other less similar cases (Ukraine, Croatia, and Slovenia) are added to the comparison, only PCON seems to keep its relevancy. Interestingly enough, however, the comparison between Serbia and these three countries yields CCUM as an important explanatory factor. Considering the pair Czech Republic and Latvia puts the emphasis on WEALTH, PCON, and PFUND. The inclusion of Estonia in the comparison only confirms the previous results, although PFUND disappears once Poland is added to the comparison. Still, these four-by-four comparisons seem to yield four relevant conditions, namely, PCON, WEALTH and, to a lesser extent, PFUND as well as CCUM (present in three of the four cases). The latter three are also deemed relevant in a three-by-three comparison between Lithuania, Slovenia, and Hungary. In this case, the importance of PI and POLAR should not be forgotten. Finally, considering the pair Romania versus Bulgaria adds PI and PFUND to the analysis.

Once the MSDO/MDSO procedure is completed, it seems clear that the number of relevant factors can be reduced to just four, namely, parliamentary concentration, CCUM, economic wealth, and PFUND. Indeed, while the last two pop up in all three analyses (i.e., MDSO+, MDSO−, and MSDO), the first two reach a high level of significance in both MDSO+ and MSDO analyses. All in all, these four conditions appear up to five times (WEALTH) or more (PCON, CCUM, and PFUND), in contrast to other less relevant (just twice), and sometimes contradictory (see pair Serbia vs. Ukraine in Table A8) factors, namely, PI and POLAR. ¹³

All in all, it is only after reducing the number of possible explanatory factors by more than three-quarters that a methodologically manageable, and certainly less complex, analysis of the “causal link/s” between those four conditions and the outcome (and/or the lack of it) can be undertaken.

Causal Mechanisms (Why?): Congruence and PT

In order to know how the previously mentioned relevant factors affect the process of PSI, I will make use in this section of two well-known case-study methods. The first one, the so-called congruence method, will help me to understand to what extent variance in the level of PSI can be explained by variance in each of the above-cited conditions. This will be done by testing both the direction and degree of change in both the outcome and the various conditions at different points in time during the process of PSI (George and Bennett 2005:181-204, 486). In particular, the above-cited method is particularly suitable for analyzing phenomena which—like PSI itself—refer to processes which not only involve specific periodizations (e.g., elections, governmental changes), but are neither monotonic, unidirectional, or finite. It is in such cases that the congruence method reveals itself to be particularly useful, allowing not only for the analysis of PSI at the end of the process but also at any particular point in time during it.

The second (PT) will allow me not only to see if there really is a common causal mechanism ¹⁴ linking the conditions and outcome but also to specifically identify the “causal chain” leading from the presence (or the absence) of wealth and/or parliamentary concentration and/or CCUM and/or PFUND to party system (non-)institutionalization (Beach and Pedersen 2013:5). The idea is that, by breaking down the rather large process of systemic institutionalization into its constituent parts, I can more easily trace the process by which each of the abovementioned conditions have produced the (expected) outcome (Caporaso 2009).

Independently of the within-case method employed, the first step in any congruence/PT-first research design is always the selection of typical cases (see Beach and Rohlfing in press or Schneider and Rohlfing, in press). Because I am equally interested in understanding both PSI and its absence, I will proceed with what I have called a BCM. ¹⁵ By combining the advantages of both comparative and within-case methods, ¹⁶ the BCM allows researchers not only to explain opposite outcomes but also to capture the “causal mechanism” behind processes which, even if facing each other, do not necessarily mirror each other. In particular, the idea is that by choosing two cases which, sharing most of the conditions, totally diverge in both the “relevant” conditions and the outcomes, I will be able to “direct [my] attention to the ways in which they differ (Gerring, 2007:133-35)” (Tarrow 2010:234), examining in particular how such opposite “causal forces” actually work.

Going back to the cases and looking again at the MSDO/MDSO results (Table A11), it follows that two of the most similar party systems with totally opposite levels of institutionalization are Hungary (positive) and Lithuania (negative). ¹⁷

The fact that economic development is one of the most important determinants of PSI in postcommunist Europe does not come as a surprise. Indeed, when we look at the state of the economy as well as the degree of systemic institutionalization in both Hungary (Figure 2(a)) and Lithuania (Figure 2(b)) at the end of each electoral period we can observe a rather clear (positive and negative, respectively) relationship. Thus, while the state of the economy in Lithuania—on every single indicator—has never been as good as in Hungary, the degree of PSI in the latter has always been superior. In this context, it should be borne in mind that while Hungary had already started a process of (limited) economic liberalization in the second half of the 1960s, ¹⁸ Lithuania remained within the Soviet “administratively centralized planned economy” until the early 1990s. This gave Hungary, itself one of the most economically developed countries within the Soviet bloc, a clear advantage over Lithuania, where bad economic performance has remained one of the main triggers of cabinet turnover and party system instability.

OPEN IN VIEWER

Figures 2.

Economic development, legislative fragmentation, and party system institutionalization (PSI) in (a) Hungary and (b) Lithuania.

But the abovementioned differences are also visible within countries. Indeed, as follows from the figure on the left, systemic institutionalization only started to increase in Hungary once the state of the economy began clearly to improve in the second half of the 1990s. In particular, while the mixed signals of the early 1990s (i.e., gross domestic product [GDP] growth while high unemployment and inflation continued) did not help the structure of interparty competition to stabilize, a clear improvement in all these indicators from the time of the second free and fair elections onward seems certainly to have fostered the process of systemic institutionalization. That is at least until the second half of the 2000s when the latter stagnated immediately after the first signs of the global economic and financial crisis began to produce their effects (i.e., growth decline, inflation, and unemployment) in the country.

On the other hand, and with the exception of the first few years when the Soviet legacy had left the economy in such bad shape that the only alternative was improvement, each post-electoral government alternation in Lithuania has been preceded by a period of economic uncertainty. In fact, and notwithstanding the sound creation of employment until 2008, the overall economic tendency since 1998 has been that of general decline, with periods of growth and moderate inflation followed by important drops in the GDP (in 1999, 2004, and, especially, after 2007) and inflation (steadily after 1999). In parallel, and as expected, the Lithuanian party system has suffered a chronic process of deinstitutionalization since the late 1990s (Ramonaitė 2006).

As follows from the abovementioned MSDO/MDSO analysis, parliamentary fragmentation also needs to be considered as one of PSI’s most important determinants. The logic is that, as thoroughly explained elsewhere (Casal Bértoa 2012, 2015), by indicating the numbers (and strength) of “streams of interaction,” the number (and size) of parties winning seats in legislative elections clearly determines the likely tactics of partisan competition and opposition as well as government formation possibilities in a country. Moreover, because the number of parties has “mechanical predispositions” in the sense that it gives us information on certain functional properties (e.g., interaction streams, coalition potential), the relationship between party system format and institutionalization can be said to be “path-dependence” as it responds to the following pattern: “the greater the number of parties (that have a say), the greater the complexity and probably the intricacy of the [interactions will be]” (Sartori 1976:120, 173). In other words, when party leaders must follow maneuvers among a large number of parties, predictability and stability in the structure of interparty competition are obviously hindered.

Figures 2a and 2b, which display the scores of both parliamentary fragmentation and PSI at the end of each electoral period in Hungary and Lithuania (respectively), show the almost perfect relationship between the above-cited two conditions. Hence, while in the institutionalized Hungarian party system the “effective” number of legislative parties—constantly below four—has decreased over time (from 3.8 to 2.4), the Lithuanian party system has suffered from a continuous and parallel process of fragmentation (from 3 to 5.8) and deinstitutionalization. Moreover, and apart from this pronounced intercountry variation, another striking pattern revealed by these data is one that is also intuitively plausible: Within each country, parliamentary fragmentation and PSI rise and fall in accord, so when the former decreases the latter increases and vice versa. In other words, and confirming previous expectations, they fluctuate not only in the same direction but also to a similar extent.

Similarly, PFUND has also contributed to the institutionalization of postcommunist party systems as it has eased the continuity of existing political options while, at the same time, reducing “the impact of those seeking to challenge the political status quo” (Scarrow 2006:629). In other words, by discouraging the entry of new parties to the system and, therefore, keeping the number of (both electoral and parliamentary) parties rather low, publicly funded party systems have been able to guarantee the supremacy of already existing parties (Katz and Mair 1995:15) and, consequently, assure the stability and predictability of the structure of competition among them.

In order to test the abovementioned statements, Table 1 compares the two party systems at hand on the basis of the number of parties winning at least .5 percent of the vote as well as the share of parties winning less than 5 percent of the vote (Scarrow 2006). No matter at which indicator we look, it seems clear that “the model of Hungarian party funding [has…] help[ed…] to consolidate the party system” (Enyedi 2007:102). The argument that this is the case derives also from the fact that, as displayed earlier, both indicators clearly improved after the introduction of public subsidies for Lithuanian political parties in 1999, although not to the same levels as its Hungarian counterpart (see also Figure 2(b)).

OPEN IN VIEWER

Table 1.

Consequences of PFUND for PSI in Hungary and Lithuania.

Country	Period (Number of Elections)	Total Number of Electoral Parties	Small Party Vote Share	Party Survival Ratea
Hungary	1990–2010 (6)	9.2	9.7	94 (40)
Lithuania	1992–2010 (5)	14.8	20.4	NA (52.4)
	1992–1999 (2)	16.5	24.4	NA (73.5)
	2000–2010 (3)	13.7	17.7	95 (31.3)

Source: Casal Bértoa and Spirova (2013:33, 35).

Note: NA = not available; PSI = party system institutionalization; PFUND = party funding.

^aThe figure in brackets refers to parties deprived of State financial support.

In a similar vein, while the Hungarian model of public funding introduced from the very beginning a clear discrimination between publicly and nonpublicly funded parties, guaranteeing the concentration of the party system among a reduced number of political options; in Lithuania, such a “reductive” effect only started to take place after 1999, when a 3 percent “payout threshold” was introduced. In fact, and as follows from the last column in Table 1, while publicly funded parties in Hungary have managed to survive election after election, in Lithuania up to 1999 the average survival rate of political parties barely reached 73 percent. However, and as expected, this percentage started to increase from that year onward for publicly funded parties, while it decreased for all those unable to pass the “payout threshold.” As a result, a process of PCON was initiated among those parties deprived of public funds with the only aim of survival: For example, Liberal Centre Union (LCS) merged with Lithuanian Liberal Union (LLS) before the 2004 elections in order to form Liberal and Centre Union (LCiS); similarly, Lithuanian Union of Political Prisoners and Deportees (LPKTS) merged into Homeland Union (TS). Equally, in 2008, both Lithuanian Nationalist Union (LTS) and Lithuanian Christian Democrats (LKD), unable to cross the payout threshold in 2004, merged with TS (Casal Bértoa and Spirova 2013:19-20, 37).

But together with a country’s economy and institutions, sociological factors have also played an important role in the process of PSI.

The idea is that when cleavages are cross-cutting, parties will have difficulties in finding ideologically contiguous partners with which to cooperate, as being close in one dimension may be accompanied by irreconcilable differences in another. On the contrary, when cleavages are cumulative (i.e. coinciding), parties will tend to interact only with other parties within the same side of the cleavage, rejecting any cooperation that would lead them to cross such a line. (Casal Bértoa 2014)

Thus, and as follows from Figure 3, the cumulative character of cleavages in Hungary has enabled a division of the political spectrum into two very antagonistic (and stable) political camps: “a socially conservative, religious, somewhat nationalist, and anti-communist camp […] and […] a secular, morally permissive and generally less nationalist camp” (Tóka 2004:322; see also Enyedi 2006). The result has been a very well-institutionalized party system in which the structure of interparty competition has pitted again and again the political forces of the cosmopolitan, postcommunist and anticlerical “left” (mainly Hungarian Socialist Party [MSZP] and Alliance of Free Democrats [SZDSZ]) against the nationalist, anticommunist, and clerical “right” (basically Fidesz/Christian Democratic People’s Party [KDNP], Hungarian Democratic Forum [MDF], and Independent Smallholders’ Party [FKgP]).

OPEN IN VIEWER

Figure 3.

Political parties and cleavages structuration in Hungary* and Lithuania.

In clear contrast, the Lithuanian party system has been characterized since the very beginning by a cross-cutting multidimensional space of interparty competition revolving around two different types of cleavage: economic and urban/rural (Duvold and Jurkynas 2004), which have divided the political spectrum into four different politico-ideological fields (Figure 3):

Socialist (strong support of state interventionism and a cosmopolitanism),

Because the social protectionist camp (socialists + agrarians) differs from the pro-market camp (conservatives + liberals) in terms of the economy, while the urban camp (socialists + liberals) differs from the rural camp (agrarians + conservatives) in terms of cosmopolitanism, parties have found it very difficult to establish stable patterns of governmental and/or electoral cooperation. In fact, in almost 20 years of democratic politics, only the first (mono-color) Lithuanian government managed to unite all parties from the same political field.

Looking at the previous “congruent” analysis, it seems clear that there is an almost perfect—in time and degree—relationship between each of the abovementioned explanatory factors and PSI. Unfortunately, it does not tell us anything about the causal mechanism linking the former with the latter. For that a more in-depth PT analysis, “detailing each of the parts of the mechanism between X [here causal factors] and the outcome, focusing on how they transmit causal forces” is needed (Beach, in press).

Interestingly enough, and implicit in previous literature (Bartolini and Mair 1990; Birnir 2005; Casal Bértoa 2012; Tucker 2006), there seems to be a common causal mechanism linking, positively or negatively, each of the abovementioned factors with the process of systemic institutionalization. Figure 4 displays a tentative formalization of the mechanism, with the top illustrating the parts and the bottom the observable implications.

OPEN IN VIEWER

Figure 4.

A socioeconomic-institutional explanation of party system institutionalization (PSI): Causal mechanism.

As can be observed from previous text, the first part of the mechanism refers to the triggers, namely, the presence of economic development, party concentration, a cumulative cleavage structure, and/or public subsidies to political parties. At this moment, we should expect to find evidences of high GDP growth and/or low inflation/unemployment rates, a moderately low number of parties in parliament, low levels of cleavage cross-cuttingness in society, and a rather high level of political parties financially dependent on the State.

The second part shows that the electorate, in light of the abovementioned favorable conditions, will remain stable in their partisan preferences. We should then expect to find relatively low levels of change in the balance of power among parties: That is, they should be able to attract a rather similar percentage of votes again and again. This would help them not only to strength the levels of partisanship (i.e., identification, closeness, membership) in society but to routinize predictable patterns of (coalitional/cooperative) behavior among them.

For the third part of the mechanism, we should see that the partisan status quo remains almost unaltered election after election. The observable implications here may consist of few parties coming or going within the electoral spectrum. As a result, and at the end of the mechanism, we should detect a relevant degree of systemic turnover. In this context, we should expect to observe quite high levels of partisan continuity at the parliamentary level. ¹⁹

Finally, the outcome should be PSI. In practical terms, we should then observe at the time of government formation stable patterns of competition among all political parties in the system, in terms of alternation, formula, and access. Note here however that if the contrary is true for every single part of the mechanism displayed earlier, then a similar but opposite process leading to weak levels of PSI would be observed.

In order to observe how the causal mechanism unpacked earlier works in practice, we will make use again of the typical (positive and negative, respectively) cases of Hungary and Lithuania. This will allow me to avoid the analysis of part 1, already explained in depth earlier: both synchronically and diachronically. Regarding part3, the most straightforward way of looking at the degree of change in the balance of electoral power among political parties is to look at the Pedersen’s index of electoral volatility for both countries in the period here examined. While on average Hungarian voters’ volatility barely passes 23 percent, and is therefore one of the most “stable” within the postcommunist region, the Lithuanian electorate—with barely 40 percent—is considered to be the most unstable in the whole European continent (Casal Bértoa 2013:417). In fact, while until 2010 Hungarian voters’ preferences became steadily more stable (from 26.3 in 1994 to 8.4 in 2006) thanks to a rather stable economic situation (see Figure 2a), a clear reduction in the level of parliamentary fragmentation (from 3.8 in 1990 to 2.4 in 2006), a change from a tripolar cleavage structure to a bipolar one in 1994 (Enyedi and Casal Bértoa 2011:123, 127-28), and a rather liberal PFUND regime (Casal Bértoa and Spirova 2013:10); the level of volatility in Lithuania has remained rather high (always above 20 percent, with a continuous increase until 2004) during the whole period (Table A12), mainly due to the unstable (almost continuously declining) economic situation, an increasing level of fragmentation (from 3 in 1992 to 5.8 in 2008), a rather stable level of cleavage cross-cuttingness (see Figure 3), and a rather restrictive party finance regime, only introduced in 1999.

Did such stable/unstable environments in Hungary/Lithuania close/open the electoral market to new political forces, while maintaining/altering the status quo of “traditional” parties? A strong test of this part is to look at the number of new parties (NNP) entering the electorate after the second electoral contest (Table A12). In consonance with the steady decrease of electoral volatility, the NNP entering the Hungarian party system until 2010 suffered a significant decline over time (from 4 in 1994 to none in 2006). In a similar vein, the fact that none of the new parties has managed to become “relevant”—in the Sartorian meaning of the term—clearly illustrates the resilience of Hungarian “traditional” parties. In clear contrast, the average NNP in Lithuania more than doubles the same figure for Hungary: five versus two, respectively. Moreover, while it decreased from nine in 1996 to four in 2000 and again to three in 2004, it increased again in 2008 when four new parties entered the electoral arena. Two of these parties (i.e., Liberals’ Movement of the Republic of Lithuania [LRLS] and National Resurrection Party [TPP]) even became part of the governing coalition, something that had already happened both in 2000 and 2004 when the recently formed New Union/Social Liberals (NS/SL) and Labour Party (DP), respectively, entered the Paksas’ and Brazaukas’ cabinets as junior partners. On the other hand, various have been the “relevant” parties obliged to disappear due to their steady electoral decline, namely, LKD, LCS, or LLS (see earlier text).

Part 4 refers to the continuity (or not) of the main political options in the party system. The best way to test this fragment of the causal chain displayed earlier is to look at the number of parties entering and leaving the party system at each election. Using Toole’s “party system turnover” (PST) index (2000:450) to calculate the latter, it is possible to observe once again a clear contrast between the two countries here analyzed. In fact, not only is the average PST in Hungary much lower than in Lithuania (.4 vs. .6, respectively), but also it has been so in every single election but for one: 2006 (Table A12). Even more, while in Lithuania

the symptoms of the deepening crisis of traditional parties were already apparent somewhat [at the time of] the 1997/1998 presidential election [or even earlier as] the parliamentary election of 1996 was marked by an increased proportion of wasted votes (Ramonaitė 2006:84);

Regarding the outcome, was there a stabilization/destabilization in the structure of competition for government? As illustrated in Casal Bértoa and Mair (2012:95, 98, 103), and explained elsewhere (Enyedi and Casal Bértoa 2011:123-29), the patterns of interparty collaboration/cooperation in Hungary have been rather stable, especially since 1994 when a tripolar structure of partisan competition gave way to a bipolar one, pitting the parties on the left (MSZP and SZDSZ) against the parties on the right (Fidesz plus other minor conservative parties). On the contrary, “the Lithuanian party system appears to be in a state of flux,” especially after the parliamentary elections of 2000, a real “turning point in [its] development” (Ramonaitė 2006:71, 84). Indeed, it was at this time when the two-bloc confrontation (i.e., socialist vs. conservatives) was disturbed by the emergence—first with the liberals, later with the populists—of a tripolar structure of competition.

All in all, the previous findings confirm that PSI is far from being a “unidirectional or irreversible” phenomenon (Stockton 2001:95). In fact, and notwithstanding the specific status of a party system at a certain point in time, what clearly follows from the “congruence” analysis undertaken earlier is that variance in one or more of the conditions can modulate the degree of PSI over time. This is not to deny that, as follows from the PT analysis, there are also a number of specific forces—common to all conditions—which bounded together clearly affect the mode in which those conditions determine the degree and direction of PSI as a whole.

Causal Combinations (How?): csQCA

Now that we know the intervening causal process by which socioeconomic and institutional conditions are linked to PSI (or its absence) and bearing in mind that not all of them are present (or absent) in all party systems, I will try to discover how such conditions have combined in each of the postcommunist countries here analyzed. For that csQCA—a methodological technique dealing with a limited number of cases in a “configurational” way—constitutes the perfect tool (Beach and Rohlfing in press; Ragin 1987).

According to the “Standards of Good Practice in QCA” (Schneider and Wagemann 2010), the analysis of necessary and of sufficient conditions, with the former always going first, needs to be separate for the outcome and for its nonoccurrence.

Conditions for PSI

In terms of necessity, the analysis (Table A13) reveals that none of the four conditions, either in its presence or its logical negation, reaches the consistency threshold of .9 recommended in the literature (Schneider and Rohlfing 2013; Schneider and Wagemann 2012). However, PFUND with a consistency threshold of .86 comes close, anticipating the importance such a condition has for the explanation of PSI (see subsequently).

In order to perform the analysis of sufficiency, it is essential to elaborate a truth table based on the four conditions. As the “contradictions-free” truth table (Table A14) shows, 13 cases fall into nine truth table rows, the remaining seven rows are logical remainders. After including only those cases when the outcome (PSI) is present (raw consistency = 1), the “standard analysis” (Ragin 2008) is performed, limiting my interpretation to the so-called intermediate solution (Schneider and Wagemann 2012). Because this solution uses for counterfactual claims all those logical remainders that comply with the directional expectations on the single conditions, the latter need to be set before proceeding further with the analysis. Thus, and on the basis of the theoretical consideration explained earlier in this article, a condition is expected to contribute to PSI when WEALTH − PCON − CCUM − PFUND.

Taking all this setup into consideration, and after the information displayed in the truth table has been logically minimized, the csQCA analysis yields an intermediate solution term with three sufficient paths toward PSI (Table 2). Their relatively low unique coverage values indicate that there are several cases displaying PSI for more than one reason, namely, Hungary, Croatia, and the Czech Republic. There are, however, four uniquely covered cases, namely, Slovakia, Slovenia, Romania, and Ukraine. The overall solution term shows perfect consistency and coverage, meaning that we are able to explain PSI in all cases.

OPEN IN VIEWER

Table 2.

Pathways to Systemic Institutionalization (PSI).

Solutions	Raw Coverage	Unique Coverage	Countries
PFUND * WEALTH	.71	.29	Hun, Cro, Cze + Slk, Slv
PFUND * PCON	.57	.14	Hun, Cro, Cze + Rom
CCUM * PCON	.57	.14	Hun, Cro, Cze + Ukr
Solution consistency		1.0
Solution coverage		1.0

Source: Table compiled on the basis of the results obtained with fsQCA 2.5.

Note: CCUM = cleavage cumulation; Cro = Croatia; Cze = Czech Republic; Hun = Hungary; PCON = party system concentration; PFUND = party funding; PSI = party system institutionalization; Rom = Romania; Slk = Slovakia; Slv = Slovenia; Ukr = Ukraine; WEALTH = economic development.

According to the formula displayed earlier, and bearing in mind that poor economic development is a necessary (but not sufficient) condition for weak systemic institutionalization, it seems clear that party systems will always be institutionalized in rich countries. This is not to say that party systems in poor countries are condemned to be underinstitutionalized. On the contrary, PSI will take place in poor countries provided that, together with a low number of parliamentary parties, they make available public funding for political parties or cleavages structure in a cumulative way.

Conditions for psi

Using the same conditions as before, but bearing in mind that causation is not essentially symmetric, I proceed now to analyze why some party systems have remained under-institutionalized during the period here examined.

Interestingly enough, the analysis of necessity reveals that poor economic development (wealth), with a consistency score of 1, is a necessary condition for weak systemic institutionalization. This has important implications for the following sufficiency analysis as I will need to block any logical remainder displaying the presence of the condition (WEALTH). This finding does not come, however, as a surprise. In fact, taking into consideration the literature on the topic as well as bearing in mind the “causal mechanism” explained earlier, the economy is the only condition producing both a closer (i.e., on the demand side) and short-term effects on the electorate. ²⁰

As far as the analysis of sufficiency is concerned, and after imposing a frequency threshold of one and a raw consistency threshold of 0.8, I specify the following directional expectations: wealth-pcon-ccum-pfund. This yields an intermediate solution term consisting again of three paths containing the necessary condition wealth (Table 3).

OPEN IN VIEWER

Table 3.

Pathways to Systemic Underinstitutionalization (PSI).

Solutions	Raw Coverage	Unique Coverage	Countries
wealth * pcon * ccum	.67	.33	Est, Lit + Pol, Ser
wealth * pcon * pfund	.5	.17	Est, Lit + Lat
wealth * ccum * pfund	.5	.17	Bul + Est, Lit
Solution consistency		1.0
Solution coverage		1.0

Source: Table compiled on the basis of the results obtained with fsQCA 2.5.

Note: Bul = Bulgaria; CCUM = cleavage cumulation; Est = Estonia; Lat = Latvia; Lit = Lithuania; PCON = party system concentration; PFUND = party funding; Pol = Poland; PSI = party system institutionalization; Ser = Serbia; WEALTH = economic development.

Similarly to the solution term for the outcome, the relatively low unique coverage values of all three paths signal that they empirically overlap. Thus, both the Estonian and Lithuanian party systems display weak levels of systemic institutionalization for different reasons. Although there are four uniquely covered cases (i.e., Poland, Serbia, Latvia, and Bulgaria), both the consistency and coverage of the overall solution term is 1.

Leaving aside the fact that party systems in rich countries will never suffer from weak levels of institutionalization, it seems clear that the combination of two of any of the other three conditions (i.e., pcon, ccum and pfund) will be enough to hinder the process of systemic institutionalization.

All in all, it seems clear that money (WEALTH * PFUND) is the most important “causal model” explaining the process of PSI in postcommunist Europe. This is not to say that PSI cannot take place in poor (wealth * pfund) countries (e.g., Ukraine), provided they have an adequate socio-institutional configuration (PCON * CCUM). However, if the latter is not true—that is, only one of these two conditions is present—then their fate is totally sealed.

Conclusions

Since Mainwaring and Scully (1995) trumpeted the important consequences PSI may have for the consolidation of democracy in post-transitional countries, much has been written about the level of institutionalization in new party systems. Yet the question of the causes of systemic institutionalization has remained, to say the least, controversial. In order to begin to solve this question, and making use of four different methodological techniques (MSDO/MDSO, congruence, PT, and csQCA), this article has tried to answer the following three questions: What factors, why and how some postcommunist party systems have managed to institutionalize while others have not?

From a methodological point of view, this article not only confirms the general benefits of using MMR, but more specifically, and clearly attuned with other articles of this special issue (Beach and Rohlfing in press; Schneider and Rohlfing 2013; Beach in press) the complementarity of both “configurational” and case-study techniques. Indeed, and adequately combined, condition-/mechanism-centered MMR designs can provide scholars with more (even if different) information than the most sophisticated quantitative analysis. Thus, while the MSDO/MDSO procedure has reduced complexity by reducing the number of possible explanatory conditions from a total of 17 to just 4, the combination of both congruence and PT has allowed me to understand the specific causal mechanisms linking each of the conditions to the outcome both at specific moments and over time, respectively. Once it was clear—thanks to the use of PT—that all causal conditions were neither always present (or absent) nor directly linked, a fully confirmatory csQCA, using Schneider and Rohlfing’s (in press) terminology, enabled me to know the manner in which they combined for specific postcommunist countries.

From a substantive perspective, the main conclusion is the following: At least until 2010, party systems in economically developed nations institutionalized to a higher degree than in economically backward countries. This is not to say, however, that poor countries could not institutionalize, as the examples of both Romania and Ukraine clearly show. However, this certainly required further efforts, that is, (1) a low number of legislative parties and (2) a system of public funding or a cumulative cleavage structure.

Moreover, and perhaps more importantly, this article seems to imply, similarly to what can be found in the democratic consolidation literature (Przeworski and Limongi 1997), the existence of a certain threshold of wealth, suggesting that during the first two decades the institutionalization in postcommunist party systems was, if not only, mostly a question of money.

Although it is perhaps too early to form any definitive conclusions, as the most recent electoral results in Hungary, Slovenia, or the Czech Republic show, pointing as they do to a certain process of deinstitutionalization (Deegan-Krause and Haughton 2015; Haughton and Kraovec 2013; Stegmaier and Linek 2014), what my findings do definitively show is the necessity to build a bridge between those scholars who exclusively emphasize either sociological or institutional dependence. Indeed, and in a similar vein to what all the contributions in this special issue suggest in terms of methodology, complementarity of different explanatory approaches constitutes the only way forward for any revision of the judgments made here.