English (pdf)
Article in xml format
Article references
Automatic translation
Send this article by e-mail
Cited by SciELO 
Similars in
SciELO 
Permalink
Not all elections have the same importance: not for parties and not for voters. The second-order election model by Reif and Schmitt (1980) and its subsequent empirical tests remind us of this. In parliamentary systems, the first-order elections are the national legislative elections, and, in presidential systems, the presidential races. The stakes in these elections are high, and voters understand their importance, since they determine the distribution of power at the most important level of political decision-making. Second-order elections, in turn, have little effect on who controls executive power. Examples of second-order elections are local elections, European parliament elections and presidential elections in weak semi-presidential regimes.
The second-order election model has been extensively used in Europe but travelled little outside the European continent’s borders. Here we seek to fill this gap by conducting an analysis of the aggregate turnout and electoral behaviour patterns of Cape Verdean voters. It is now almost 30 years since the Cape Verde regime became a multi-party democracy, a process that was formalised by the 1991 legislative elections. These elections led to an immediate change in the incumbent and peaceful alternation in political power since then (Évora, 2001, 2009) - an important characteristic of consolidated democratic systems, which, along with other factors, causes Cape Verde to be portrayed often as a haven of democracy, an exception in the African panorama (Baker, 2006; Meyns, 2002). By applying a traditional model of aggregate electoral patterns to a recent democracy in Africa, we aim at contributing to the flourishing field of African electoral politics research by testing the explanatory power of a tool designed for consolidated democracies in their younger counterparts.
The Cape Verdean context seems ideal for testing the assumptions made by the second-order elections model outside the European context in which it was initially developed. First, Cape Verde is a recent democracy, and this often means that second-orderness assumptions are less clearly confirmed due to the fluidity of party systems and attachments (e.g. Schmitt, 2005). Second, Cape Verde is an archipelago, which may mean that local elections could matter more than in continental democracies, merely due to the geographical distance from the central national government experienced by several islands; also, research has shown than in African polities local elections are often very participated, which would deny the idea that citizens look at them as less important (Bleck & Walle, 2019). Therefore, to what extent are local elections in Cape Verde indeed second-order? Third, the fact Cape Verde’s political system is semi-presidential (and hence the president has more powers than the head of state in parliamentary regimes) leads us to question whether voters actually consider presidential elections to be second-order, especially in a region of the world in which both presidentialism and the president’s formal and informal powers are mainstream and paramount (Bleck & Walle, 2019; Walle, 2003).
We set out four hypotheses derives from Reif and Schmitt’s (1980) second-order elections model. Thus we expect different electoral behaviour across parliamentary and presidential and local elections in Cape Verde. To test our expectations, we examined aggregate data regarding every presidential, legislative and local election that took place in the country since 1991. Some of the data used in this analysis was found on the official webpage of Cape Verde’s National Election Commission (CNE - Comissão Nacional de Eleições), while the older data was obtained at the CNE headquarters. This data is analysed as a whole and within specific electoral cycles determined by the legislative election, which provides the most appropriate context to test second-order election model-related hypotheses (e.g. Freire, 2005).
This article is divided into six sections. The first is dedicated to the theoretical framework, namely the second-order election model assumptions and its test in new democracies. We then turn to a discussion of the presidential elections as second-order elections within semi-presidential regimes, with a special focus on the Cape Verdean context, and to what extent can local elections be defined as second-order elections within the African context. In the fourth section, we present our goals, the methodology adopted, and the hypotheses to be tested. We then test our hypotheses using the official aggregate data. The article concludes with a discussion of the main patterns identified through the empirical analysis.
The second-order election model in old and new democracies
In their seminal article on the European Parliament election, Reif and Schmitt (1980) called for the qualitative differentiation of election types and for a distinction between first- and second-order elections. At the heart of these two election types is the idea that voting behaviour and election results are determined by the perception voters have of what is at stake in the election.
According to Reif and Schmitt (1980), legislative (in parliamentary and semi-presidential systems) and presidential elections (in presidential regimes) are first-order contests, since their result determines how national executive power will be allocated, with important consequences in terms of decision-making process (Reif & Schmitt, 1980). First-order elections are marked mainly by an increase in conflict and salience (Schmitt, 2005), providing voters with an arena to make a critical decision about who should govern the country and what public policies should be adopted (Norris, 1997). Second-order elections, however, have no direct influence on who controls national executive power, and consequently are perceived as minor elections by both voters and political actors (Reif & Schmitt, 1980). Instead of factors directly having to do with the arena or function that the elected will cover, in a second-order election voting behaviour is determined by the voter’s assessment of the political parties based on their domestic policy, government popularity and economic performance (de Vries et al., 2011). This will have a two-fold impact: on the one hand, in terms of voter turnout and, on the other, voting choices (Hobolt & Wittrock, 2011; Norris, 1997; Reif & Schmitt, 1980; Eijk & Franklin, 1996).
Voter turnout tends to be lower in second-order elections simply because individuals are less likely to go to the polls if they believe what is at stake is of lesser importance (e.g. Freire & Magalhães, 2002). However, when they do vote, they tend to base the vote on their concerns about the national government, often treating these elections as an opportunity to express dissatisfaction with the party of government (de Vries et al., 2011). In fact, second-order elections tend to offer minor parties more chances of success, often denied them in the first-order elections, which tend to be dominated by mainstream parties (Hix & Marsh, 2007; Hobolt & Wittrok, 2011; Reif & Schmitt, 1980; Eijk & Franklin, 1996). In this sense, political actors involved in first-order elections can be punished directly through second-order elections (Marsh, 1998), especially when second-order elections are held in a moment of the electoral cycle in which the incumbent’s popularity is low (Fortes & Magalhães, 2005; Reif & Schmitt, 1980; Eijk, Franklin & Marsh, 1996).
Reif and Schmitt (1980) argue this behaviour has two causes: 1) the absence of strategic voting - second-order elections are a means by which voters can express their party preferences (Ferrara & Weishaupt, 2004); and 2) protest voting - voters use such elections to express their dissatisfaction with the performance of the government and as a way to punish the incumbent party (Eijk & Franklin, 1996).
While its original formulation was aimed first and foremost at explaining differences in voting behaviour between national and European Parliament elections, the second-order election model was later extended to other types of competitions: local elections (Anderson & Ward, 1996; Freire, 2005; Norris, 1997; Schmitt, 2005; Sotillos, 1997), mid-term elections in the United States (Kernell, 1977; Stimson, 1976; Eijk, Franklin & Marsh, 1996), parliamentary by-elections in the United Kingdom (Norris, 1997) and presidential elections in semi-presidential regimes (Fortes & Magalhães, 2005).
In 1997, Reif, Schmitt and Norris conducted a theoretical review of second-order elections and highlighted some shortcomings in the original 1980 article, such as the fact it was restricted to consolidated European democracies. Looking at the changes that have taken place over the past 40 years, it is about time to ask if the second-order election model remains accurate and valid when applied to new situations - such as in countries that have only recently become democracies. From 2004 on, research in this field of electoral politics has encompassed the Central and Eastern European (CEE) new democracies, which has brought some of the flaws of the second-order election model to light. Some studies suggest Reif and Schmitt’s original theory is less useful in terms of explaining and forecasting differences between elections in new democracies due to the feeble patterns of party identification and more fluid party competition trends that characterize these contexts. For instance, Koepke and Ringe’s 2006 analysis of four CEE democracies concluded that voters do not use second-order elections to engage in protest voting against their national governments, and that electoral losses of larger parties decreased as the first-order electoral cycle proceeded. Hix and Marsh (2007) observed a cleavage between the new and old democracies in respect of the electoral success of small and mainstream parties in the 2004 European elections. Despite incumbent parties largely losing electoral support in both old and new European democracies when compared to smaller parties, in the post-communist states there were unique features related to party competition, with high levels of volatility, creating a more complex panorama. Schmitt (2005) reached a similar conclusion in his comparison of the 2004 PE election results with the previous first-order election in each EU country: in the new Eastern democracies, voting losses do not follow the cyclical arrangement of second-order elections, in the sense that the small parties do not perform consistently better in EP elections compared to mainstream parties. According to Schmitt (2005) and Mainwaring and Torcal (2005), this is explained by the fact that post-communist democracies did not enjoy a stable, consolidated party system and thus experienced high levels of electoral volatility. Indeed, the party organisational and programmatic instability makes it more challenging for voters to identify the actors involved in party competition, assign accountability and assess the government’s performance. To a certain extent, these are also traits of African new democracies (Bleck & van de Walle, 2019), but perhaps to a lower extent in Cape Verde, where party system institutionalization levels are comparatively high (Sanches, 2018).
To our knowledge, the second-order election model was never implemented in an African consolidated new democracy such as Cape Verde. Considering that, in terms of postdemocratization electoral dynamics, Africa can be compared to the experience of former Soviet Union and Eastern European democracies (Bleck & van de Walle, 2019), an educated guess would state that the likelihood of fit between the second order model theoretical assumptions and the empirical patterns would be low. In the following two sections we lay out the grounds of the assumptions regarding to what extent may presidential and local elections in Cape Verde, despite its celebrated exceptionality vis-à-vis the African general panorama, be less viewed as second order.
Presidential elections in Cape Verde - first- or second-order?
As mentioned above, presidential elections in semi-presidential systems may be characterised as second-order elections (e.g. Fortes & Magalhães, 2005), albeit the use of this framework in this type of elections is less common in the literature. Resorting to the definition of semi-presidentialism, in which presidential power is limited, and the president does not govern (Elgie, 1999), Marsh and Franklin (1996) argue such elections are strong candidates to be described as second-order. Yet, a comparative study of semi-presidential systems by Fortes and Magalhães (2005) shows that the applicability of the second-order model to presidential elections in semi-presidential regimes is not straightforward and depends mainly on the powers the president has. The argument is that the more power they have, the more important the election will be to voters, making them more likely to be first-order elections; however, if the president’s power is limited, then voters will treat the election as less important and, therefore, as second-order (Fortes & Magalhães, 2005; Brug, Eijk & Marsh, 2001).
With this in mind, what can we say about Cape Verde? In fact, there is no consensus among researchers in respect of the nature of the system of government established by Cape Verde’s 1992 Constitution (Canas & Fonseca, 2007; Madeira, 2015): while some describe it as a form of rationalised or mitigated parliamentarism, others prefer to call it semi-presidential (Costa, 2009). In any case, the system in place distinguishes Cape Verde from several other African political systems, in which presidentialism is the most popular choice in terms of system of government (Bleck & van de Walle, 2019) - and in which presidential elections are clearly first-order.
The powers granted the President of Cape Verde by the 1992 Constitution do not allow the holder of that office to make big decisions independently; rather, the most important decisions can only be achieved with the consent of the political forces represented in the National Assembly and, in some cases, take into account the result of national election (Canas & Fonseca, 2007; Costa, 2009). It would therefore seem the powers granted the president of Cape Verde and the rather limited way in which they can be exercised would make Cape Verde’s presidential elections strong candidates to be perceived as second-order. However, it is worth noting that in a comparison of the power of popularly elected presidents in semi-presidential systems, the powers held by the president of Cape Verde are deemed greater that those exercised, for instance, by the Portuguese president since 1982, although they are not as extensive as the powers available to the presidents of France or São Tomé and Príncipe (during the 1980s) (Fortes & Magalhães, 2005).
An initial test of the second-order model in Cape Verde proved that voter turnout in presidential elections was consistent with second-order theory predictions (Fortes & Magalhães, 2005). This means the second order tag provided an accurate characterisation of presidential elections in this country, at least in terms of turnout. However, these conclusions are drawn from the analysis of a very small set of Cape Verde elections, and do not focus on the other assumptions of the model. It is now 15 years since the article by Fortes and Magalhães was published, and a considerable number of elections happened in Cape Verde since then, so it is important to ask if the elections that have taken place in this increasingly consolidated democracy since then also show signs of these patterns.
Local elections in Cape Verde: second-order?
The expectation that local election is, in terms of electoral behaviour, second-order has been empirically tested in a number of countries. For instance, studies on the United Kingdom, Netherlands and Belgium indicated those elections are “one-and-three-quarter order” (Heath et al., 1999), as they take a middle position in the first- and second-order elections dichotomy. This is so because voter turnout is generally lower than national elections but higher than in EP elections (Heath et al., 1999; Lefevere & van Aelst, 2014; Marien et al., 2015; Rallings & Thrasher, 2005). Moreover, empirical studies have shown that personal ties with local candidates are more important than national preferences (Fuentes & Villodres, 2010) and the election of the local government is less a matter of ideology and more about personality, experience and personal network (Marien et al., 2015). However, as voters do not judge the track record of that specific level of government, and the amount of information about local politics is scarce (Berry & Howell, 2007; Lefevere & van Aelst, 2014), the mechanism of electoral accountability is not always guaranteed at regional and local level.
What about local elections in Cape Verde? The broader literature about these elections in Africa would lead us to question their second-orderness in the eyes of the electorate. In terms of turnout, Bleck and van de Walle (2019) find that:
African citizens seem to turn out for local races in relatively high numbers, at least compared to other regions of the world […] these patterns diverge from those in elections in the United States and Europe, where turnout for presidential, legislative, and gubernatorial elections is consistently higher than that for local elections. (p. 226)
Several reasons may be behind this: local politicians may be more visibly responsible for delivering goods and services to their constituents, thus increasing the apparent importance of the election (and this may be even more the case in Cape Verde, due to the fact that it is composed of islands scattered in the Atlantic ocean); the lack of incumbency advantages one finds in presidential races may mean that electors might think they are more likely able to affect the election’s outcome, because incumbent candidates do not enjoy the same degree of incumbency advantage presidents or prime-ministers have; mobilization may also be more successful since in some contexts these elections are dubbed “elections of proximity” in which people mobilizes members from their local networks, which contrast with more elite-based legislative and presidential races (Bleck & van de Walle, 2019).
Also, it is possible that no behavioural change vis-à-vis the elections can be spotted because, as Bleck and van de Walle (2019) suggest, voters may feel that voting for the same party that controls higher levels of government will make coordination and resource delivery to the local level easier. This would mean no punishment of the incumbent party whatsoever.
Goals, hypotheses and methodology
This article analyses the existence of differences in turnout and electoral choice in different types of elections, in line with the second-order elections model defined by Reif and Schmitt (1980), in the Cape Verdean context. Cape Verde is an interesting case to study this because it shares the features of a consolidated democracy with those of a recent democratic system and the challenges of African electoral politics. Therefore, being Cape Verde a young but consolidated democracy in a somewhat different African landscape, differences in participation levels between first- and second-order elections can be understood as an additional example of democratic maturity on the part of Cape Verdean voters. Cape Verde is an example of semi-presidentialism regime in which presidential elections can be more or less seen as second-order elections; the same applies to local elections for the reasons debated above. In turn, legislative elections, through which deputies are elected to the National Assembly, are without doubt first-order elections, insofar as the results have a direct impact on the allocation of political positions at the national level and, consequently, within the governmental apparatus.
The time period covered by this research ranges from the first free elections in 1991 to the most recent elections, which took place in 2016. This article uses electoral data published in Cape Verde’s official bulletins. Part of the data was sourced from Cape Verde CNE’s official website, and the remainder gathered directly from the CNE offices. The elections analysed can be found in Table 1.
Resorting to Reif and Schmitt’s (1980) second-order election theory, we formulate four hypotheses which will guide our analysis of the aggregate data. These are designed in order to access the second-order nature of local and presidential elections in the country.
Hypothesis 1: Participation levels are expected to be lower in local and presidential elections than in parliamentary elections in Cape Verde.
Participation rates tend to be lower for second-order elections because voters are less inclined to take part since they believe there is less at stake. As such elections do not make a direct contribution to the distribution of executive power, voters tend to be less engaged with them compared to legislative elections, which directly contribute to the allocation of power, and thus will be less likely to turn out to vote (Freire, 2005; Freire & Magalhães, 2002; Reif & Schmitt, 1980).
Hypothesis 2: Smaller parties in Cape Verde are more likely to perform better in local and presidential elections than in parliamentary elections.
There is a consensus among researchers that second-order elections tend to benefit smaller parties, while first-order elections tend to favour the larger parties, also because the latter may receive votes from supporters of the former, in a process of strategic or tactical vote (Eijk & Franklin, 1996; Eijk et al., 1996; Norris, 1997; Hobolt & Wittrok, 2011). In the case of presidential elections, which are not partisan stricto sensu, we will analyse the results of the candidates backed up by specific parties.
Hypothesis 3: There will be more blank and spoiled papers in local and presidential elections in Cape Verde than there will be in parliamentary elections.
Voters may express their discontent with the parties and/or candidates by spoiling their papers in second-order elections or expressing no preferences in the ballot. As a result, there may well be an increase in the number of spoiled and blank papers in second-order elections (Reif & Schmitt, 1980).
Hypothesis 4: The national incumbent party will be penalised in local and presidential elections by receiving fewer votes than it did when it was elected.
Since second-order elections are not considered crucial elections for determining the distribution of power, especially in the period between first-order elections, discontent voters will be more willing to use these elections to send the governing party a message. Voters will often use second-order elections to punish the incumbent party (Freire, 2005; Reif & Schmitt, 1980; Eijk & Franklin, 1996; Eijk et al., 1996).
Results
Elections in Cape Verde (1991-2016): Overview
Table 2 shows the results of all legislative elections in Cape Verde between 1991 and 2016. During this period there were six legislative elections, with only the two largest parties, the Movement for Democracy (MPD - Movimento para a Democracia) and the African Party for the Independence of Cape Verde (PAICV - Partido Africano da Independência de Cabo Verde), being able to govern. Since the 1991 multiparty elections, these two parties have exerted executive power for relatively equal periods: the PAICV governed from 2001 to 2016 and the MPD from 1991 to 2001 and then since 2016. Of the smaller political parties that have stood for parliament, up until now only three parties had elected deputies to the National Assembly: the Democratic Convergence Party (PCD - Partido da Convergência Democrática) in 1995, the Democratic Alliance for Change (ADM - Aliança Democrática para a Mudança) in 2001 and the Democratic and Independent Cape Verdean Union (UCID - União Cabo-verdiana Independente e Democrata) in 2011. None of these other parties have managed to elect a deputy to the National Assembly: Social Democratic Party (PSD - Partido Social Democrático), Democratic Renewal Party (PRD - Partido da Renovação Democrática), Labour and Solidarity Party (PTS - Partido do Trabalho e da Solidariedade) and the Popular Party (PP - Partido Popular). Voter turnout has declined over the past decade, particularly compared to the 1990s; however, the abstention rate was lower in the last two legislative elections. For details on the background and grounds for party competition in these elections, see Sanches (2018).
Table 2 Legislative elections in Cape Verde (1991-2016)
Source: CNE. All values were rounded to one decimal place
In turn, Table 3 displays the results of all six presidential elections held in Cape Verde between 1991 and 2016. Only candidates supported by the two main parties, the MPD and PAICV, have been elected president. Pedro Pires, the PAICV candidate, won in 2001 and 2006, while the two MPD’s candidates won twice each: in 1991 and 1996 (António Mascarenhas Monteiro) and 2011 and 2016 (Jorge Carlos Fonseca). Independent candidates stood for the presidency in 2001 (Jorge Carlos Fonseca and David Hopffer Almada), 2011 (Aristides Lima, in dissidence with PAICV, and Joaquim Monteiro) and 2016 (Albertino Graça and Joaquim Monteiro), with none receiving enough votes to participate in a second round. After a first presidential race, still in the transition period (1991), marked by high turnout (this election was considered of paramount importance due to the context and the charisma of the competing leaders; Sanches, 2018), turnout rates for presidential elections have slightly declined over the years, with considerably low values in 1996 (a non-competitive election) and 2016. Interestingly enough, the time between presidential and legislative elections in each electoral cycle is usually very short, varying between one and seven months.
Table 3 Presidential elections in Cape Verde (1991-2016)
Source: CNE. All values were rounded to one decimal place
Lastly, Table 4 presents the results of all seven local elections that have taken place in Cape Verde between 1991 and 2016. As with the other elections, there is a tendency for the large parties to dominate, with the MPD coming out on top in 1991, 1996, 2004, 2008, 2012 and 2016, and the PAICV in 2000. This trend was spotlighted by Janira Hopffer Almada, PAICV leader, last April when she set up as a goal to break the MPD’s hegemony1 on local elections by winning eight municipalities in 20202. A number of small parties and independent groups have run in these elections since the early 1990s, but with some exceptions their results are often modest.
Table 4 Local elections in Cape Verde (1991-2016)
Source: CNE. All values were rounded to one decimal place
A total of 15 small parties and independent groups took part in the first local elections in 1991, with all but one obtaining less than 5% of the vote: the exception was the Movement for the Renaissance of St. Vincent (MPRSV - Movimento para o Renascimento de São Vicente) with 6%. Just nine small parties and independent groups took part in the 1996 local elections, with the MPRSV gaining 10.5% and beating the PAICV, which obtained 7.4%. The 2000 local elections were contested by 13 small parties and independent groups, with the PCD and the Action for Work and Solidarity (ATS - Acão para o Trabalho e para a Solidariedade) winning 6.6% and 7%, respectively. In 2004, none of the nine small parties and independent groups contesting the local elections obtained more than 5%, and in 2008 local elections, none of the smaller competitors took more than 4%. Again, while a total of nine small parties and independent groups stood in the 2012 local elections, their results were below the 5% threshold. Finally, of the nine smaller parties and groups that contested the 2016 local elections, only UCID won more than 5% of the vote (5.1%).
It is worth noting that the time between these elections and the previous legislative elections varied between one month and three years, but that most local elections are not held close to the legislative election date (the January 1996 election, held one month after the legislative contest, is a clear outlier in this set of local elections). Turnout at local elections generally varies between 55% and 60%, but it was higher in 2008 and 2012.
Hypotheses testing
Our four hypotheses are tested in two phases. First, we look at the democratic period in Cape Verde as a whole, comparing average differences between the patterns observed in presidential and local elections and the preceding first-order elections; then we test them through an analysis of the electoral cycles, composed of legislative elections and the local and presidential elections that followed. The first electoral cycle is that of 1991, the year in which the country held its first multiparty elections. The second cycle includes the years 1995, 1996 and 2000, with legislative and presidential elections in the first two years and local elections in the second and the latter. The third electoral cycle ran from 2001 to 2004, with presidential and legislative elections in 2001 and local elections in 2004. The fourth electoral cycle corresponds to the years 2006 to 2008, with presidential and legislative elections in the former and local elections in the latter. The fifth cycle took place from 2011 to 2012, with presidential and legislative elections in the former and local elections in the second. Finally, the sixth electoral cycle took place in 2016, a year in which all three types of election were held. While the overall analysis identifies general patterns, the election cycle analysis allows us to test our hypotheses with more precision and robustness, since second-orderness effects in terms of participation and vote choice are expected to be linked with the specificities of legislative election cycles (Reif & Schmitt, 1980).
Overall analysis of the period 1991-2016
We begin by analysing the global average differences between turnout rates in local and presidential elections and the legislative election that preceded them (Table 5). Taken as a whole, presidential and local elections display negative figures, which suggest that more people voted in the previous first-order elections (the overall average difference is of almost 10 percent points). Despite the differences being much greater in the case of presidential elections (-14.4) than local elections (-3.6), these patterns confirm our first hypothesis.
The next two columns of Table 5 present the data used to test the second hypothesis (small parties obtaining better results in elections other than the parliamentary competitions). We see an overall average difference of 5.5 percentage points, with both the presidential and local elections showing positive values of 2.4% and 7.6%, respectively. This confirms the second hypothesis. It is nevertheless interesting to see that, if we look at the overall average difference in the electoral result of large parties between the two election types, while we observed an overall difference that is congruent with our expectations (negative: -5.4), the average difference for presidential elections is actually positive (5.1%), while local elections are those in which a negative difference can be spotted (-17.5%).
Looking at the local and presidential elections as a whole, it is not possible to spot more blank and spoiled papers than in the previous legislative election (our third hypothesis), since the overall average difference was of 0.3 percentage points. This is because the average difference for the presidential elections is negative (-1.1%), while for the local elections the figure is positive (1.9%). This means the hypothesis is only confirmed for local elections. Interestingly enough, in the case of the local elections, the difference is positive for every election, while in the case of presidential elections it is always negative.
Finally, looking at the hypothesis stemming from the argument that presidential and local elections may be used to punish the party of government, the data points to an overall average of -3.1 percentage points difference in the incumbent’s results (Table 5). The average for presidential elections is positive (4.0%), while for local elections it is negative (-11.2). This partially confirms the fourth hypothesis, only for local elections.
Analysis of electoral cycles
Looking at the election results in each cycle (Tables 5 and 6), we see that the first hypothesis (lower voter turnout in presidential and local elections in Cape Verde compared to the turnout in first-order elections) is confirmed in most electoral cycles (1991; 1995/1996/2000; 2011/2012; and 2016). However, this was not the case in the 2001/2004 electoral cycle, when both local and presidential elections had a higher turnout than the first-order elections. This, however, may be due to the specific context of the 2001 legislative election, when alternation was expected after the MPD faced change in leadership, corruption accusations, internal turmoil and dissidence (Meyns, 2002; Sanches, 2018), which may cause their voters to demobilize. However, the hypothesis is confirmed for the first round of the presidential elections in this 2001/2004 cycle, which had a lower voter turnout than the first-order elections. Considering that the two candidates had virtually the same result in the first round, it is not surprising that voters felt more mobilized in the second round - due to a very intense campaign (Meyns, 2002). Also, the hypothesis is confirmed for the 2006/2008 electoral cycle only when comparing the legislative and presidential elections. Again, legislative elections were poorly participated, very much due to contextual reasons that would be overcome in the 2011 race (Sanches, 2018).
Therefore, the fact that the hypothesis about turnout is not confirmed in the second round of the presidential elections and the local elections of the 2001/2004 electoral cycle, and its partial confirmation in the 2006/2008 electoral cycle, may be due to the climate of public opinion. While in some cases government popularity could be the cause of increased voter turnout in second-order elections, this pattern may also be explained by voter “discontent” with the positions adopted by the party of government, with the increased participation being accompanied by a decline in support for the incumbent party or its favoured candidate in second-order elections when compared to the results obtained in first-order elections. This argument gains weight especially when we realise that in all the non-parliamentary elections that took place during these two electoral cycles, with the exception of the 2001 presidential elections, and even when there was an increase in voter turnout rates, the incumbent party lost support (in terms of vote share) compared to the results it had received in the preceding first-order elections.
As for the second hypothesis (greater probability of smaller parties in Cape Verde obtaining better results in local and presidential than in first-order elections), our analysis highlighted three different situations. First, the hypothesis is partially confirmed in the 1995/1996/2000, 2001/2004 and 2006/2008 electoral cycles, only when we compare the sum of the results achieved by small parties in local elections with their results in legislative elections. The hypothesis was confirmed, however, for all elections in the 2011/2012 and 2016 electoral cycles. Lastly, the hypothesis could not be tested at all in the 1991 election cycle (because there were not small parties in the 1991 legislative race) and for presidential elections in the 1995/1996/2000 (because there was only one candidate, supported by the MPD) and 2006/2008 electoral cycles (due to the absence of candidates who were either independent or supported by small parties).
The third hypothesis (that there would be more blank and spoiled papers in presidential and local elections in Cape Verde than in first-order elections) is confirmed in all electoral cycles, but only for local elections. The hypothesis is indeed not confirmed when we compare the percentage of spoiled papers in legislative elections with those in presidential elections in each election cycle; in fact, there were actually fewer spoiled papers in every presidential election than in any of the legislative elections. One of the explanations for this hypothesis not be confirmed could be the fact that presidential elections are more personalized, so it is possible voters feel it is easier to express their views.
Lastly, the fourth hypothesis (incumbent party punished in local and presidential elections) is partially confirmed for local elections in the electoral cycle of 1991, the year in which the first multi-party elections were held, as well as in the 1995/1996/2000, 2001/2004 and 2006/2008 electoral cycles, when the incumbent party suffered a substantial decline in vote share relative to the elections it had won. Yet, we should mention that it was not possible to carry out a valid test of the 1995/1996/2000 cycle in what regards the presidential election, because there was only one candidate, António Mascarenhas Monteiro, supported by the MPD, the incumbent party at the time3.
The hypothesis is ultimately confirmed in the 2011/2012 electoral cycle, when the incumbent party received a lower vote share than it had secured in legislative elections in both the local and presidential elections. The government-supported candidate in the 2011 presidential elections lost, with a slight decline in their percentage share of the vote - however, this was much due to the fact that another PAICV personality decided to run as presidential candidate (Sanches, 2018). Moreover, not only there was a large fall in the share of votes won by the incumbent party in the 2012 local elections, but this meant that it lost those elections to the opposition parties. In turn, this hypothesis was rejected in the 2016 electoral cycle, the last year in which elections were held, and in which the party of government displayed a better electoral performance in both second-order contests. The non-confirmation of the fourth hypothesis in the 2016 electoral cycle may be due to the fact the MPD (the winner of the first of these three elections) had been the main opposition party for three mandates (15 years). In this sense, Cape Verde voters may have seen this election an opportunity to punish the PAICV, which had governed the country for more than a decade, more than the newly elected party with whom they were still honeymooning.
Conclusions
This article sought to apply Reif and Schmitt’s (1980) second-order election model to Cape Verde by assessing if there are any differences in voter turnout and electoral choice in legislative, presidential and local elections consistent with the model proposed. To accomplish this goal, we examined all the elections that took place between the first multi-party elections in 1991 and the most recent elections held in 2016.
The empirical patterns observed allow us to highlight some points that help understand the dynamics of participation and voting behaviour in local and presidential elections in Cape Verde. First, the results suggest without doubt that Cape Verde voters view local elections as second-order: lower participation, more blank and spoiled papers, better results achieved by smaller parties and the party of government being punished. However, in the case of presidential elections, their second-order nature is less clear and not always verified: if, on the one hand, turnout is almost always lower than that of the previous legislative race, as expected, the other trends are either completely absent or present in but a few presidential races.
There could be two reasons for this. The most likely explanation is the fact that legislative and presidential elections have almost always been held within a few months of each other. Presidential elections therefore happen in the honeymoon phase of the first-order electoral cycle. This may not lead to a case in which a considerable amount of voters are discontent and so seek to penalise the party of government in subsequent presidential elections, either by voting for the main opposition party, spoiling their ballots or voting for smaller parties (Reif & Schmitt, 1980). Indeed, both Costa (2009) and Sanches (2010) mention a sort of contagion effect between the legislatives and the presidential elections in Cape Verde due to the timing, with the candidates supported by the winner of the legislative race having good results in the presidential election. The fact that presidential elections are scheduled to occur soon after the legislative elections and do not produce any changes in terms of who governs would actually be a distinct sign of these elections’ second-orderness. Nevertheless, despite Cape Verde being a semi-presidential regime, the president does have a number of powers (Costa, 2009; Fortes & Magalhães, 2005; Madeira, 2015), which make it possible to argue that perhaps the Cape Verdean voters do not regard, or at least do not always regard, presidential elections as being actually second-order, and therefore do not vote in line with Reif and Schmitt’s (1980) theoretical expectations. This would actually put Cape Verde in line with a clear trend in terms of electoral politics in African democracies: that of a considerable weight of presidents and presidential elections (e.g. van de Walle, 2003). Subsequent research, perhaps benefiting from more variation in the distance between the legislative and presidential election dates in Cape Verde, may shed light on this interesting puzzle.
