Inglés (pdf)
Articulo en XML
Referencias del artículo
Traducción automática
Enviar articulo por email
Citado por SciELO 
Similares en
SciELO 
Permalink
It is well established that free and fair elections are a necessary but not sufficient condition of democracy (Schmitter & Karl, 1991), and Africa is no exception to this convention. Following the third wave of democratisation, the proportion of countries organising multiparty elections rose from about 25% in 1988 to 92% in 2012; 80% of these had held four or more elections, and 60% six or more elections (Van Ham & Lindberg, 2015). But what has the transformative impact of elections in Africa been? A decade ago, Lindberg claimed that repeated elections, even if not entirely free and fair, would promote democratic qualities in the long run (Lindberg, 2006a, p. 149). However, 30 years after the inception of democratic experiments in Africa, it is clear that most countries did not move linearly towards democracy. Although almost all countries in the region have introduced liberalisation reforms and conducted multiparty elections, only a few transitioned to a liberal democracy and are experiencing democratic consolidation (Bratton & Walle, 1997; Cheeseman, 2015; Lynch & Crawford, 2011).
In a study that covered elections held between 1990 and 2010, Lynch and Crawford demonstrate that while some African countries experience democratic progress, others slide back towards authoritarianism; but more importantly, most countries remain stuck as “‘hybrid regimes’, which are neither fully democratic nor classically authoritarian” (Lynch & Crawford, 2011, p. 281). Bleck and Walle (2019) recently argued that elections failed not only to generate democratic consolidation, but also to change the patterns of electoral competition. The authors demonstrate how incumbent presidents rely on an array of mechanisms - access to state resources, control of clientelistic and patronage networks - to survive electoral contests. Nevertheless, a handful of democratic success stories - Cabo Verde, Ghana, São Tomé and Príncipe, Mauritius and Senegal - show that frequent turnovers in government can go hand in hand with democratic developments (Sanches, 2018a). Thus, the relationship between elections, turnover and further democratisation is at best far from univocal, and includes experiences of success and failure (Cheeseman, 2015).
This study examines whether elections are vehicles for change (a term we borrow from Cheeseman, 2010) in terms of both democratic performance and alternation in government. It addresses two fundamental questions: Does having more regular elections improve democratic performance in Africa? And have elections transformed the patterns of alternation in government? These questions will be successively analysed in two empirical analyses.
The first empirical analysis tests whether elections have contributed to further democratisation in Africa. It takes democratic performance as the dependent variable and focuses on government alternation (in the founding multiparty elections and subsequent elections) as the key explanatory variable. In theory, consolidated democracies are expected to pass the “two-turnover” test so as to prove losers’ acceptance of the rules of the game (Cheeseman, 2010; Gasiorowski & Power, 1998; Huntington, 1991). Political alternation in this sense represents a moment when democratic procedures become accepted and routine, and are supported by the majority of the society, including the political actors (Linz & Stepan, 1996). However, in practice things are far more complex. On the one hand, most elections in Africa have not promoted democratic consolidation, which means democratic performance varies greatly (Bratton & Walle, 1997; Lynch & Crawford, 2011; Cheeseman, 2015). On the other, although new opposition parties emerged during and after democratic transition, it is uncertain whether they were able to push democracy forward and accept the democratic rules of the game. Therefore, turnover may represent fresh opportunities for changes in the rules of competition but may not necessarily result in democratic gains if parties fail to act as democratic forces (Cheeseman, 2010; Wahman, 2014).
The second empirical analysis examines whether elections have changed the patterns of alternation in government. It has been argued that elections have brought little political change to Africa and that authoritarian parties/presidents have managed to thrive in the new multiparty era (Bleck & Walle, 2019; Bogaards, 2004; Rakner & Svåsand, 2002). Nevertheless, great variation has been found with some countries experiencing more episodes of transfer of power than others (Cheeseman, 2010; Sanches, 2018a). Understanding the sources of this variation is the main motivation here. The second analysis takes government alternation as a dependent variable, while alternation in government in the founding multiparty elections, political competition and fairness of elections are the key independent variables. Overall, the two studies will allow us to understand whether elections have been a vehicle of change in Africa.
The two empirical analyses draw on an original time-series cross-sectional dataset of 179 elections in 27 African countries, from the founding multiparty elections until May 2019. The main finding for the first study is that while government alternation in the founding elections impacts democratic performance over time, opposition victories in subsequent elections do not produce democratic gains. This means that opposition parties/leaders - as much as incumbent parties/leaders - fail to act as agents of democracy. The results of the second study reveal that alternation in government in the founding elections, the level of political competition and the quality of elections significantly increase the odds of government alternation in future elections.
Overall this study reveals that alternation in government and democratisation strengthen each other. On the one hand, alternation in the founding multiparty elections shapes the prospects of democratisation; on the other, democratic principles such as political competition and clean elections increase the likelihood of turnover. These findings contribute to literature linking elections, democracy and turnover in Africa.
The study is organised as follows: The two theoretical sections outline the arguments and hypotheses linking elections to changes in democratic performance and in patterns of alternation in government over time. The methodological section clarifies how the main hypotheses will be tested, operationalises the dependent and independent variables in each study, and presents data on democratisation and government alternation in Africa. The empirical section discusses the main results of the statistical models estimated. The conclusion reflects on the implications of our findings for future research on democracy, elections and turnover in Africa.
Elections as a vehicle of democratic performance
Democratisation is far from a linear process as it involves complex dynamics, timing and sequences between liberalisation, transition and consolidation (Schneider & Schmitter, 2004). Focusing on democratisation rather than its outcome - democracy - allows us to observe regime performance in Africa over time and space, and why some countries manage to progress while others experience democratic backslide or remain stuck in hybrid regimes. Democratisation is usually explained on the basis of certain requisites such as homogenous social structure, economic development (or modernisation), pro-democratic political culture, power sharing political institutions and international context (Gasiorowski & Power, 1998; Norris, 2008; Teorell, 2010). Democratic performance, as it is used here and will be explained later, is assessed using as references political rights, civil liberties, checks and balances, rule of law, and the quality of elections.
In the African setting, scholars have put forward innovative theses of democratisation, one of which addresses the role of elections. In his seminal work The Surprising Significance of African Elections, Staffan Lindberg (2006a) made an important contribution to the prolific debate on how elections affect democratisation. Whilst acknowledging that elections are not the only nor the main explanatory factor for democratisation, Lindberg suggests that holding repeated elections in the new electoral regimes in Africa fosters and develops democratic qualities (Lindberg, 2006a, p. 149). Under Lindberg’s main argument, even if regular elections in the case of sub-Saharan Africa were flawed, they contributed to democratisation in the long run as they stimulated processes and behaviour that change citizens’ and institutions’ practices and attitudes toward democracy. In this sense, elections turn citizens into voters who demand responses and hold their representatives accountable by encouraging them to pressure political leaders to defend democratic principles, thus promoting a democratic mind-set in society. Moreover, civic organisations learn and create social capital to defend political rights and civil liberties, state institutions and officials see the importance of defending civil liberties, and the media takes on an increased role in protecting and promoting democracy (Lindberg, 2006a).
Reacting to Lindberg’s argument on “democratisation by elections”, Bogaards (2013) argues that it is also true that multiparty elections helped perpetuate non-democratic leaders on the continent. Democratisation by elections also has different patterns of stability and change, with free countries remaining free and electoral autocracies remaining autocracies. Based on the analysis of 324 multiparty elections up to 2011 in 43 countries in sub-Saharan Africa, Bogaards concludes that only five of those that held at least two consecutive multiparty elections went through a democratisation process by elections, and two of these reverted to electoral authoritarianism. In the same vein, based on the cases of Southeast Asia where authoritarian regimes oversaw 88 competitive elections from 1945 to 2015, Morgenbesser and Pepinsky argue that elections did not necessarily result in democracy in the region over time, since only two countries - the Philippines and Indonesia - are currently electoral democracies (Morgenbesser & Pepinsky, 2018). According to Bogaards, these findings call for the revision of the argument of democratisation by elections and the need to combine electoral reform with conditions that favour alternation in power, which is considered a key element for democratisation in the recent growing literature of the field (Bogaards, 2013).
Turnover is an important benchmark of the democratisation process (Lodge, 2013). The argument is that democracy establishes itself when an incumbent party loses an election, and it is consolidated when at least two turnovers occur after the introduction of democratic elections (Huntington, 1991). As mentioned previously, the regularity of elections on the continent did not necessarily result in full-fledged democratisation. One of the critical points in the process is turnover, with many examples of leaders or parties maintaining themselves in power despite the regularity of elections and the introduction of limits on terms (Cheeseman, 2010; Tull & Simons, 2017).
Most studies in the field focus on power transfers at the presidential level, showing that specific traits and trends of African politics have prevented the continent from democratically ensuring a transfer of power (Cheeseman, 2010). Military takeover of power is one such trait and, albeit not specific to the continent, it has implications on turnover. In Africa, leaders who took power militarily were found to be reluctant to relinquish power voluntarily and tend to stay in power until they die or their health fails (Durotoye, 2016). Although the African Union and international donors’ organisations oppose the unconstitutional prolonging of terms, they have been neither effective nor proactive in enforcing such rules and have allowed African leaders to resort to these practices in impunity (Durotoye, 2016; Tull & Simons, 2017).
The so-called “third-termism” or “third term bids” is another growing trend; this hallmark of Africa’s democratic backsliding is characterised by the manipulative stance of democratically elected leaders altering the Constitution to change the terms’ limit to remain in power longer, thus hampering the prospects of alternation in power (Durotoye, 2016; Tull & Simons, 2017). From 1990 to 2016, over 10 African presidents tried to amend the constitution to alter the term limits (Durotoye, 2016; Tull & Simons, 2017) and there has been a democratic backsliding in the continent (Bogaards, 2013). Even countries formally hailed as success stories like Mozambique are experiencing a democratic backsliding (Muchemwa & Harris, 2018). Term limits are key mechanisms for the institutionalisation of power and for democratisation because they influence the way leaders come to office or leave power in a context where, according to some scholars, politics have been historically dominated by “big men” that are particularly resilient and good at manipulating institutions to stay in power (Tull & Simons, 2017).
Given this scenario, what is the (democratic) role of opposition parties? Can turnover at the government level lead to further democratisation? Rakner & Walle (2009) posit that strong opposition parties are a central component of any strategy of democratisation by elections, and that “opposition parties’ evolving ability to compete politically should in theory correlate with the level and quality of democratic practice” (Rakner & Walle, 2009, p. 109). However, several studies suggest that the democratic role of opposition parties is far from guaranteed.
Drawing on a dataset consisting of 210 elections in electoral regimes in Africa, Lindberg (2006b) demonstrated that opposition parties’ participatory behaviour - that is their decision to either contest or boycott elections - significantly improved the democratic qualities of elections and he therefore concluded that opposition behaviour was an “important determinant of democratisation by elections” (Lindberg, 2006b, p. 133). On the other hand, Cheeseman (2010) relies on illustrative case studies to highlight the variation and unpredictability of turnover (at the presidential level) in Africa: while Benin witnessed positive improvements after the transfer of power in 1991, this was not the case in Zambia or Kenya. Using data on competitive authoritarian regimes between 1989 and 2008, Wahman (2014) offers additional insights into this discussion. The study starts by demonstrating that newly elected governments are not more prone to democratic development than re-elected ones, before specifying the conditions under which this may vary. Wahman argues that when elected governments are uncertain about the prospects of their re-election - which is more common in weakly institutionalised party systems - they will be more reluctant to reduce the incumbents’ advantage through further democratisation (Wahman, 2014, p. 221). This argument is illustrated with the case studies of Senegal, Kenya and Ghana.
We aim to contribute to this discussion by re-evaluating the conditions under which government alternation leads to democratisation. This is innovative as the trend in the field has been to analyse power transfers at the presidential level (Bratton & Walle, 1997; Cheeseman, 2010), thus neglecting the governmental arena, and the role of political parties. The first explanation here proposed focuses on path dependency, and argues that government alternation during transition helps improve democratic performance in the future. Several scholars have developed a path dependency argument to explore how the mode of transition from authoritarian rule shaped the prospects of democratic consolidation in Southern Europe and Latin America (Munck & Leff, 1997; O’Donnell & Schmitter, 1986; Schmitter & Karl, 1991). Our goal is to extend this logic in order to understand the long-term impact of government alternation during transition on democratisation.
The minimalist definition of path dependency is that “what has happened at an earlier point in time will affect the possible outcomes of a sequence of events occurring at a later point in time” (Sewell, 1996, p. 263, cited in Mahoney & Schensul, 2006, p. 459). This is expected because as the process advances the “costs of exit” - switching to some previously plausible alternative - rise (Pierson, 2000, p. 252). Democratic transitions can be conceived as “formative or founding moments” that shape subsequent political developments (Munck & Leff, 1997, p. 343). They affect “the form of post-transitional regime and politics through” their “influence on the pattern of elite competition, on the institutional rules crafted during the transition, and on key actors’ acceptance or rejection of the new rules of the game” (Munck & Leff, 1997, p. 345).
During the African democratic wave, many countries held multiparty elections for the first time ever or in decades. Bratton and Walle (1997, p. 196) allude to the relevance of founding elections when saying that “they heralded regime transition in the sense that they laid down new procedures of political competition that broke down with monopolistic authoritarian antecedents and established the potential for democratic rule”. The numerous opposition parties that formed during this period and participated at the polls had different degrees of success, as incumbents managed to be elected in many countries. Nevertheless, the losers’ acceptance of the results in countries like São Tomé and Príncipe, Cabo Verde or Benin, to mention just a few examples, demonstrated a new attachment to democratic rules and principles (Bratton & Walle, 1997, p. 197). In sum, founding elections can be perceived as key moments when pro-democratic behaviours and norms emerge, and this can shape regime performance over time. Experiences of turnover at the transition stage are, therefore, meaningful, not because they immediately lead to democracy but because they introduce changes in the status quo, hence:
Hypothesis 1 (path dependency): Government alternation in the founding multiparty elections positively impacts democratisation over time.
However, the democratic gains of turnover after the transition are less obvious over time. As previously discussed, opposition parties do not always act as forces of democratisation (Cheeseman, 2010) and once in government they may try to revert democratic gains if they fear electoral defeat (Wahman, 2014). The current trends of democratic backslide and authoritarian resurgence in countries that have experienced turnover - at the executive and/or presidential level - suggest that a power shift may not necessarily lead to further democratic development. Thus:
Hypothesis 2 (democratisation by turnover): Government alternation following the founding elections has a smaller effect on democratisation.
Elections as vehicles for government alternation
Several studies suggest that multiparty elections have had a limited transformative impact on African party systems: incumbent authoritarian parties and leaders managed to be re-elected (Bleck & Walle, 2019; Rakner & Svåsand, 2002) and dominant party systems have proliferated in most countries (Bogaards, 2008). However, more variation emerges when we revisit the studies that analysed the impact of elections on party systems using indicators of stability (Lindberg, 2007) or institutionalisation1 (Kuenzi & Lambright, 2001; Rield, 2014; Sanches, 2018a). But regardless of the metric used, these studies reveal that party systems differ substantively in how they operate - some countries have experienced more change between elections while others less.
To explain diversity in the level of stability or institutionalisation, scholars have resorted to a set of explanations based on history, economy, political institutions and social structure. It has been argued that authoritarian legacies - strategies of power accumulation and clientelistic networks established during authoritarianism - have a greater impact on African party systems than economic performance and type of political institutions (Riedl, 2014). Ferree (2010) and Weghorst and Bernhard (2014) provide some support for social structure explanations when showing that parties’ ability to form winning coalition of ethnic groups helps stabilise the party systems. Finally, incumbency is an important advantage in the context of an uneven playing field where opposition parties’ lack of access to state resources affects their ability to distribute patronage and develop strong nation-wide structures (Bleck & Walle, 2019; Cheeseman, 2010; Wahman, 2017).
In what regards political alternation, several studies draw interesting descriptive inferences from the analysis of presidential elections (Bratton & Walle, 1997; Cheeseman, 2010; Lodge, 2013). Looking at elections from the 1990s until 2010, Lodge (2013) finds that 15 countries experienced an orderly succession of presidents of different party affiliation either during or following the transition, while 12 countries did not experience power transfer - not even once. Some variation is also captured by Cheeseman (2010) using data from presidential elections held between 1990 and 2009. Finally, Sanches (2018a) measured the patterns of alternation in government in Africa using data on 178 elections held in 30 (hybrid or) democratic countries between the founding multiparty elections and 2016. The study shows great variation across countries, with alternation in government occurring in half of the elections held in the observed period.
While these studies are important contributions for our understanding of electoral politics in Africa since the introduction of multiparty elections, the question of why some countries experience more alternation in government than others remains unanswered. To contribute to this discussion, the explanatory model developed here focuses on the effects of path dependency, the level of political competition and the quality of elections as key explanatory factors. The rationale connecting each of these factors to alternation in power is discussed next.
The first explanation is that current trends of electoral competition in Africa trace back to the transition stage. Parties’ ability to succeed in the founding multiparty polls crucially shaped both competition in the short-term and future rounds of elections (path dependency). As explained previously, this is expected because the democratic transitions are critical junctures of party formation and institutional crafting during which party elites (incumbents and newcomers) seek to minimise the odds of losing power. The choices they make at this point will influence the kind of competition that will emerge in the future even if they are confronted with high levels of uncertainty (Lupu & Riedl, 2013, p. 1344). Studies from South European countries are illustrative of the importance of the transition. In fact, they show how this period had a “significant impact in shaping the new parties (party unit level) and relations among parties (party system level)”, namely by determining patterns of coalition and alternation in power over time (Cotta, 1996, p. 73).
In the African setting, several studies make path dependency a plausible argument when explaining patterns of government alternation; we elaborate this further. It has been demonstrated that second elections in Africa resembled the early founding round in terms of winners and losers and levels of political participation (Bratton, 1998). Voters “generally sought continuity in political leadership in the aftermath of turbulent interludes of regime transition” (Bratton, 1998, p. 61). And studies covering more years and elections reveal that the events of the early 1990s crucially shaped future rounds of elections (Bleck & Walle, 2019). The incumbent’s success in the founding multiparty elections was crucial for political dominance/stability over time and to reduce the opportunities for alternation; but once a country experiences one alternation, the probability of another in future elections increases (Bleck & Walle, 2019). Thus we formulate that:
Hypothesis 3 (path dependency): Countries that experienced alternation in government in the founding elections are more likely to experience it again in future rounds of elections.
The second explanation observes levels of competition during elections. According to Bartolini and Mair (1990), electoral volatility is influenced by the number of parties in competition as the probability that each individual voter will vote for the same party in two consecutive elections will decline as the number of different available options increases. In other words, limited political offer is expected to lower the level of volatility and thus the likelihood of electoral shifts in government. Early empirical assessments of elections in Africa recognise that leadership alternation2 was associated with close races (Bratton, 1998, p. 64); and subsequent research found a positive relationship between the number of electoral parties and party system institutionalisation (Kuenzi & Lambright, 2001; Sanches, 2018a) or stability (Ferree, 2010). This discussion informs the following hypothesis:
Hypothesis 4 (level of political competition): The larger the number of relevant electoral parties, the greater the probability of alternation in government.
The final explanation concerns the quality of elections. The minimalist definition of democracy requires political leaders to be chosen in competitive, free and fair elections (Dahl, 1971). However, elections in Africa are the world’s worst in terms of electoral integrity (Norris & Grömping, 2019). Twenty years ago, Bratton (1998) noted that the quality of multiparty elections in Africa deteriorated over time: in the second elections held after the transition, presidents resorted to executive power to rig the rules against their challengers. He also added that:
In contrast to electoral volatility and incumbent turnovers in second elections in Eastern Europe and Latin America, dominant parties in Africa have usually been able to reinforce their supremacy the second time around. After a period of turbulent transitions, African politics is returning to an institutional legacy of “big man” rule, and the electoral alternation of leaders is again becoming abnormal. (Bratton, 1998, p. 65)
More recent appraisals of elections and democracy in Africa note that vote buying, intimidation and harassment have been used to skew the competition in favour of the incumbent parties (Bleck & Walle, 2019; Cheeseman & Klaas, 2018). Hence, the final hypothesis postulates:
Hypothesis 5 (quality of elections): The cleaner the elections, the greater the probability of alternation in government.
Methods and data
To test the hypotheses raised in this study, we draw on an original dataset of 179 parliamentary elections held in 27 countries from the founding multiparty elections until May 2019. As can be seen in annex A, most elections were conducted between 1990 and 2019 but the analysis also includes the cases of Botswana, Mauritius and The Gambia where multiparty elections have been held since the 1960s. The countries included in this study score at least partly free in the latest Freedom House report, they hold competitive elections regularly3, and have (minimum) legal provisions that guarantee political rights and civil liberties.
The dependent variables
Two types of analysis will be performed using democratic performance and alternation in government as dependent variables. Democratic performance is measured by the Varieties of Democracy Project’s Liberal democracy index which varies from low to high (0-1). This index takes into account political rights, civil liberties, checks and balances, rule of law, and the electoral component of democracy.
As can be observed in Figure 1, the countries included in this study vary substantively in terms of democratic performance. In the overall depiction, Cabo Verde, São Tomé and Príncipe, South Africa, Mauritius, Botswana, Senegal and Benin are exemplary cases of democratic progress and they have been systematically rated the best democracies in Africa - despite some internal variations. Cases such as Comoros, Guinea-Bissau, Mozambique, Madagascar or Lesotho have witness periods of both progress and, more recently, democratic retreat.
Authors’ elaboration based on V-Dem data (https://www.v-dem.net/)
Figure 1: Democratic performance (country, averages)
The second dependent variable - alternation in government - measures whether there was wholesale, partial or no alternation in power following the elections (Casal Bértoa & Enyedi, 2016; Mair, 1996). In the first scenario (wholesale alternation), the incumbent government leaves office and is totally replaced by a new party or coalition. In the second scenario (partial alternation), the new cabinet contains both new parties and old ones from the previous government. In the third scenario (no alternation), alternation is complete absent and the same party or parties remain in exclusive control of government (Mair, 1996). To compute the variable, we analysed all parliamentary/general elections and verified whether there was a transfer of power at the government level following the elections.
As exhibited in Figure 2, there are striking variations across countries: 87 out of 179 elections produced no alternation in power, while wholesale alternation occurred 50 times and partial alternation 42 times. Botswana, Mozambique, Namibia, South Africa and Tanzania never experienced alternation of power. The dominant parties, respectively the Botswana Democratic Party (BDP), Frente de Libertação de Moçambique (Frelimo), South-West African People’s Organisation (SWAPO), African National Congress (ANC) and the Chama Cha Mapinduzi (CCM), successfully managed to close political competition and deter new party entries. Cabo Verde, Ghana, Kenya, Madagascar, and Niger are among the countries where elections produce regular alternation in power. The two major parties in Cabo Verde - Partido Africano da Independência de Cabo Verde (PAICV) and Movimento para a Democracia (MPD) - have peacefully rotated in government since the founding democratic elections. Just a few months after its foundation in March 1990, the MPD unexpectedly won the founding multiparty elections in 1991 and repeated the win in 1995 and more recently in 2016. After losing parliamentary elections in 1991 and 1995, the PAICV won subsequent elections (2001, 2006, and 2011) with broad parliamentary support (more than 50% of the seats). Similarly, in Ghana the National Democratic Congress (NDC) and the New Patriotic Party (NPP) are the two largest parties in the system and the only ones to have legitimate expectations of forming one-party cabinets. The NDC won elections in 1992, 1996, 2008 and 2012, while the NPP took power in 2000, 2004, and 2016.
In contrast, several countries appear to have a more unpredictable electoral market. Benin has the most open party system, with partial alternation in government in all elections since 1995. The only exception is the founding multiparty election in which there was wholesale alternation. In São Tomé and Príncipe and Mauritius, most elections also resulted in the formation of cabinets combining both new and incumbent parties. Finally, elections in countries such as Lesotho and Burkina Faso led to partial, wholesale or no alternation in government.
The independent variables and controls
In the first study, where the dependent variable is democratic performance, there are two main independent variables: alternation in government in the founding elections, which is coded 1 if there was wholesale or partial alternation during the transition, and 0 otherwise; and alternation in government following the founding elections, which is also measured dichotomously (1= wholesale or partial alternation; 0= non alternation) but takes into account all elections held from the transition until 2019. These variables allow us to test the underlying assumptions of our two hypotheses, namely whether alternation in government in founding elections (hypothesis 1) and subsequent elections (hypothesis 2) impacts democratic performance over time.
These hypotheses are tested alongside a set of controls that are usually linked to democratisation. Ethnicity is often related to democratic performance, but the relationship between the two is far from straightforward. Reilly (2001), for instance, argued that highly fragmented societies can experience democratic progress if no group can control power alone. Ethnicity is measured with Alesina et al. (2003) Fractionalisation Data, which varies from 0 (homogeneity) to 1 (heterogeneity) and collapses the degree of linguistic, ethnic and religious fractionalisation within a country. GDP growth (annual %) is retrieved from the World Bank and accounts for the economic explanations for democratisation. The models also account for type of electoral system (1= plurality; 0= otherwise), form of govern (1= presidential; 0= otherwise), and time (measured as years since the first multiparty elections) for different reasons. Plurality formulas are known to increase the potential for conflict, and democratic breakdown given their winner takes all nature, while proportional representation favours proportionality and power sharing between competing forces (Norris, 2008). There is lack of consensus when it comes to forms of government: while some warn of the perils of presidentialism (van Cranenburgh, 2008; Walle, 2003), others show that presidentialism is not inherently bad and suggest that it bears negatively on democratisation when combined with high levels of party fragmentation and ethnic politicisation (Sanches, 2018b). Time since the first multiparty elections accounts for the more temporal dynamics of democratisation in Africa, and provides for an additional control of the democratisation by elections argument. Given the temporal and cross-sectional nature of the data and type of the dependent variable, we estimated a linear regression correlated panels corrected standard errors (PCSEs) in STATA with the command xtpcse.
The second study, for which alternation in government is the dependent variable, has three main independent variables as indicated in hypotheses 3, 4 and 5: alternation in government in the founding elections which is measured dichotomously as in the first study (1= wholesale or partial alternation; 0= non alternation); political competition which is calculated following Laakso & Taagepera (1979) Effective Number of Electoral Parties (ENEP) formula; and, finally, quality of elections which is measured by the Varieties of Democracy clean elections index4 (varies from 0= low to 1= high). These independent variables will enter the regression models alongside a set of controls, considered relevant to electoral competition in Africa.
As in the prior study, we control for ethnicity (as measured by Alesina et al., 2003) which significantly shapes party formation and competition in Africa (Ferree, 2010; Weghorst & Bernhard, 2014). Least squares index (LSq) is the second control, measuring the disproportionality between the distribution of votes and of seats; it is therefore of a finer measure of the effects of electoral systems on party systems. It is expected that higher disproportionality will favour the incumbent and major parties vis-à-vis opponents and smaller parties (Colomer, 2005). The robustness of civil society is measured by the Varieties of Democracy Core civil society index5 (varies from 0= low to 1= high), which allows us to control for the fact that the strength of civil society might influence the kind of alternatives that can emerge and put pressure on the incumbent government (Bratton & Walle, 1997). A set of variables - drawn from the World Bank - are also added to account for the macro-structural changes that can foster political change. These include GDP growth (annual %), urban population (as %) and unemployment (% of total labour force; modelled ILO estimate). Finally, a variable measuring number of elections was included to identify any significant changes over repeated elections. Once again, given the temporal and cross-sectional structure of the data and the binary nature of the dependent variable, we estimated logistic regressions in STATA with the command xtlogit. Basic descriptive statistics of the independent and control variables used in both studies are presented in annex B.
Main results
The first study tests the effects of alternation in government on democratisation, and the results of the models estimated are presented in Table 1. Model 1 includes the key independent variables - alternation in the founding elections, and alternation following the founding elections; Model 2 adds the controls.
The results confirm our theoretical expectations. First, if a country has experienced a power transfer during democratic transition, this strongly influences its capacity to continue to democratise over time. In other words, political alternations are unique formative moments when new democratic values and expectations are formed. Over time, countries that managed to democratise the most during the transition stage and experienced a peaceful transfer of power, experienced more positive democratic developments. However, and as most literature suggests, whether or not a country experiences alternation in future elections is not important to democratisation (Cheeseman, 2010; Wahman, 2014). As our data show, in addition to there being no significant effect of government turnover on democratisation, the association is negative. That is, opposition victories did not bring democratic gains and, ultimately, this means democracy is poorly rooted at the elite level, both in governing and opposition elites. In short, this first study confirms hypotheses 1 and 2. Though our expectations were confirmed it is important to note that the models have relatively low explanatory power (Model 1 R-square = 0.02; Model 2 R-square = 0.15) which suggest that other variables must be included in the analysis.
Table 1 The effect of alternation in government on democracy in Africa
Notes: significant at +p<0.10; *p<0.05; **p<0.01; ***p<0,001. Dependent variable is Quality of democracy measured by the Liberal democracy index which varies from low to high (0-1). Given the temporal and cross-sectional structure of the data and the type of dependent variable, the regression models were estimated with the command xtpcse in STATA. Multicolinearity is within acceptable standards (that is <10): in Model 1 the mean VIF = 2.36; and in Model 2 the mean VIF = 3.09 (estimated with the command vif, uncentered).
Turning to the sources of government alternation: Which factors matter the most? The second study provides answers to this question. Table 2 presents two models seeking to test our assumptions on the role of alternation in government in the founding multiparty elections, level of political competition and quality of elections on government turnover. Using a similar structure, we first estimated a model with the three main independent variables and a lagged dependent variable to account for the prior outcome for alternation in government (Model 3), and then added the controls (Model 4).
Taken together, the results corroborate hypotheses 3, 4, and 5. Hypothesis 3 posited that countries that experienced alternation in government in the founding elections would be more likely to experience alternation in successive rounds of elections, and this is confirmed in both models: the two variables are positively and significantly correlated, hence confirming the importance of transitional legacies and their structuring of competition in the future. Hypothesis 4 tested the assumption that close races had a significant impact on the likelihood of government alternation; this is confirmed by our results: a larger number of effective electoral parties has a significant and positive effect on alternation in government. Finally, hypothesis 5 predicted that clean elections favour alternation; and indeed, the results show a strong association between the two variables in both our models. This set of results reveals the significance of democratic credentials to alternation. In fact, both political competition and quality of elections are part of minimalist definitions of democracy.
Table 2 The sources of alternation in power in Africa
Notes: significant at +p<0.10; *p<0.05; **p<0.01; ***p<0,001. Dependent variable is Alternation in Government measured dichotomously (1= wholesale or partial alternation; 0= non-alternation). Given the temporal and cross-sectional structure of the data and the type of dependent variable, the regression models were estimated with the command xtlogit in STATA. Multicolinearity is within acceptable standards (that is <10): in Model 3 the mean VIF = 1.31; and in Model 4 the mean VIF = 1.63 (estimated with the command collin).
Overall, the two studies conducted herein reveal the importance of path dependence, more specifically, that of transition legacies to future political developments in Africa. They also show that government alternation and democratisation are strongly interrelated; especially at the early phase of democratisation, when new (democratic) values are adopted by the political actors, and start embodying the political game from the beginning of the process.
Conclusion
Although multiparty elections have become commonplace in Africa, it is also true that most countries have not seen democratic gains and that the same political class continues in power. So do more elections bring more democracy to Africa? And have they transformed the patterns of competition for government? The great variation found across African countries suggests that the answer is far from straightforward. In fact, it is necessary to revisit the arguments linking elections to democracy and turnover, and cover a larger time span and more countries than prior studies.
We relied on an original dataset of 179 elections in 27 countries to examine whether elections are vehicles for improved democratic performance and alternation in government; and performed two tests with democratic performance and government alternation as dependent variables. The first set of findings reveals that although government alternation in the founding elections positively impacts democratic performance, opposition victories in subsequent elections do not produce democratic gains. This finding underlines the role of democratic transitions as critical junctures when the seeds for the future flourishing of democracy are planted. However, the fact that alternation in government in subsequent elections has a limited impact on democratisation is revealing of the weak democratic role of opposition parties, which - like incumbent parties/leaders - fail to improve the democratic credentials of the regime once in power.
The second set of findings relate to why some countries experience more transfers of power than others. The results of our statistical models demonstrate that government alternation during transition, the level of political competition and the quality of elections significantly increase the odds of turnover over time in Africa; this implies that alternation in government results from a mix of structural and contextual factors.
Together, these findings further the existing literature in three complementary ways. First, they show that what happened during transition is still important and shapes democratic developments and competition for power in Africa today. While prior studies explored the impact of colonial heritage on democracy (Bernhard et al., 2004), or the role of authoritarian and transition legacies on party system development (LeBas, 2011; Riedl, 2014; Sanches, 2018a), to the best of our knowledge this is the first large-N study that assesses the impact of transitional outcomes - more specifically turnover in founding elections - for the prospects of democratisation and government alternation in Africa. Second, they highlight the notion that the fairness of elections matters for government alternation which, on a more negative note, is revealing of the uneven playing field in Africa and of the ways in which incumbency advantage, electoral malpractices and authoritarian practices continue to distort the character of African elections and nature of representation. Thirdly, the results are also telling of the fact that alternation and democratisation strengthen each other. On the one hand, alternation in government in the founding multiparty elections shapes the prospects of democratisation; on the other, democratic principles such as political competition and clean elections increase the likelihood of turnover over time.
Overall, this study shows that elections produce a diversity of outcomes in terms of both democratic developments and alternation in government. This variation is largely explained by historical legacies and continuous struggles over the rules of the game. It also raises issues on why political alternation after transition does not necessarily lead to further democratisation. This is an unexplored area deserving further study. An insight that emerges from this study is that the conditions that lead political elites to embrace democratic values in Africa matters for understanding democratisation. It would also be interesting to explore political culture and socialisation in future studies, by analysing the political trajectories and socialisation of the incoming elites that replace the incumbents when political alternation occurs.
