Abstract
Giovanni Sartori once described African party systems as “formless.” Our contribution challenges this view in an era of resurgent multipartism that swept through the subcontinent in the early 1990s and continues until today. The article brings together contemporary research on African party systems with the wider disciplinary literature on party system institutionalization. Using a data set including all continuous election sequences in Africa from 1950 to 2008, we find that Africa has some of the most volatile party systems ever recorded and yet, tremendous diversity across regimes. We test the relative impact of political institutions, economic performance, the history of party system development, and social cleavage structure on party system institutionalization in Africa. We find that its party systems have been shaped by a set of factors unique to the subcontinent, but some of the general global patterns of party system development hold true in Africa as well.
Over thirty years ago Giovanni Sartori characterized the pervasive weak party systems of Africa as “formless polities”, propounding an exceptionalist view of African party systems that has been echoed in decades of subsequent scholarship (1976, chap. 8). Newly independent countries lacked political institutions that mediated between state and society and most commonly turned into fragile one-party states vulnerable to coups and military rule. Today the African exceptionalism argument is given less credence: “Africa is . . . a place where issues and theories that have been worked on in other areas can now also be studied” (Lindberg, 2009b). In this article, we examine party system institutionalization in Africa with the experience of other newly democratizing regions as a theoretical backdrop.
Party system institutionalization varies tremendously across the subcontinent. While dictatorship became ubiquitous in the period following liberation, some countries experienced intervals of democracy and the dictatorships often did not endure for long. The waves of mass protest and liberalization of dictatorships that rocked the continent in the late 1980s and early 1990s have yielded even more diverse systems of rule (Bratton & van de Walle, 1997). Some democratic experiments have sustained competitive elections while others returned to familiar forms of mono- or no-party dictatorship. The emergence of ubiquitous hybrid or competitive authoritarian regimes (van de Walle, 2002) casts additional uncertainty on the fate of many of Africa’s democratic experiments. In such regimes, governing parties hope to derive the legitimizing effects of democratic competition (Ake, 1996), while limiting real political contestation by maintaining an “unlevel playing field” that favors incumbents over challengers (Levitsky & Way, 2010). Today African competitive party systems are not formless but are both diverse and complex. This raises two important questions: Are Africa’s party systems really guided by mechanisms unique to the subcontinent? How does Africa’s level of party system institutionalization compare with other regions of the world and other periods of democratization?
Our article begins by highlighting the significance of party institutionalization in general and follows with a discussion of the extant research about African party systems. We then bring these two discussions together to build a complex theoretical model that takes account of both general theory and the particularities of Africa. Our theory engages four overarching explanations of party system development: social structure, economic performance, institutional constraints, and historical legacies. We test the theory using a time-series cross-sectional model that explores differences between party systems as well as variation within them longitudinally. Our analysis shows that party systems in sub-Saharan Africa are volatile but institutionalizing over time. Volatility brought about by the entry and extinction of ephemeral parties is on the decline, while volatility from healthy political competition is rising. Furthermore, historical, institutional, economic, and social factors contribute to this new pattern, just as they do in other regions.
Party Institutionalization in Africa in Comparison With Other Newly Democratizing Regions
Party systems are the most important structures of political society in democracies. Their role in representation is critical to making democratic systems responsive. They enable citizens to voice political preferences through the ballot and to hold politicians and parties accountable through the threat of withdrawing support in subsequent elections. At the same time, the distribution of legislative party strength resulting from elections is key to forming governments and passing legislation essential to stable governance. In serving these two functions, party systems lie at the core of effective democratic rule.
The importance of this topic is reinforced by an extensive literature on the correlates of party system institutionalization in new democracies from regional and global perspectives. Scholars have examined party system institutionalization in regions affected by the Third Wave of democratization, including Latin America (Mainwaring & Scully, 1995; Roberts & Wibbels, 1999), postcommunist Europe (Bernhard & Karakoç, 2011; Bielasiak, 2002; Powell & Tucker, 2013; Tavits, 2005), East Asia (Stockton, 2001), as well as cross-regionally in new democracies (Bielasiak, 2005; Mainwaring & Zoco, 2007). Much of this research shares a focus on how explosions of political activism surrounding democratization translate into political organization. Posttransition party systems are forged in a process that entails great uncertainty. With the first postdictatorial elections, parties form and disappear with only a few establishing themselves as viable electoral actors. This “great electoral lottery” (Innes, 2002) is more volatile in regions where there is little tradition of democracy, whereas there is less uncertainty where historical parties reemerged during liberalization (as in Latin America and Southern Europe).
Sub-Saharan Africa features a mixture of fledgling democracies and competitive authoritarian regimes and thus may challenge dominant findings in the literature on party system institutionalization. However, as in Latin America and Southern Europe, antecedent parties have survived democratic transitions in Africa. The most durable of these parties are extensions of the dictatorial regime and continue to dominate political space in democracies and even more so in competitive authoritarian regimes (van de Walle & Butler, 1999; van Eerd, 2010). When these parties are strong at transition, they can sow the seeds of party system institutionalization (Riedl, 2014). Because established ruling parties discourage credible competitors for power, the emergent opposition has been extremely “fluid” (Lindberg, 2007; Riedl, 2014; van Eerd, 2010), with failure to dislodge incumbents leading to high rates of entry and exit. Another complication is Africa’s dearth of political competition and party development under colonialism and postcolonial dictatorship. Unlike in Latin America where phases of competitive politics alternated with dictatorship, most contemporary episodes of competitive politics in Africa represent the first time citizens have organized and participated in a competitive party system.
The discussion above raises important differences in the roots of party systems in Africa compared with Latin America, postcommunist Europe, and East Asia. Should we conclude that party systems in Africa are not comparable with other regions and that theories drawn from the extant literature are not relevant? Using conventional measures of party system institutionalization to situate sub-Saharan Africa among other regions of the world suggests otherwise.
As Table 1 shows, the levels of party system institutionalization measured as either legislative seat or vote volatility are similar to patterns found in new democracies in Latin America and postcommunist Europe despite differences in how the party systems arose. Even if the formative stages of party systems vary in important ways, volatility outcomes in the party systems are similar. We thus believe that existing theory should be tested in the African context to establish whether African party development is unique or a simply another set of variant cases that existing theories can explain. Where existing theory fails, it is an opportunity to consider its limitations and formulate alternatives. In the next section, we discuss recent contributions to the study of African party systems and, in particular, highlight explanations of party system institutionalization that have emerged in this literature.
Levels of Volatility in Different Regions.
Lessons on Party System Institutionalization in Africa
Much like Mainwaring and Scully’s (1995) investigation of Latin American party systems, van de Walle and Butler (1999) and Kuenzi and Lambright (2001) provide “taxonomic” accounts and conceptual sketches of patterns emerging from the resurgence of multiparty elections in the early 1990s. These works, along with a number that followed, note several important trends. First, there is diversity in party systems. Countries lacking experience with democratic competition and programmatically identifiable parties often have weakly institutionalized and highly volatile party systems (Bogaards, 2008; Kuenzi & Lambright, 2001; van de Walle & Butler, 1999). However, regimes where one party “dominates” electoral competition have shown signs of party system consolidation (Bogaards, 2008; van Eerd, 2010).
Many highlight the link between party system institutionalization and openness of the political playing field. That more democratic regimes should have more vibrant party competition is a straightforward expectation, because competitive authoritarian regimes attempt to undermine challengers. Yet, the literature does not automatically assume that democracy yields more institutionalized party systems. If Bogaards (2008) is correct, electoral authoritarianism should lead to disorganized, fragmented, and short-lived opposition parties. Yet, some research shows that African democracies have less institutionalized party systems (Kuenzi & Lambright, 2005; Lindberg & Jones, 2010; Manning, 2005). In short, existing theory offers no consensus and further empirical investigation is needed.
Studies of African party systems also stress the significance of ethno-political cleavages, pervasive clientelism, and institutional legacies of colonialism. Many argue that African voters overwhelmingly support ethnic parties and multiethnic coalitions (i.e., Bates, 1974; Chandra, 2004; Latin, 1986; Posner, 2005; Young, 1976). In an environment where parties fail to distinguish themselves according to policy, ethnic solidarity is a powerful mechanism by which parties credibly commit to look after the interests of in-group members. Given otherwise unexpressed policy preferences, sharing ethnicity with a candidate might convince a voter that the candidate’s political position resembles his or her own (Ferree, 2006). Ethnicity thus may represent a cleavage structure similar to relevant religious, class, and urban–rural divisions found throughout the rest of the world (Cheeseman & Ford, 2007).
Mozaffar, Scarritt, and Galaich (2003) claim that ethnic fractionalization reduces the number of parties in a system, but that when societies are highly divided and parties have geographically concentrated bases of support, ethnic heterogeneity leads to numerous, poorly institutionalized parties. Similarly, Ferree (2010) and Mozaffar and Scarritt (2005) argue that volatility is higher when no single ethnic group can form a minimum winning coalition (MWC) and groups must coordinate to form coalitions before and after elections. Ferree (2010) also shows that volatility is higher when nested groups can form multiple MWCs and that party systems are less volatile when a single identity group can form a MWC.
Others argue that interparty competition in sub-Saharan Africa is strongly structured by clientelism. Instead of being based on shared interests, ideological commitments, or policy platforms, parties function as distributors of patronage, often along ethnic lines (van de Walle & Butler, 1999; van de Walle, 2003). Having a copartisan occupy a key political office unlocks state resources that parties can distribute privately. Clientelism also circumvents the accountability that elections impose on poorly performing governments as the distribution of private goods can be used to bolster flagging electoral support. The implication of this research is that clientelism renders the system unresponsive to performance evaluations that party systems should reflect. We join others in arguing the African voters and party systems are not so different as to preclude the use of existing theory (Brambor, Clark, & Golder, 2007; Erdmann, 2004; Lindberg & Morrison, 2008; Weghorst & Lindberg, 2011) and consequently we should expect African party systems to respond to economic performance.
As in any context, political institutions have mechanical effects on party systems that are conditioned by the peculiarities of local political traditions and cleavage structures. Presidentialism reigns supreme throughout the subcontinent and 44 of 49 regimes are presidential or semipresidential. The fortunes of parties centered on flag bearers rise and fall with those leaders (van de Walle, 2003). It also enhances the extent to which elected leaders can personalize the state and use public resources to co-opt potential challengers. This orients political competition away from legislative bodies and weakens them (Manning, 2005). Presidentialism simultaneously raises barriers for party success but also incentivizes more parties to compete. If this is the case, then it will be more difficult for party systems in presidential regimes to institutionalize.
Electoral systems are generally a legacy of colonial history rather than a product of political bargaining and choice—French and especially British colonies tended toward winner-take-all electoral systems, while Belgian and Portuguese colonies implemented proportional systems upon independence (Manning, 2005). 1 The general literature on party systems suggests that multimember districts lead to higher electoral volatility because they reduce the barriers to party entry. The implications for Africa may not be as straightforward because smaller parties based on concentrations of ethnic voters in majoritarian systems are up for grabs through patronage, thus increasing volatility. Previous research in Africa has shown either that differences between electoral systems are insignificant or that volatility is actually higher under majoritarianism. 2
This literature review highlighted specificities of African political development that may lead to different forms of party system institutionalization. In the next section, we bring this insight together with broader lessons on party system development drawn from around the world.
Understanding Africa in a Comparative Theoretical Perspective
In this section, we build a theoretical framework for understanding party system development in Africa within the confines of the general comparative literature while respecting what is unique about politics in Africa. The approach draws inspiration from broad cross-national and regional research outside Africa (Latin America, Europe, Asia). Its operationalization will adapt theory to Africa’s historical specificities. In this way, we hope that what is unique about Africa will contribute to the broader literature on party institutionalization.
Our Approach Toward Party System Institutionalization
Mainwaring and Scully’s (1995) framework has been used fruitfully to study party system institutionalization cross-nationally and at the macro-level (see Bernhard & Karakoç, 2011; Kuenzi & Lambright, 2005). They pinpoint four dimensions of party system institutionalization: (a) regularity of competition, (b) constituency stability, (c) legitimacy of actors and outcomes, and (d) programmatic political appeals to voters. The extent to which these attributes are present determines whether a party system has acquired “the value and stability” essential to institutionalization (Huntington, 1968, p. 12). We explore the degree of institutionalization in terms of constituency stability. This captures the ability of parties to attract and hold constituencies, so the first dimension of regular party competition must hold. This scope condition requires at least two consecutive multiparty elections for inclusion in our analysis.
Our measure of party system institutionalization is legislative seat volatility between elections. Pioneered by Mogens Pedersen (1979), volatility captures the extent to which party competition has stabilized over time. In institutionalized systems, small shifts at the margins from party to party will determine electoral outcomes. Volatility in well-established party systems will be low except during watershed events like party realignments (Bartolini & Mair, 1990). The classical literature anticipates that new democracies will experience high rates of volatility which will attenuate over time. Third Wave democratizers have not seen volatility fall as fast as it did historically in Western Europe and currently have more weakly institutionalized party systems than expected (Bielasiak, 2002, 2005; Bernhard & Karakoç, 2011; Drummond, 2006; Mainwaring & Scully, 1995; Roberts & Wibbels, 1999; Tavits, 2005). We recognize that volatility captures one dimension of the overarching concept of party system institutionalization but, like Mainwaring and Torcal (2006), believe that volatility allows us to precisely operationalize constituency stability and the underlying concept of party system institutionalization. 3
Factors Driving Party System Institutionalization
Sartori argues that party systems are structured by social cleavages, the system of electoral competition, and antecedent party development (1994, chap. 3). Our investigation will look at these three clusters of factors, but add a fourth. Economic performance, as has been highlighted in the voluminous literature on economic voting, shapes the fortunes of parties, particularly ruling parties, and is thus also necessary to study volatility.
Social cleavages are one of the foundational factors in party system institutionalization. Lipset and Rokkan’s (1967) well-known work on Europe highlighted urban/rural, class, and religious cleavages as essential to establishing patterns in European party systems. In the context of Africa, three cleavages are potentially important: urban/rural, class, and ethnicity. The literature on Africa stresses the predominance of ethnicity (Fridy, 2007; Ishiyama & Fox, 2006; Lindberg & Morrison, 2005). Ethnic diversity in some areas of the world has actually stabilized party systems, notably in postcommunist Europe (Birnir, 2007). We focus not just on ethnic diversity but also on the distribution of ethnic populations and their ability to form ethnic voting blocs and party coalitions (Ferree, 2010).
Many downplay the role of economic and class cleavages in Africa. However, as African economies develop and populations urbanize, class and urban/rural cleavages may play a more important role, even without fully displacing ethnicity. Higher levels of development are usually seen as an indicator of an established modern class system. As development increases, class cleavages become more stable and play an important role in the life of the party system (Lipset & Rokkan, 1967). The question remains whether such processes are advanced enough in Africa, which is still widely underdeveloped. It is possible that in its early phases, development may increase volatility by creating a more complex dual-class structure split between traditional rural and modern urban social segments, but over time it should tip the balance between sectors and reduce volatility. Unlike in other more developed regions, development may actually increase volatility in Africa.
Some authors expect that rural African voters have less stable patterns of political support because of low levels of information about incumbents, weak penetration of parties into rural areas, and the propensity of rural voters to change their affiliation on the basis of clientelistic payoffs, policy changes, or both (Stokes, 2005; Weghorst & Lindberg, 2011). Others have argued that rural voters are more stable because they hold strong partisan loyalties (Ishiyama & Fox, 2006) and yet others, less stable, as they view incumbents with greater hostility (Harding, 2010). Finally, the extent to which any social cleavage is polarized by distributional conflicts (inequality) should promote volatility (Bernhard & Karakoç, 2011).
The second group of factors is the short-term effects of economic performance. Poor economic performance should increase volatility as voters will move away from government incumbents and vote for opposition parties (Drummond, 2006; Mainwaring & Zoco, 2007; Roberts & Wibbels, 1999; Stockton, 2001). Both inflation and recession should elicit this sort of reaction from voters.
Third we investigate how institutions shape party systems. Generally, more proportional systems make party entry easier and increase volatility. At the same time, greater proportionality makes legislative survival easier and thus diminishes the prospects for the total extinction of parties, limiting volatility in comparison to more punishing systems (e.g., single-member district, majoritarian).
The nature of the executive power also affects party system institutionalization. The importance of parties to parliamentary rule promotes more stable party systems, whereas presidentialism’s separation of power may impede party system development. In Africa, the centralization of power in presidential systems undermines parties and legislatures and thus the ability of parties to retain constituencies (Mozaffar & Scarritt, 2005). An additional institutional factor is coalition government. Evidence from other regions shows that smaller coalition partners suffer when voters choose to punish incumbent governments (Fidrmuc, 2000; Tucker, 2006).
The final cluster of factors is the character of the antecedent political system. In other regions, greater attention has been paid to the age of existing parties (Mainwaring & Zoco, 2007). How the origins of Africa’s party systems shape their development requires new theoretical perspectives and a distinct set of variables to capture its unique historical legacies.
The ubiquity of postindependence dictatorships and scarce interparty competition means that the number of consecutive multiparty elections is critical to establishing both durable individual parties and party systems. Thus, it is reasonable to expect party systems with a longer history of competitive elections to have less volatility. In addition, Africa’s experience with colonialism left a weak history of political competition and self-rule, a second important antecedent factor. Experience with independent rule should decrease volatility in competitive party systems, with each year since independence/statehood increasing prospects for institutionalization. Furthermore, colonial legacies depend greatly on the colonizing power. We believe the strongest competitive legacies to be in British colonies because elected assemblies were created before independence. The French held elections but on a less widespread basis and often to elect representatives to the French National Assembly. Other powers did not even hold colonial elections at all (Collier, 1982; Zolberg, 1966).
A third-party system legacy is the degree of democracy. This is clearly very important in Africa with its mix of democratic and competitive authoritarian systems. Given Van Eerd’s (2010) theory that opposition is likely to be highly fluid in dominant party systems—an ever-changing cast of weak opponents—we believe that higher levels of democracy will promote greater institutionalization and that this will increase over time.
Because we are looking at Africa in the context of the theories generated by the literature on other Third Wave democracies, our models will be highly specified with a large number of variables. Our purpose in making this choice is to see whether existing intutions work in the African context and how the results from Africa might inform the rethinking of emergent theory in the party system institutionalization literature. We summarize our expectations in Table 2.
Theoretical Expectations of What Drives Party System Volatility.
MWC = minimum winning coalition.
We now test this complex set of expectations. First, we discuss our data, measurement, and methods, and then report our empirical findings.
Data and Measurement
Dependent Variable: Legislative Seat Volatility
We collected data for all elections held in sub-Saharan Africa since independence (or for noncolonies, the commencement of national elections). We measured party system volatility for all instances where a country held two multiparty elections in a row without interruption or intervening events like the extralegal dissolution of a legislative body or a coup d’état. 4 The observations in our analysis cluster in the two most recent decades during Africa’s “third wave,” but stretch as far back as the 1960 elections in Benin and up to the end of 2008. Our unit of analysis is a country-election and volatility is computed for country-election pairs, for which we have a total of 141 observations in 41 countries. 5 The Online Appendix Table A2 presents descriptive statistics of all variables we use in our analysis. 6
We first calculated volatility using the Pedersen approach, which sums the absolute value of the difference in percentage seat shares of all parties with legislative representation in either electoral period and then divides this total by 2. 7 This variable has a possible range from 0—a case where legislative parties have the exact same distribution of seats across two consecutive elections—to 100, where all parties that won seats in a given election fail to obtain any seats in the next. We compute volatility of parties’ legislative seat share because we are interested in questions of party system institutionalization, rather than voting behavior. 8 Data were mostly collected from Nohlen, Krennerich, and Thibaut (1999) and augmented by Lindberg’s (2009a) Elections and Democracy in Africa database and reports from national electoral commissions, Inter-Parliamentary Union (2011), International Institute for Democracy and Electoral Assistance’s (IDEA, 1998) ACE Project, and reports from the Electoral Institute for Sustainable Democracy in Africa (n.d.).
We also separate two distinctive components of volatility: volatility due to entry and exit of political parties from the system (Type A) and volatility due to established parties winning and losing voters to other established parties (Type B). Type A and B volatility are computed like the Pedersen approach, differing only in terms of which parties constitute “entering/exiting” parties and which fluctuations count for “vote-switching volatility.” 9 The two measures added together constitute “total volatility,” or the conventional Pedersen measure. While distinguishing which parties are “established” or “stable” is subject to interpretation, using legislative seat volatility makes the coding decision straightforward. For each volatility observation, we consider any political party that held one or more legislative seat in only one of the two elections and none in the other as eligible for the Type A volatility measure and any party that held one or more legislative seats in both elections for the Type B measure.
Figure 1 shows the distributional patterns of all three measures of volatility as violin plots. Each contains a box plot representing central tendencies of the data, including the median (the hollow circle) and the interquartile range of volatility for all three measures. The shaded area illustrates the broader distribution of volatility values with a Kernel density plot. 10 All three measures feature a skew with volatility scores densely located between 10 and 20 and a few observations in excess of 50. Thus, while mean total volatility is 27.6, its standard deviation is 20.7 and observations cluster around lower volatility levels. Our total measure of volatility is similar to previous studies of volatility in Africa—Ferree (2010) finds a mean Pedersen score of 26, while Kuenzi and Lambright’s (2005) is 28. At face value, Types A and B seem similar, with Type A’s mean volatility at 13.8 and Type B’s at 14.3. However, such summary statistics overlook important differences between them in our data. First, they have distinctive distributional tendencies: Type A volatility is generally lower but features a substantial rightward/upward skew up to a nearly complete defeat of seat-holding parties in a subsequent election. Second, Type B has a higher median level of volatility.
Distribution of legislative volatility by volatility type.
Independent Variables
Our explanatory variables capture the four central themes we discussed previously: social cleavages, economic performance, political institutions, and historical legacies. Explanatory factors that vary in the short run—such as economic performance and inequality—are operationalized with 1-year lags. For variables that are time invariant in the short run or pertain to outcomes of a given election (e.g., district magnitude), we use the election year observation.
Multiple ethnic MWCs
If there exists more than one ethnic MWC in a state, this dummy variable is coded as 1, and 0 otherwise. This variable captures when an ethnic group by itself constitutes a MWC and is nested within a larger group that also makes a MWC (i.e., following Ferree’s (2010) illustration, 70% of the Beninese population is made of Southerners, while the largest group within Southerners (the Fon) is 55.5% of the total Beninese population). Volatility should be greater in such polities compared with regimes with a single MWC (i.e., the Hausa in Niger).
No ethnic MWC
This is a dummy variable coded as 1 if there is no ethnic majority in a state, and 0 otherwise. This and the previous measure were based on Scarritt and Mozaffar’s (1999) data using Ferree’s coding instructions (2010). The omitted category for this and the previous measure captures observations with one ethnic MWC, so the coefficients of these two variables capture differences in volatility “relative to regimes with a single MWC.”
Rural population percentage
The variable is a continuous measure of the percentage of the total population living in rural areas (World Bank, 2012).
Level of development and economic performance
Development is measured by the log of GDP per capita (Penn World Tables, US$ 2005 purchasing power parity adjusted) one year prior to elections. Economic performance is the rate of change of the unlogged measure. These data come from Heston, Summers, and Aten’s (2012) work.
Basic needs satisfaction
We measure subsistence inequality in terms of food supply as kilocalories per capita per diem. When the measure increases, it indicates an improving food supply for poor segments of the population (Reenock, Bernhard, & Sobek, 2007). Unlike poor populations, the well-off do not eat more as they grow wealthier, but eat higher quality and more expensive food. Thus upward movement in this measure represents the narrowing gap between the caloric consumption of rich and poorer populations. This measure has several advantages over income distribution in poorer countries. First, it is available for a much wider sample of countries from 1960 to the present. Second, in underdeveloped contexts, income distribution may miss improvements in the life chances of poorest segments of the population. The data were collected from FAOSTAT (2009).
Inflation
We use the GDP implicit deflator, which captures the overall rate of price change in a country’s economy, computed at constant levels of local currency (World Bank, 2012). The distribution of this variable is skewed, with its tail pulled rightward due to catastrophic hyperinflation in places like Zimbabwe. It also has negative values (deflation) in a handful of instances. Following Shadden and Zorn (2011), we address both distributional issues by rescaling the data with the “V-curve” approach. This entails log transforming the absolute value of inflation and including a dummy variable that flags instances of deflation.
Mean district magnitude
This is a continuous variable, ranging from 1 (SMD systems) to highly proportional systems (60). The data come primarily from Bormann and Golder’s (2013) work and are augmented by Beck, Clarke, Groff, Keefer, and Walsh (2001) and the Inter-Parliamentary Union (2011).
Parliamentarism
This dummy variable codes parliamentary systems as 1 and presidential and semipresidential systems as 0. These data were coded by the authors.
Previous government coalition
This variable is coded as 1 if, for a given election, the previous government included more than one party, and 0 otherwise. This was coded by the authors.
Colonialism
This set of dummy variables identifies former British colonies and “other or no colonialism.” These data come from Hadenius and Teorell’s (2005a) work. French colonies are the omitted group, so the coefficients of these two variables compare volatility with former French colonies.
Combined polity
We use Hadenius and Teorell’s (2005b) Imputed Freedom House (FH) Polity 2 Index. The measure has an advantage over Polity or FH alone in superior coverage across countries and over time. It also combines evaluations of de jure political institutions (Polity) and de facto political rights (FH) that should both influence party system outcomes. The measure has the range [0, 10].
Years after colonialism
This measures the number of years a given observation is from the date of colonial independence. In the case that a state never experienced colonialism, it is coded at the onset of statehood. It was computed by the authors.
Elections in a row
This captures the number of consecutive uninterrupted elections held prior to that point in time, including the current election.
Analysis and Results
Our estimation strategy uses a time-series, cross-sectional model. This allows us to account for both change within political systems over time and differences between countries arising from slow moving or time-invariant factors—like colonial legacy. We prefer this strategy over ordinary least squares (OLS) regression for several reasons. First, country panels are unbalanced. Because many countries in sub-Saharan Africa have experienced episodes of competitive multiparty elections punctuated by military interventions, an OLS regression model could overrepresent countries with longer standing multiparty traditions and sequences of unbroken democratic competition (Mauritius, Botswana). Second, a central lesson of party institutionalization is the importance of longitudinal trends, and we want our analysis to capture this. We further have evidence that repeated elections in Africa enhance interparty competition over time (e.g., Lindberg, 2006). Finally, patterns are significantly different across country units, suggesting that a time-series cross-sectional procedure is superior to OLS estimation. 11 We use a random-effects model organized by country elections. We capture time with a variable that counts the total number of multiparty elections held by a country prior to that date. 12 We begin the results section by reporting the models in which total volatility is the dependent variable (see Table 3).
Random-Effects Estimations of Total Legislative Seat Volatility.
Models 2-5 control for instances of deflation. Standard errors are given in parentheses. MWC = minimum winning coalition; PPP = purchasing power parity; MDM = mean district magnitude; FH = Freedom House.
p < .1. **p < .05. ***p < .01.
Explaining Total Volatility
Ferree (2010) argues that in regimes where there exists one ethnic minimum winning coalition (MWC), dynamics of interparty competition will be more stable, as parties know the cleavage on which to seek out supporters. When this is not clear—there exist multiple ethnic MWCs or none at all—campaign and recruitment strategies for party elites are less straightforward. We find partial support for this theory. Multiple MWCs are linked to greater volatility but systems with no ethnic MWC are not systematically more volatile.
Do party systems in the Africa reflect other social cleavages? Do they create group identities that parties can capture and turn into reliable constituencies? We do not find that economic development reduces volatility yet in Africa as it does in other regions. It is signed positively, as we theorized that it could be at low levels of development, but it misses conventional levels of significance. The findings for total volatility also do little to resolve the conflicting theories over whether a sharper rural/urban divide increases or decreases party system volatility. Our analysis does not provide evidence that urbanization influences party system institutionalization.
We do, however, discover that inequality in basic needs satisfaction increases levels of volatility. In regimes with high levels of inequality, conflicts over redistribution are central to interparty competition. A cleavage structure built upon such distributionally harsh divides fosters polarized parties and leads to substantially higher levels of volatility. Importantly, the relationship represents a trait common to African party systems and other developing regions of the world. It also reinforces the point that economic voting in developing democracies requires attention not only to economic performance but to distribution as well (Bernhard & Karakoç, 2011).
Next we turn to the impact of economic performance on total volatility, where tests do not support the hypothesis that positive performance dampens volatility. Neither economic growth nor inflation drives variation in party system institutionalization. While this might lead to the conclusion that African voters are not economic voters, additional findings we present below paint such an assessment as hasty.
Turning next to the institutional variables, greater district magnitude is associated with lower levels of volatility. Recall that general theory predicts that district magnitude affects the number of parties within a political system, increasing barriers to entry and thus reducing volatility (Taagepera & Shugart, 1989). In contrast, our study follows others in finding the opposite in Africa (Bogaards, 2008; Ferree, 2010), although we are unique in establishing statistical significance. Tavits (2005) provides theoretical justification for such a finding. She argues that higher mean district magnitude can decrease volatility because under highly proportional voting systems that admit a large number of parties, dissatisfied voters may “shop around” and choose alternate parties, without causing the kind of drastic party exit that occurs in winner-take-all and other highly restrictive voting systems.
Regimes governed by coalitions generally seem to experience substantially higher volatility in the subsequent election in our sample, although the result is not robust to our fully specified model. This is consistent with the expectation that governments that must balance the demands of multiple political parties will be less effective, particularly if compromise between parties alienates supporters. We do not find evidence that parliamentary systems promote more institutionalized political parties. Presidential regimes in Africa are no more volatile than parliamentary ones.
Finally, we consider the impact of historical and temporal variables. In comparison with former French colonies (the omitted category), countries that were colonized by a power other than Britain and France and those not colonized are less volatile. The average volatility level for former French colonies in our sample is about 30% higher than the mean for British and other/noncolonies (24.8 and 24.7, respectively, in comparison with 32.9 for French colonies). Former British colonies on average do have lower levels of volatility than French ones, but this does not meet conventional levels of significance. But, our finding regarding the “other or no colonialism” category is opposite longstanding expectations. We think the finding on this third category may be the product of the types of observations filling this group. Many of these regimes have particularly stable interparty competition because they are either dominant party systems (e.g., Mozambique), highly repressive (Rwanda), or some combination of both (Ethiopia).
Democratic and authoritarian regimes share similar patterns of party system institutionalization, a finding that is robust to many specifications. 13 The same holds for dominant versus nondominant party regimes (not displayed; see Online Appendix A4). While the level of democracy does not seem to matter, more experience with self-government, independent of regime type, measured as the number of years since the end of colonialism, actually contributes to higher volatility. This aligns with other research finding that volatility in sub-Saharan Africa is actually increasing over time (Bogaards, 2008), despite the expectation that volatility will diminish as parties develop stable support bases and deter new entrants. Even in democracies, we have not yet seen the beginnings of this process, as the interaction term between the level of democracy and the number of uninterrupted elections is not systematically related to volatility.
Disentangling Party Entry and Exit (Type A) From Vote Switching (Type B)
Following Powell and Tucker (2013), we believe that the Pedersen measure of volatility aggregates two different trends: volatility stemming from the entry and exit of political parties (Type A) and that from an increase or decrease of seat shares among well-established parties (Type B). As in the previous section, we use a random-effects time-series model to estimate the drivers of Type A and Type B volatility. Table 4 shows the resulting estimations for our fully specified model (Model 5 in Table 3). The findings in Table 4 illustrates how total volatility may reflect different things: variables that affect the combined total volatility in the system, or which only contribute to the birth and death of parties, or those which affect interparty competition between existing competitors.
Comparing Total, Type A, and Type B Volatility.
All models control for instances of deflation. Standard errors are given in parentheses. MWC = minimum winning coalition; PPP = purchasing power parity; MDM = mean district magnitude; FH = Freedom House.
p < .1. **p < .05. ***p < .01.
What makes for a system of ephemeral parties that fail to win sustained support over time? Our analysis offers few conclusions on what drives Type A volatility. The one strong finding seems to be coalition governments hold potentially high costs. Governing coalitions lead to potentially fatal ramifications for parties on whom voters place blame for government shortcomings (Tucker, 2006). Why might this be the case? The primary members of any coalition—often longstanding incumbent parties in Africa—want coalitions with weak parties who will be amenable to their government objectives. Observers often relate that incumbents can manipulate opposition challengers and use various tools of co-optation to gain sympathetic partners and maximize their control within the regime (Lust-Okar, 2004; Svolik, 2009), turning credible rivals into “cosmetic opposition.” Such parties have little long-term viability, contributing to Type A volatility, because when voters choose to punish incumbents, such weak and compliant parties stand little or no chance in the next round of elections.
While our total volatility analysis could be used to argue that African party systems are different from other regions of the world, our examination of Type B volatility shows a much greater degree of similarity. Volatility due to vote switching between parties—considered a hallmark of accountability in well-functioning electoral regimes—occurs as a product of both patterns of social cleavage and economic performance. We see that where political parties can build support bases from multiple ethnic groups, volatility is higher. This is evidence that parties are competing for citizens along better established cleavage lines with some degree of effectiveness. We observe a similar change with the finding regarding rural populations. In our analysis of total volatility, it did not have a significant effect (although it did approach significance). When we decompose volatility into Types A and B, we see that the variable is signed in different directions, and that the negative direction on Type A hides the importance of rural population share in determining the level of vote switching between parties. The positive effect of more rural populations on Type B volatility provides evidence that the persistence of the traditional sector promotes interparty competition. The macro-level nature of our analysis does not allow us to pinpoint which voters switch parties in more rural societies, but this view is consistent with theories that argue that rural African voters are more likely to switch the party for which they vote (Harding, 2010; Stokes, 2005; Weghorst & Lindberg, 2011).
The contrast between the aggregated and disaggregated findings also highlights important differences in how economic performance and basic needs satisfaction affect volatility. Once again for economic performance, we find that the coefficients for Types A and B are signed in the opposite direction and, when we disaggregate the two, growth emerges as an important deterrent to vote switching between established parties. This is congruent with theories of economic voting and the findings for many other regions of the world and cross-regional samples. In contrast, basic needs satisfaction, which was systematically associated with total volatility, does not achieve significance in the tests for either Type A or B. It also raises an important consideration for scholars when theorizing about different volatility types and which factors drive Type A and Type B volatility. Variables like basic needs satisfaction—which jointly capture the viability of class-based parties and their sustainability (likely to be reflected in Type A volatility) and voter evaluations of how incumbent governments distribute the gains or losses of economic performance under their rule (likely reflected Type B)—may shape party system volatility, but only when considering their impact on both types of volatility combined. We note a similar loss of significance for the previous colonial power variables when we decompose total volatility into Types A and B.
Finally, the decomposition of total volatility in Types A and B provides additional insight into the nature of volatility over time. As countries move further from colonialism, they experience more party system volatility in general. Such a finding casts doubt on whether African party systems are evolving in the direction of greater institutionalization and if they will soon exhibit the stable patterns of party competition that have been associated with established systems. When we disentangle the two types of volatility, however, we find that the fashion in which volatility is increasing in African party systems may not be so troubling.
New party systems experience high rates of entry and exit of parties and we expect, as a party system institutionalizes, the volatility induced by such parties to drop over time. As the system coalesces into viable long-term political alternatives, volatility should increasingly occur between these organizations. By disaggregating total volatility into Types A and B, we see that increasing interparty competition over time largely drives the relationship between more experience with self-government and volatility. This is evidenced by the positive and significant coefficient for years after colonialism for Type B (interparty) volatility.
If we examine trajectories of Type A and Type B volatility in our sample, we see evidence of a maturation of African party systems. Figure 2 shows that the rate of entry and exit for parties without long-term organizational capacity is diminishing. At the same time, we also see an increase in interparty competition since independence.
Type A and Type B volatility in Africa over time.
The figure plots the mean levels of Type A and Type B volatility in sub-Saharan Africa since 1960. The patterns differ dramatically. First, Type A is overall much higher than Type B generally, but this begins to change as we approach the present. Type A spikes significantly in the 1990s when multiparty elections were held for the either for the first time or after decades of dictatorship. However, as countries have adjusted to multiparty competition, the players have become more stable, illustrated by the drop-off in Type A volatility in the 2000s. Type B volatility has steadily increased since its trough in the mid-1980s. It has remained relatively high but stable across the era of competitive politics with some variation up and down over the years. The figure reinforces how analyses of total volatility might mask important distinctive trends in volatility Types A and B, and further, the importance of accounting for longitudinal patterns in volatility data. We take this as evidence of the beginnings of party system institutionalization. The players are establishing themselves and competition for voters, rather than the institutional failure of parties, explains why total volatility has remained relatively high in the period of competitive politics.
Discussion and Conclusions
Political parties and party systems in sub-Saharan Africa have traversed a path from the upheavals of independence and repression under postcolonial dictatorships to a rapid resurgence of multiparty politics in the late 20th century. Electoral politics in Africa are frequently described as distinct from other regions in the world and as guided overwhelmingly by clientelism and parochial ethnic identities. Yet, such accounts of its party systems overlook wider evidence of polities in formation. Sustained political competition has allowed parties to establish track records and develop new capacity. Voters have become more sophisticated in evaluating the political system. Our article builds on a number of recent works studying what drives party system institutionalization in African polities. Using a new data set covering all uninterrupted multiparty election pairs in the continent from 1960 to 2008, we have presented an analysis of the determinants of legislative volatility. The article’s approach accommodates the specificities of African regimes while also bringing to bear conventional explanations of party volatility around the world.
Our account suggests that some of Africa’s unique social and historical factors have important implications for understanding interparty competition and the nature of party systems there. In particular, we confirm that when nested ethnic groups can form multiple MWCs, the option that members of this group have to coalesce as a political constituency in one or the other group generates less party system stability. Furthermore, the legacy that colonialism has left in African politics is striking and its effect on party system institutionalization is no different. For one, former French colonies in Africa have dramatically higher levels of volatility than other countries, even when taking into account the specific electoral institutions that they left behind. And even more importantly, more time since the end of colonization generally drives higher overall volatility and volatility as a result of the transfer of votes from one established party to another.
We also show that in some ways African party systems resemble those found in other regions of the world. First, some evidence about the impact of social structure aligns with classical accounts of party system development. To this end, we show that party systems with more rural voting populations have higher levels of volatility, but that this is driven by competition between existing parties and not the entry and exit of parties. Moreover, we found that a failure to see to the incompressible social needs of the population (measured by the average per capita caloric supply) leads to a divisive distributional politics that promotes volatility.
When we examined total volatility, we had a reason to believe that economic performance does not systematically affect volatility, but when we discomposed volatility into Types A and B, we had to reconsider. Economic performance did not affect the entry and exit of parties, but was a robust predictor of volatility as a product of change in seats between existing parties. On the whole, we believe our data present a message that when governments manage development poorly, the party system responds, in ways unique to Africa but in line with long-held beliefs about economic voting.
We also note some very important differences between Africa and the rest of the world. Institutions seem to have little impact compared with other regions of the world. And where they do matter, they operate in markedly different ways. Despite arguments in the literature that presidentialism should contribute to volatility, we find no evidence of this in Africa. The one variable that does attain significance, mean district magnitude, operates in a direction opposite of what the literature expects. Greater proportionality increases volatility rather reducing it as it does in many other regions. And coalitions governments only seem to have an effect on Type A volatility, increasing it, and thus promoting the exit of existing parties and the entry of new ones.
A final finding of our study joins in a growing debate about the nature of party system stability over time. Scholars from Lipset and Rokkan forward have suggested that party system volatility might rise during the establishment of multiparty politics but eventually cleavages would “freeze” in place and systems would stabilize. Yet, patterns of party system development in new democracies—particularly in postcommunist Europe and sub-Saharan Africa—have raised doubt about the applicability of this hypothesis in the late 20th and early 21st century. Like others, we found that as countries move away from colonialism, volatility actually rises.
Nevertheless, we noted important emerging patterns in competition in African party systems. Our decomposition of total volatility in Types A and B showed that volatility as a product of entry and exit of parties has diminished substantially since the advent of competitive multipartism in Africa in the 1990s. At the same time also observed was an increase in volatility as the product of competition between established parties. We see these trends as important evidence of the beginning of party system institutionalization. Assuming that they continue, the prospects for stable interparty competition seem hopeful. Our use of these disaggregated forms of volatility and their utility in uncovering what lay behind our estimates of the correlates of total volatility and what they left hidden affirms Powell and Tucker’s (2013) ideas on moving beyond the original Pedersen index. In sum, our article presents an image of party system institutionalization in Africa as both intriguing in its unique character but far from as different as many accounts have suggested in the past. Much work still remains to be done in this vein of the literature. The time is nigh for taking seriously party system institutionalization in sub-Saharan Africa and studying it with the attention other regions have received
Footnotes
Acknowledgements
The authors wish to thank Dong-Joon Jung, Adrienne LeBas, Samuel Handlin, Rachel Riedl, and Tristan Vellinga for valuable feedback on the manuscript, as well as three anonymous reviewers for suggestions which enhanced this piece. As always, errors and omissions are the sole responsibility of the authors.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.