European Journal of Political Research 47: 537-555, 2008 doi: 10.HH/j.1475-6765.2008.00782.x
537
On the linkage between electoral volatility and party system instability in Central and Eastern Europe
MARGIT TAVITS
Department of Political Science, Washington University in St Louis, MO, USA
Abstract. Electoral volatility is assumed to be a precursor to, or even an indicator of, party system instability. Such an assumption has strong implications for the underlying elite-mass electoral linkage and for the prospects of party system stabilisation in young democracies. This article demonstrates that electoral volatility follows from, rather than leads to, changes in the supply of parties. Thus, the choices of elites may be more responsible for instability in the early stages of party system development than the erratic behaviour of voters.
Students of democracy argue that few institutional developments are more critical for democratic stabilisation than the evolution of stable party systems (Bielasiak 1997,2002; Diamond & Linz 1989; Elster et al. 1998; Kitschelt et al. 1999; Kreuzer & Pettai 2004; Pridham 1990; Toka 1995). Existing studies of democratic consolidation frequently have pointed out the alarming rates of electoral volatility and party system turnover in nascent democracies (Birch 2003; Bielasiak 1997, 2002; Mozaffar & Scarritt 2005; Olson 1998; Sikk 2005; Toka 1998). New parties continue to emerge and old ones continue to die. Voters seem to change their loyalties from election to election. The volatile voter is often accused of impeding the development of stable party systems and, hence, the consolidation of the regime.
In a democracy, political parties are prime mechanisms for integrating diverse interests and social forces in the governing institutions, as well as for regulating social conflict, formulating policy alternatives and holding officials accountable to the public. If voters continue to vote erratically, the argument goes, parties cannot build stable support bases and strong organisations in order to fulfill the above tasks successfully. This failure will make democracies vulnerable to instability and takeover (Innes 2002). The stakes of party system instability are high. However, is it voters switching their vote choice from election to election that makes elites change the supply of parties or do voters merely react to the unstable supply of parties? The purpose of this study is to answer these questions and untangle the relationship between electoral volatility and party system instability.
® 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
Published by Blackwell Publishing Ltd., 9600 Garsington Road, Oxford, OX4 2DQ, UK and 350 Main Street, Maiden, MA 02148, USA
538
MARGIT TAVITS
Uncovering this relationship is crucial in the context of democratic consolidation, and also has broader implications. Whether it is parties or voters who adapt is a debate that reaches beyond specific studies about volatility and party system stabilisation. This debate is in the centre of the literature on voting behaviour and is echoed also in the disagreement between the proximity and directional theories of spatial voting (Hinich & Munger 1997; Lewis & King 2000). As such, this article contributes to the debate about the competing views of mass-elite linkage in all democracies, old or new. The major finding is that in Central and Eastern Europe electoral volatility responds to, rather than triggers, party schisms and elite level manipulations of supply. This finding highlights the importance of elites - actors that are often ignored in favour of voters or institutional features of a system - in the stabilisation process. The results of this study clearly imply that, at least in the initial phase of party system evolution, instability is more likely to result from erratic elites than from inexperienced voters.
Electoral volatility and party system instability: Conceptualisation and causality
'Electoral volatility' is defined as the minimum shift in the vote based on aggregate election results (see also Pedersen 1979), while the concept of 'party system instability' includes frequent changes in the supply of parties (Mainwaring 1999; Mair 1997; O'Dwyer 2004; Toole 2000). A considerable number of studies have assumed that electoral volatility leads to party system instability, and many even use it as an indicator of such instability. Are these processes really the same? Although some authors have criticised this assumption and called for conceptual as well as empirical separation between these processes (Mair 1997), there is still a lack of evidence about exactly how electoral behaviour relates to changes in the party system.
Problems with conceptualisation: Separating demand from supply
Existing studies of party system change pay surprisingly little attention to these questions and, consequently, do not offer adequate answers. Furthermore, most of the existing studies do not differentiate between the roles of the masses and of the elites in the stabilisation of party systems. Indeed, a considerable number of studies on Western European or OECD countries have used electoral volatility as an indicator of party system instability (Crewe 1985; Dalton et al. 1984; Evans 2002; Maguire 1983; Pennings 1998; Pedersen 1979; Rose & Urwin 1970; Shamir 1984; Taagepera & Grofman 2003).1 A similar
S> 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
PARTY SYSTEM INSTABILITY IN CENTRAL AND EASTERN EU 539
approach is used by students of new democracies in Africa, Latin America and Eastern Europe (Bielasiak 1997, 2002; Coppedge 1998; Kuenzi & Mabright 2001; Mainwaring 1998a, 1998b; Mozaffar & Scarritt 2005; Roberts & Wibbels 1999; see also Birch 2003). To provide a few specific examples, Pedersen (1979: 1) treats volatility as a 'developmental aspect of party systems'; other authors refer to the instability, decay or upheaval of party systems when actually explaining electoral volatility (Roberts & Wibbels 1999; Tavits 2005). Similarly, Taagepera & Grofman (2003: 660) state: '[Vjolatility is a key element in understanding party system stability.' Przeworski (1975) uses the term 'party system institutionalization' while actually referring to electoral volatility. These examples illustrate how the concept of 'party system instability' is often reduced to the electoral level. This focus on the demand side of the political market ignores the choices of elites and, thus, implicitly assumes that a volatile electorate is responsible for system instability. Underlying this widespread approach seems to be the assumption derived from the literature on Western democracies that parties are automatic by-products of pre-existing societal cleavages (see Kreuzer & Pettai 2004: 621-623).
Several authors, who are critical of treating electoral shifts and party system shifts as equivalent, have argued for separating the role of elites and masses in the process of party system stabilisation (Crewe 1985; Mair 1997; Toole 2000; see also Kreuzer & Pettai 2004; Laver & Benoit 2003). Mair (1997) explicitly calls for abandoning the electoral focus in defining and studying party system change. He argues that party system change can be conceptualised as an entirely elite level phenomenon - as a change in the relationship between parties. The stabilisation of party systems would thus occur when the patterns of interaction among political parties become predictable, rather than when the patterns of representation - voter and party alignments - become predictable. In contrast to the conceptualisation that centres on the demand-side of party system change, this approach moves to the other extreme: it defines party systems mostly in terms of the supply side of the electoral market. The interesting puzzle, however, is whether and to what extent supply influences demand and vice versa. An exclusive focus on either side does not permit exploring this relationship. However, studying the association between these concepts helps uncover the origins of party system stabilisation.
Interaction between supply and demand
Several studies have recognised that party system stabilisation may be a product of both voter demand and elite supply (see Kitschelt et al. 1999; Rose & Munro 2003). This recognition is further echoed in studies of strategic coordination in the electoral arena (Cox 1997) and the emergence and success
S> 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
540
MARGIT TAVITS
of new parties (see especially Hug 2001). Unlike Crewe (1985), Mair (1997) and Toole (2000), these studies focus specifically on the level of electoral politics: voters try to coordinate on the choice of a few viable parties in order to avoid wasting their votes; and elites, when deciding on entry or withdrawal from the electoral arena, try to anticipate which parties will be viable in the eyes of the voters (Cox 1997:157-178).
Recognising the distinct roles of elites and masses in producing stability is, however, only the first step in understanding the relationship between these concepts. Equally crucial is to uncover the direction of causality or at least the temporal order between the supply and demand. Does the volatile electoral behaviour of voters cause the change in the supply of parties? Or does the inconsistent behaviour of party elites (in terms of withdrawing old parties and launching new ones) cause electoral volatility? It is intuitive to argue that each side reacts to the other to a certain extent, but it is important to tease out empirically whose behaviour works as the cause.
The existing literature provides arguments to support either causal direction. First, some studies argue that electoral volatility causes changes in the supply of parties because elites try to anticipate voter behaviour and act accordingly. Specifically, several authors state that electoral volatility increases the supply of parties (Birch 2003; Maor 1997; Mair 1997), or that in volatile systems, new parties are more likely to emerge and old ones to disappear (Mainwaring 1998a; Toka 1998). Furthermore, the organisational weakness of parties in new democracies is often blamed on voters - due to low party identification or a weak civil society (Elster et al. 1998). Because a high level of electoral volatility leads to election outcomes that are less predictable (Mainwaring 1998a), it affects the extent to which elites have predictable bargaining power in government formation. This, in turn, influences the durability of cabinets and the stability of inter-party relations (Mainwaring 1998c; Powell 1982). All these arguments echo the view that elites react to the behaviour of the electorate.
A high level of electoral volatility, the argument goes, indicates that large parts of the electorate have not developed loyalty towards any of the existing parties. This, in turn, signals to the elites of potential new parties that the probability of getting electoral support is relatively high - or at least that their potential electoral viability is not worse than that of existing parties. Given such an expectation, they are more likely to launch a new party (Bielasiak 1997; Cox 1997; Mair 1997; Reich 2004). For existing parties, on the other hand, the uncertainty and the high level of availability of the electorate poses a threat and a potential cost rather than an opportunity. Therefore, if electoral volatility is high, parties that have already suffered some electoral defeat may prefer to withdraw, rather than invest in a costly campaign. It follows that
S> 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
PARTY SYSTEM INSTABILITY IN CENTRAL AND EASTERN EU 541
electoral instability leads strategic elites to increased entries and increased withdrawals.
The causality, however, may also run in the opposite direction. In young democracies, elites may not necessarily be so strategic and prudent. Rather, they may be inconsistent, impatient and in search of instant gratification. Elites may not bother to build strong party organisations and develop grassroots connections. They also may not present clear choices or stand for identifiable values, and thus frequently merge, split, dissolve and create parties that lead to continuing party system instability. Voters simply are not given a chance to vote consistently because the choices with which they are presented differ significantly from election to election. It has, indeed, been argued that rather than being a cause, electoral volatility is an effect of party system change (Pedersen 1979).
A similar conclusion is reached by qualitative studies of some specific countries or elections, such as Moldova (King 2000), Ukraine (Birch 2000), Russia 1999 (Rose et al. 2001), Hungary, Poland and the Czech Republic (Toole 2000). Tworzecki (2003) suggests that in the new democracies of Eastern Europe, it is the parties, not the voters, that are unstable. Similarly, Zielinski (2002) argues that elites, not voters, have to solve the coordination problem of supplying parties in order for stability to occur. The counter argument thus posits that elites trigger voter response, not vice versa. Elite level instability - the change in the supply of electoral choices, the withdrawal of old parties and the creation of new ones - makes it impossible for voters to develop loyalties towards parties. Furthermore, one would also expect there to be a certain mechanical effect: if a new party is launched and it attracts some votes, this process should be reflected in electoral volatility. The same is true of the withdrawal of an old party that used to attract at least some share of votes (Pedersen 1979). The existing literature therefore poses two opposing causal chains for the relationship between the supply of parties and electoral volatility. These are summarised in Figure 1. The next section will try to tease out the causal order between these variables empirically.
Analysis
A total of 13 new democracies of Central and Eastern Europe were selected as cases with which to test the propositions. Such case selection is justified because party system stabilisation and high electoral volatility are the specific concerns of new democracies. It is here that electoral and partisan instability might inhibit the consolidation of democracy or even threaten the existence of the regime. The debate about the source of instability is most apparent in the
S> 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
542 MARGIT TAVITS
Electoral volatility as the cause
Low voter loyalty		High level of electoral volatility		Unpredictable election results	->	(1) the emergence of new parties (2) the withdrawal of risk averse parties
		Electoral volatility as the effect				
Impatient elites		(1) the emergence of new parties (2) the withdrawal of risk averse parties	->	Low voter loyalty	->	High level of electoral volatility
Figure 1. The opposing causal chains for the relationship between the supply of parties and electoral volatility.
context of these new democracies. The countries and elections included are listed in Appendix A. All of the countries included are in essence parliamentary systems, although some have directly elected and strong presidents. The actual number of cases differs across analyses depending on the availability of information on specific variables and the number of lags included.
The purpose of the analysis is to establish the nature of the link between electoral volatility and the supply of parties. To do this, I will employ instrumental variable regression. This technique is appropriate for determining the causal order between variables. Based on the theoretical discussion above (and the empirical analyses below), one would predict a high level of mutual dependence between volatility and the supply of parties. Given such endo-geneity, ordinary least squares regression remains inadequate for determining the effects of each variable on the other. Similarly, simply observing, case by case, the trends in volatility and the supply of parties is likely to uncover only correlation, but not establish causality or the actual size of the effects. I will explain the technique of instrumental variable regression in more detail below.
Electoral volatility is measured by the Pedersen (1979) index:
V = 'L\cu_l-cu\/2,
where V is volatility, cI>( is the vote share for a party / at a given election (t) and d,t-i is the vote share of the same party / at the previous election (t— 1). The change in the Supply of parties is measured by an index that combines two variables: the count of parties in a given election that field candidates for the first time, and the count of parties that fielded candidates in the previous election, but not in the given one.2,3 For the purposes of the analysis, both of these count variables are standardised - that is, centred around zero with a standard deviation of one, and averaged into an index. Such a transformation
S> 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
PARTY SYSTEM INSTABILITY IN CENTRAL AND EASTERN EU
543
gets rid of the count data and allows using the appropriate estimation techniques: ordinary least squares regression with panel corrected standard errors (PCSE) and instrumental variable regression.4
The estimations also include a set of control variables used in studies of party system change. First, poor economic performance may trigger overall vote shifts and encourage change in the supply of parties (Hug 2001). This variable is measured by GDP growth rate.5 Cleavage structure is another commonly used control variable in studies of party system instability (Roberts & Wibbels 1999). To measure this, I use Vanhanen's (1999) index of Ethnic heterogeneity accounting for the number and size of different ethnic, racial and religious groups. Electoral systems can also constrain aggregate vote shifts and new party entries (Hug 2001; Roberts & Wibbels 1999). The mean District magnitude captures the disproportionality of the system and the cross-national variance in the level of institutional permissiveness towards new contestants in the electoral arena (Taagepera & Shugart 1989).6 Furthermore, party support may also decrease simply because voters withdraw. Turnout change is measured by the difference in voter turnout in the previous and current election. The value of the variable is 0 when turnout between the two elections is exactly the same; it is negative when turnout has decreased and positive when it has increased.7 Additionally, time itself, or rather democratic development and the crystallisation of party identification among voters, may be an important factor influencing the extent of instability in the electoral arena (Tavits 2005). The Age of democracy is measured by years since the first democratic election.
Many of the countries included in the analysis have directly elected and relatively strong presidents, including Lithuania, Moldova (until 2000), Poland, Romania, Russia and Slovakia (since 1999). This has been argued to increase the instability of party systems by encouraging new party entries (Shugart & Carey 1992). Siaroff (2003) has generated an index of presidential power by counting how many of the following characteristics are present: popular election of president; concurrent election of president and legislature; discretionary appointment powers; chairing of cabinet meetings; right to veto; long-term emergency and decree powers; central role in foreign policy; central role in government formation; and ability to dissolve the legislature. This index, however, is highly correlated with the variable measuring the supply of parties (r = 0.60). When both of these variables are entered in the same model, their independent effects are impossible to determine; indeed, both of them appear insignificant. Since knowing the effect of the supply of parties is crucial for testing the argument, I decided not to include the presidential power index in the analyses. Descriptive statistics for all variables included in the analyses below are provided in Appendix B.
S> 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
544
MARGIT TAVITS
A test of causality
Before tackling the issue of causality, Table 1 presents the results of the ordinary least squares regressions of the level of electoral volatility and the change in party supply. Both Models 1 and 2 include the same control variables and use PCSE. The latter take care of panel-heteroskedasticity (Beck & Katz 1995). Lagrange multiplier tests showed no serial correlation of errors in either model, which is why no time-series techniques are employed.
Model 1 estimates the effect of change in the supply of parties (from election t-1 to t) on the level of electoral volatility (from election t-1 to t). The results show that change in the supply of parties significantly influences the level of electoral volatility: the greater the number of entries and exits, the higher the electoral volatility.8 Model 2, in turn, estimates the effect of electoral volatility on change in the supply of parties. The underlying logic here is that prior high levels of electoral volatility may encourage the entrance of new contestants in the electoral arena or discourage some existing ones from entering. Given that candidates at election t are fielded before the volatility scores from election t-1 to t can be calculated, the only information that elites can use is the level of volatility prior to election t (i.e., from election t-2 to t-1).
Table 1. Ordinary least squares estimations of the mutual effects of electoral volatility and change in the supply of parties
OLS with PCSE
Model 1 Model 2
(DV = Electoral volatility)       (DV = Supply of parties)
Variables	b(PCSE)	b (PCSE)
Electoral volatility		2.164** (0.761)
Supply of parties	0.077** (0.0170)	
District magnitude (log)	-0.012 (0.0190)	0.006 (0.067)
GDP growth	0.001 (0.0010)	-0.007 (0.007)
Ethnic heterogeneity	0.001** (0.0002)	0.002 (0.003)
Turnout change	0.006** (0.0020)	-0.016 (0.024)
Age of democracy	0.005* (0.0030)	0.002 (0.023)
Constant	0.240** (0.0620)	-0.925** (0.229)
R2	0.35	0.27
N	39	27
Note: Table entries are unstandardised regression coefficients with panel corrected standard errors in parentheses. ** p < 0.01; * p < 0.1, one-tailed test.
O 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
PARTY SYSTEM INSTABILITY IN CENTRAL AND EASTERN EU 545
As the results in Model 2 show, the level of electoral volatility from prior elections is, indeed, associated with increases in change in the supply of parties.
Based on these results, one would conclude that electoral volatility and change in the supply of parties have reciprocal effects on each other. However, as argued above, neither electoral volatility nor the supply of parties is a strictly independent variable, although both are used as such in the above models. Given such endogeneity, conclusions about their effects based on the OLS models are likely to be flawed. A stronger test of the causal order between these variables can be performed by using instrumental variable regression. The start-up assumption of this estimation is that both variables, volatility and the supply of parties, are endogenous. In order to determine whether each variable is exogenous - that is, whether it is causally prior to the other variable - we need to find plausible instruments for both. These instruments should be correlated with the endogenous regres-sors, orthogonal to any other omitted characteristics and not correlated with the outcomes of interest through any channel other than their effect via the endogenous regressors (Acemoglu et al. 2003). Lagged values of the variables in the model provide natural candidates for such instruments (Greene 2003: 79-80).
More specifically, consider Model 1. Here I find that the change in the supply of parties has a significant effect on electoral volatility. However, both of these variables are likely to be influenced by past electoral volatility. This causes a correlation between the Supply of parties and the error term in the model, which in turn leads to biased (inflated) estimates. Instruments are used to isolate the information in this variable that is not correlated with the error term. Specifically, instrumental variable regression is estimated in two stages. The first stage creates a new variable that replaces the problematic causal variable - the Supply of parties. In order to do so, the Supply of parties is regressed on its instrument (lagged values of this variable) and other independent variables in the main regression. The predicted values of the Supply of parties from this first stage regression are then used in the second stage model as the instrumental variable to get the exogenous effect of the change in the supply of parties on electoral volatility. In essence, the instrumental variable is the part of the Supply of parties that is purged of any endogenous effects. If this instrumental variable is still significantly related to electoral volatility, then we can conclude that it is exogenous or causally prior to volatility.
The results of the instrumental variable regression are presented in Table 2. Since panel corrected standard errors are not available for this estimation technique, I have used Huber-White-Sandwich robust standard errors within country clusters to deal with the country-level heteroskedasticity. Models 3
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
Table 2. Instrumental variable regression estimations of the mutual effects between electoral volatility and change in the supply of parties
Instrumental variable regression: First stage Instrumental variable regression: Second stage
Model 5 Model 6
Model 3                             Model 4                       (DV = Electoral (DV = Supply
(DV = Electoral volatility)       (DV = Supply of parties)               volatility) of parties)
Variables                                  b (robust SE)                      b (robust SE)                    b (robust SE) b (robust SE)
Electoral volatility	0.383** (0.162)			2.965 (6.904)
Supply of parties		0.288* (0.1970)	0.1140** (0.0580)	
District magnitude (log)	-0.099 (0.104)	-0.034*** (0.0140)	-0.0220 (0.0160)	0.077 (0.208)
GDP growth	-0.018 (0.026)	-0.001 (0.0020)	0.0060** (0.0030)	-0.006 (0.017)
Ethnic heterogeneity	0.003 (0.005)	0.001** (0.0007)	0.0010* (0.0007)	0.002 (0.011)
Turnout change	-0.047** (0.025)	0.001 (0.0020)	0.0090* (0.0070)**	-0.017 (0.021)
Age of democracy	-0.001 (0.049)	0.014* (0.0110)	0.0006 (0.0060)	0.013 (0.156)
Constant	0.395 (0.598)	0.153* (0.1020)	0.2870*** (0.0990)	-1.389 (1.420)
R2	0.41	0.49	0.29	0.40
N	29	19	29	19
Note: Table entries are unstandardised regression coefficients with robust standard errors in parentheses. The instruments for both variables are the lagged values of these variables. *** p < 0.01; ** p < 0.05; * p < 0.1, one-tailed test.
PARTY SYSTEM INSTABILITY IN CENTRAL AND EASTERN EU
547
and 4 refer to the first stage regressions used for generating the instrumental variables for electoral volatility and the supply of parties respectively. The coefficients of lagged values of both variables are statistically significant, indicating that these variables serve as strong instruments. Given that lagged values of both variables are used, the number of observations is reduced compared to the analyses in Table 1.
Model 5 and 6 contain the second stage estimation. Model 5 presents the effect of change in the supply of parties on electoral volatility. In essence, this is a re-estimation of Model 1, except that the Supply of parties is replaced by the predicted values of the first stage regression. The coefficient of the instrumental variable represents an unbiased exogenous estimate. Its high level of statistical significance supports the conclusion that changes in the supply of parties are temporally and causally prior to electoral volatility.
Model 6 estimates whether the opposite is also true: perhaps there is a reciprocal relationship between these variables? Again, Model 6 re-estimates Model 2 using the predicted values of electoral volatility from the first stage regression. The results of this analysis support the null hypothesis: correcting for endogeneity, electoral volatility has no significant effect on the supply of parties. Indeed, the effect of electoral volatility remains indistinguishable from zero even when all control variables are removed from the model. This indicates that the causality between the change in the supply of parties and electoral volatility runs from the former to the latter, but not vice versa.
As argued above, the explanation for such a finding is intuitive. Frequent changes to the number and types of parties do not allow voters to form loyalties to any electoral contestant. Their party of choice may not even contest in the next election and they will be forced to switch their vote or abstain. Similarly, new parties emerging may be ideologically closer to some voters and offer policies more to their liking than their previous vote choice. This makes it rational to switch one's vote. Keeping the supply of parties constant reduces the need for voters to switch their allegiance. Thus, consistency in elite behaviour becomes crucial for determining the size of the party system as well as the stability of voter choices.
These inferences about the direction of causality follow directly from the statistical technique employed. Unfortunately, the evidence supporting this conclusion cannot be presented in a more intuitive manner. When describing specific cases, it is impossible to sort out the independent exogenous effect of either variable on the other and we would only be left with the conclusion that they are correlated. In a simplified manner, the logic of the argument can be illustrated by the following specific examples. These, however, should not be taken as explanations of the intuition behind the results obtained by the instrumental variable regression or evidence of the direction of causality.
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
548
MARGIT TAVITS
Consider, for example, Latvia, where electoral volatility dropped by 4 per cent from 1995 to 1998. If trends in volatility predicted elite level behaviour, this drop should have encouraged stabilisation in the supply of parties for the following election. To the contrary, however, the number of entries and exits increased to ten compared to eight in 1998 and four in 1995. Similarly, in Slovakia, electoral volatility decreased by 3 per cent from 1994 to 1998. This, however, did not lead to increased stability in the supply of parties, but rather increased the entry and exit of parties from four in 1998 to ten in 2002. At the same time, in both countries, increased instability in the supply of parties was always followed by increased electoral volatility.
The results of the current study echo the findings of some previous qualitative case studies that have also suggested that party systems can change or remain stable independently of electoral change. Korasteleva (2000) argues that in Belarus there is extensive vote switching, but it does not have a significant effect on the structure of competition between parties. Toole (2000) reaches a similar conclusion in his analysis of Poland, Hungary and the Czech Republic: using different measures, the study finds that party system stabilisation has occurred in these countries despite continuing high level of electoral volatility. The analyses in this article have provided more systematic evidence to support this claim. Indeed, the results consistently suggest that elite level choices and behaviour may be more consequential for the stabilisation of the regime than voter-level behaviour. Focusing on elites may provide original and invaluable insights for better understanding the process of stabilisation. This does not rule out the possibility that elites take into consideration anticipated voter reaction when launching a new party or withdrawing an existing one. The empirical regularities established simply state that past voter behaviour is not significantly influencing such decisions. The results also leave open the possibility that past and/or anticipated voter behaviour influences other types of elite-level decisions about their electoral strategies, such as party ideological placement or adaptation (Adams et al. 2004).
Conclusions
This article was motivated by the puzzle of how new party systems emerge and stabilise, and what the respective roles of elites and voters are in this process. The common assumption that electoral volatility equals party system instability does not get us very far in answering these questions. Thus, in this study I have tried to separate these concepts and consider whether and how electoral choices relate to different aspects of party system development. Causality tests
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
PARTY SYSTEM INSTABILITY IN CENTRAL AND EASTERN EU
549
suggested that rather than triggering change in the supply of parties and hence party system instability, electoral volatility merely reacts to it. The results provide consistent support for the important role of elites in the emergence and stabilisation of party systems in young democracies.
These findings provide novel insights and implications for the process of stabilisation. The role of elites has been often ignored in previous literature, probably because of the explicit electoral bias in describing and analysing party system change and stability in advanced democracies. The idea of an elite-driven party system development has occasionally been discussed in the context of new party systems. This further implies that the logic of party system stabilisation in young democracies may be different from the one applicable to advanced democracies. The findings in this article reinforce and provide important empirical evidence for the suggestions that party politics in new democracies is very much elite-dominated (Van Biezen 2003; Mair 1997) and that elite strategising may significantly condition the effects of other explanatory variables on the emerging party systems (Gunther 1989).
Following the developments in the young democracies in Central and Eastern Europe over time has helped to put forward a novel logic of the initial stabilisation of a new party system. Party systems in young democracies do not necessarily 'freeze' as a result of stable patterns of party support among the electorate. Rather, elite choices in designing the party system may help stabilise electoral alignments. If party elites can make the available choices more stable and consistent, they may be able to bring about more coherent voter preferences and more predictable electoral alignments. The novelty of this implication lies in moving away from the 'inevitability' and the notion of 'lengthy stabilisation process' created by the argument that electoral stabilisation forms the basis for party system and government stabilisation in new democracies. The findings here suggest that much more emphasis should be put on understanding the incentive structures of elites that encourage or discourage stability on their part.
The relative freshness of the implications of these findings undoubtedly points to the need for further research. We still need to understand better the aspects of elite behaviour beyond the supply of parties that potentially trigger mass level response. For example, one could examine the effects of party membership volatility, candidate switching (see Kreuzer & Pettai 2003; Laver & Benoit 2003; Shabad & Slomczynski 2004), or party ideological coherence on stabilisation. Furthermore, as already mentioned, if elites dominate party system stabilisation, we need to understand their incentive structures. Following the development of young democracies remains an invaluable source for such theory development and revision.
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
550
MARGIT TAVITS
Acknowledgements
I would like to thank Taavi Annus, Kenneth Janda, Lisa Pohlman, Richard Rose and Daniel Joseph Young for reading and commenting on different versions of this article. An earlier version was presented at the Annual Meeting of the Midwest Political Science Association, Chicago, IL, 7-10 April 2005.
Appendix A. The number of entries and exits
		Number	Number
Country	Election	of entries	of exits
Bulgaria	1991	9	0
Bulgaria	1994	4	2
Bulgaria	1997	3	0
Bulgaria	2001	6	1
Czech Republic	1992	9	3
Czech Republic	1996	1	1
Czech Republic	1998	1	2
Czech Republic	2002	4	0
Estonia	1995	9	3
Estonia	1999	2	4
Estonia	2003	1	3
Hungary	1994	3	1
Hungary	1998	2	2
Hungary	2002	1	2
Latvia	1995	3	1
Latvia	1998	4	4
Latvia	2002	6	4
Lithuania	1996	6	1
Lithuania	2000	5	2
Lithuania	2004	4	2
Moldova	1998	8	5
Moldova	2001	7	2
Poland	1993	4	4
Poland	1997	4	7
Poland	2001	4	4
Romania	1992	8	1
Romania	1996	8	0
Romania	2000	4	3
©2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
PARTY SYSTEM INSTABILITY IN CENTRAL AND EASTERN EU
Appendix A. Continued.
551
Number Number
Country	Election	of entries	of
Romania	2004	7	3
Russia	1995	15	2
Russia	1999	7	4
Slovakia	1992	7	2
Slovakia	1994	6	2
Slovakia	1998	2	2
Slovakia	2002	9	1
Slovenia	1992	7	0
Slovenia	1996	2	1
Slovenia	2000	3	2
Slovenia	2004	4	0
Appendix B. Descriptive statistics
Variable	N	Mean	Standard deviation	Minimum	Maximum
Electoral volatility	39	0.323	0.117	0.141	0.663
Supply of parties	39	-0.081	0.767	-1.050	2.830
District magnitude (log)	39	2.384	1.206	0.672	5.011
GDP growth	39	-1.539	8.702	-34.090	9.000
Ethnic heterogeneity	39	28.666	26.073	4.000	90.000
Turnout change	39	8.079	5.337	0.280	22.370
Age of democracy	39	7.641	3.876	1.000	14.000
Notes
1. See also Mair (1997: 214-215) for a similar argument.
2. Both of these measures are based on the sources used to determine the new parties in Tavits (2008). The same sources were used to determine the parties that withdrew. Because the entry and exit of very small parties is unlikely to have destabilising consequences for the party system or democratic development, and thus are of little interest, only those parties that received at least 0.3 per cent of the vote were counted. Rose (1996), Olson (1998) and Birnir (2005) use even higher thresholds: 1 per cent of the vote or parliamentary representation. Admittedly, there is no 'right' threshold for including parties (see Tavits (2008) for an extensive discussion of this issue). However, even a party that in the end fails to get much support in the electoral arena has by its presence changed the nature of the electoral game.
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
552
MARGIT TAVITS
3. This remains a rather crude measure of the change in the supply because it does not account for changes within parties that may also disorient voters. Underestimating the extent of change is likely to weaken the statistical relationship between the supply and electoral volatility and pose an uphill battle for detecting significant relationships. The actual relationships are therefore likely to be even stronger than suggested by the findings here.
4. The raw numbers of entries and exits are presented in Appendix A.
5. The measures are coded from Easterly & Sewadeh (2002), and for information beyond 2001 from Eurostat and the CIA World Factbook 2002, 2003, 2004. Replacing the GDP growth rate with other measures of economic performance such as unemployment rate and inflation rate did not alter the substantive results.
6. The variable is coded mostly from Beck et al. (2001). It is common in the literature to code countries with mixed electoral systems (such as Hungary, Lithuania and Russia) as having district magnitude of 1. This is true only for the single member district part of the election, but disregards the proportional tier. I have used a different method by taking the total number of candidates (elected from both tiers) and dividing it by the total number of districts in both tiers. For example, if there are 300 candidates elected, 150 of them from single member districts and 150 from one nationwide district, 300 is divided by 151 (the total number of districts). Using the conventional measurement does not alter the results.
7. Source: International Institute for Democracy and Electoral Assistance (http:// www.idea.int).
8. The substantive effects in these analyses are difficult to interpret given that the supply of parties is measured by a standardised index (i.e., the specific values of this variable are un-interpretable).
References
Acemoglu, D. et al. (2003). Institutional causes, macroeconomic symptoms: Volatility, crises
and growth. Journal of Monetary Economics 50(1): 49-123. Adams, J. et al. (2004). Understanding change and stability in party ideologies: Do parties
respond to public opinion or to past election results? British Journal of Political Science
34: 589-610.
Beck, N. & Katz, J.N. (1995). What to do (and not to do) with time-series cross-section data
in comparative politics. American Political Science Review 89(3): 634-647. Beck, T. et al. (2001). New tools in comparative political economy: The database of political
institutions. World Bank Economic Review 15(1): 165-176. Bielasiak, J. (1997). Substance and process in the development of party systems in East
Central Europe. Communist and Post-Communist Studies 30(1): 23^-4. Bielasiak, J. (2002). The institutionalization of electoral and party systems in postcommunist
states. Comparative Politics 34(2): 189-210. Birch, S. (2000). Elections and democratization in Ukraine. Basingstoke: Macmillan. Birch, S. (2003). Electoral systems and political transformation in post-communist Europe.
New York: Palgrave Macmillan. Birnir, J.K. (2005). Public venture capital and party institutionalization. European Journal of
Political Research 38(8): 915-938. Coppedge, M. (1998). The dynamic diversity of Latin American party systems. Party Politics
4(4): 547-568.
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
PARTY SYSTEM INSTABILITY IN CENTRAL AND EASTERN EU
553
Cox, G. (1997). Making votes count: Strategic coordination in the world's electoral systems. Cambridge: Cambridge University Press.
Crewe, I. (1985). Introduction: Electoral change in Western democracies: A framework for analysis. In I. Crewe & D. Denver (eds), Electoral change in Western democracies. London/Sydney: Croom Helm.
Dalton, R.J., Flanagan, S.C. & Beck, P.A. (eds). (1984). Electoral change in advanced industrial democracies: Realignment or dealignment? Princeton, NJ: Princeton University Press.
Diamond, L. & Linz, J.J. (1989). Introduction: Politics, society and democracy in Latin America. In L. Diamond, J.J. Linz & S.M. Lipset (eds), Democracy in developing countries, Vol. TV: Latin America. Boulder, CO/London: Lynne Rienner/Adamanture Press.
Easterly, W. & Sewadeh, M. (2002). Global development network growth database. Washington, DC: World Bank.
Elster, J, Offe, C. & Preuss, U.K. (1998). Institutional design in post-communist societies. Cambridge: Cambridge University Press.
Evans, J.A.J. (2002). In defence of Sartori: Party system change, voter preference distribution and other competitive incentives. Party Politics 8(2): 155-174.
Greene, WH. (2003). Econometric analysis. Upper Saddle River, NJ: Prentice Hall.
Gunther, R. (1989). Electoral laws, party systems and elites: JTie case of Spain. American Political Science Review 83(3): 835-858.
Hinich, M.J. & Munger, M.C (1997). Analytical politics. Cambridge: Cambridge University Press.
Hug, S. (2001). Altering party systems. Ann Arbor, MI: University of Michigan Press. Innes, A. (2002). Party competition in postcommunist Europe. Comparative Politics 35: 85-104.
King, C. (2000). The Moldovans: Romania, Russia and the politics of culture. Stanford, CA:
Hoover Institution Press. Kitschelt, H. et al. (1999). Post-communist party systems: Competition, representation and
inter-party cooperation. Cambridge: Cambridge University Press. Korasteleva, EA. (2000). Electoral volatility in postcommunist Belarus. Party Politics 6(3):
343-358.
Kreuzer, M. & Pettai, V. (2003). Patterns of political instability: Affiliation patterns of politicians and voters in postcommunist Estonia, Latvia and Lithuania. Studies in Comparative International Development 38(2): 76-98.
Kreuzer, M. & Pettai, V. (2004). Political parties and the study of political development: New insights from the postcommunist democracies. World Politics 56: 608-633.
Kuenzi, M. & Mabright, G. (2001). Party systems institutionalization in 30 African countries. Party Politics 7(4): 437^168.
Laver, M. & Benoit, K. (2003). JTie evolution of party systems between elections. American Journal of Political Science 47(2): 215-233.
Lewis, J.B. & King, G. (2000). No evidence on directional versus proximity voting. Political Analysis 8(1): 21-33.
Maguire, M. (1983). Is there still persistence? Electoral change in Western Europe, 1948-1979. In H. Daalder & P. Mair (eds), Western European party systems: Continuity and change. London: Sage.
Mainwaring, S. (1998a). Rethinking party system theory in the third wave of democratization: The importance of party system institutionalization (Working Paper 260). Notre Dame, IN: Helen Kellogg Institute for International Studies, University of Notre Dame.
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
554
MARGIT TAVITS
Mainwaring, S. (1998b). Electoral volatility in Brazil. Party Politics 4: 523-545.
Mainwaring, S. (1998c). Party systems in the third wave. Journal of Democracy 9: 67-81.
Mainwaring, S. (1999). Rethinking party systems in the third wave of democratization: The case of Brazil. Stanford, CA: Stanford University Press.
Mair, P. (1997). Party system change: Approaches and interpretations. Oxford: Oxford University Press.
Maor, M. (1997). Political parties and party systems: Comparative approaches and the British
experience. London/New York: Routledge. Mozaffar, S. & Scarritt, J.R. (2005). The puzzle of African party systems. Party Politics 11(4):
399-421.
O'Dwyer, C. (2004). Runaway state building: How political parties shape states in postcom-
munist Eastern Europe. World Politics 56: 520-553. Olson, D.M. (1998). Party formation and party system consolidation in the new democracies
of Central Europe. Political Studies 46: 432^-64. Pedersen, M.N. (1979). The dynamics of European party systems: Changing patterns of
electoral volatility. European Journal of Political Research 7(1): 1-26. Pennings, P. (1998). The triad of party system change: Votes, office and policy. In P. Pennings
& J.-E. Lane (eds), Comparing party system change. London: Routledge. Powell, B.G. (1982). Contemporary democracies: Participation, stability and violence.
Cambridge, MA: Harvard University Press. Pridham, G. (1990). Securing democracy: Political parties and democratic consolidation in
Southern Europe. London/New York: Routledge. Przeworski, A. (1975). Institutionalization of voting patterns, or is mobilization the source of
decay? American Political Science Review 69(1): 49-67. Reich, G. (2004). The evolution of new party systems: Are early elections exceptional?
Electoral Studies 23: 235-250. Roberts, K.M. & Wibbels, E. (1999). Party systems and electoral volatility in Latin America:
A test of economic, institutional and structural explanations. American Political Science
Review 93(3): 575-590. Rose, R. (1996). What is Europe? A dynamic perspective. New York: HarperCollins. Rose, R. & Munro, N. (2003). Elections and parties in new European democracies. Washington, DC: Congressional Quarterly Press. Rose, R. & Urwin, D.W. (1970). Persistence and change in Western party systems since 1945.
Political Studies 18(3): 287-319. Rose, R., Munro, N. & White, S. (2001). Voting in a floating party system: The 1999 Duma
election. Europe-Asia Studies 53(3): 419-443. Shabad, G. & Slomczynski, K. (2004). Interparty mobility among political elites in post-communist East Central Europe. Party Politics 10(2): 151-176. Shamir, M. (1984). Are Western European party systems frozen? Comparative Political
Studies 17(1): 35-79.
Shugart, M.S. & Carey, J.M. (1992). Presidents and assemblies: Constitutional design and electoral dynamics. Cambridge: Cambridge University Press.
Siaroff, A. (2003). Comparative presidencies: The inadequacy of the presidential, semi-presidential and parliamentary distinction. European Journal of Political Research 42: 287-312.
Sikk, A. (2005). How unstable? Volatility and the genuinely new parties in Eastern Europe. European Journal of Political Research 44: 391-412.
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)
PARTY SYSTEM INSTABILITY IN CENTRAL AND EASTERN EU
555
Taagepera, R. & Shugart, M.S. (1989). Seats and votes: The effects and determinants of
electoral systems. New Haven, CT: Yale University Press. Taagepera, R. & Grofman, B. (2003). Mapping the indices of seats-votes disproportionality
and inter-election volatility. Party Politics 9(6): 659-677. Tavits, M. (2005). The development of stable party support: Electoral dynamics in post-communist Europe. American Journal of Political Science 49(2): 283-298. Tavits, M. (2008). Party systems in the making: The emergence and success of new parties in
new democracies. British Journal of Political Science 38(1): 113-134. Toka, G. (1995). Parties and electoral choices in East-Central Europe. In G. Pridham & P.G.
Lewis (eds), Stabilising fragile democracies: Comparing new party systems in Southern
and Eastern Europe. London/New York: Routledge. Toka, G. (1998). Party appeals and voter loyalty in new democracies. Political Studies 46:
589-610.
Toole, J. (2000). Government formation and party system stabilization in East Central
Europe. Party Politics 6(4): 441-461. Tworzecki, H. (2003). Learning to choose: Electoral politics in East-Central Europe. Stanford,
CA: Stanford University Press. Van Biezen, I. (2003). Political parties in new democracies: Party organization in Southern
and East-Central Europe. New York: Palgrave. Vanhanen, T. (1999). Domestic ethnic conflict and ethnic nepotism: A comparative analysis.
Journal of Peace Research 36(1): 55-73. Zielinski, J. (2002) Translating social cleavages into party systems: The significance of new
democracies. World Politics 54: 184-211.
Address for correspondence: Margit Tavits, Assistant Professor, Department of Political Science, Washington University in St Louis, Campus Box 1063, One Brookings Drive, St Louis, MO 63130, USA. Tel.: +001 314 935 5810; Fax: +001 314 935 5856; E-mail: tavits® wustl.edu
© 2008 The Author(s)
Journal compilation © 2008 (European Consortium for Political Research)