Coalitions and democratization by elections

The world has seen a rapid increase in the number of electoral authoritarian regimes since the end of the Cold War. These regimes combine formal democratic rules with authoritarian governance and they use elections as their primary source of legitimization (Schedler Reference Schedler2002b). As counter-intuitive as it might sound, authoritarianism without election is nowadays a rare combination (for an illustration of the spread of the electoral authoritarian regime type see Hadenius and Teorell Reference Hadenius and Teorell2007). Although it has become common sense not to equate elections with democracy,Footnote ³ a number of authors have asked themselves whether elections could act as a promoter of democratization, even though the elections themselves were not democratic.

Empirical research has made this claim plausible (Lindberg Reference Lindberg2006; Teorell and Hadenius Reference Teorell and Hadenius2009). Countries with a longer electoral history tend to be more democratically developed than those with a shorter experience of arranging elections. How the process of democratization by elections should be perceived is, however, still a largely unsettled issue. Two divergent interpretations of this concept have been used in democratization research. In Lindberg's original work (Reference Lindberg2006) on this subject, the idea was promoted that elections should not only be considered an indicator of democracy but also an independent variable that promoted democratic practices. In this way, holding elections would give rise to longer-term beneficial effects, such as a more active citizenry, a more developed civil society and more institutionalized political parties. Elections can also be moments when internal and external actors can increase pressure for reform. In other contributions to the field, democratization by elections is understood as a more radical process. According to this view, elections sometimes become impossible for the incumbent regime to control and hence lead to drastic regime changes. This way of looking at democratization by election became especially attractive after the ‘colour’ revolutions in Eastern Europe and Central Asia (see Bunce and Wolchik Reference Bunce and Wolchik2010; Hale Reference Hale2005). Because of these instances, it became increasingly relevant to ask why some of these elections resulted in electoral turnovers, while others were comfortably held under control by the incumbent regime.

A clear member of the latter tradition is the work by Howard and Roessler (Reference Howard and Roessler2006). When investigating potential determinants for ‘liberalizing electoral outcomes’, Howard and Roessler found opposition coalitions to be the most prominent explanation for such changes. This is, of course, a highly interesting finding, but it is one in need of more systematic testing. This study will make an effort to deepen our understanding of how opposition coalitions affect electoral outcomes.

In the work by Przeworski et al. (Reference Przeworski, Alvarez, Cheibub and Limongi2000), alternation is seen as the defining feature that separates democracy from autocracy. However, as suggested by other authors, it is possible to change regime without changing regime type (Levitsky and Way Reference Levitsky and Way2010; Schedler Reference Schedler2002b). Democratization is here defined, along the lines of classic Dahlsian polyarchy, as a process of increased political competition and participation along with more effective enforcement of political liberties and civil rights (Dahl Reference Dahl1971). From this theoretical perspective it becomes clear that democratization does not by default follow alternation in power.

In Howard and Roessler's article (Reference Howard and Roessler2006), most of the presented mechanisms, connecting opposition coalitions with democratization, are related to an increased chance of opposition victory or, at least, decreased support for the incumbent when the opposition is coordinated. It does undoubtedly make intuitive sense that opposition parties with pooled administrative and economic resources and strategically coordinated electoral behaviours would increase their electoral capacity. It is, however, unclear whether coordination would also change the risk of electoral manipulation and repression by the incumbent regime. Quite the contrary: there have been theoretically convincing arguments suggesting that manipulation would be more common in cases where the incumbent is threatened by a strong opposition (Molina and Lehoucq Reference Molina and Lehoucq1999).

The question remains whether the effect of opposition coalition building is really one of general democratization or merely a consequence of an increased probability of opposition electoral success. This distinction is important for several reasons. First, our understanding of the effects of coalition building is limited if we do not know why it tends to produce democratically beneficial outcomes. If the aim is to increase our understanding of elections under authoritarianism, we should not aggregate concepts more than necessary. Rather than discussing several possible mechanisms by which coalitions could create democratization, it might be possible to narrow it down to one question: whether coalitions increase the probability for turnovers.Footnote ⁴

Second, recent experience has shown that opposition success should not be equated with democratization. In a number of divergent cases, such as Madagascar and Georgia, it has become clear that newly elected governments might be reluctant to abolish incumbent electoral advantages and increase electoral fairness (Kalandadze and Orenstein Reference Kalandadze and Orenstein2009). Alternation in power is not by definition democratization, but it should be considered a possible independent variable to longer-term democratization. Also, democratization through elections could be created without turnovers. If incumbent regimes feel pressured, from inside or outside, they might give in to such pressure and increase the fairness of elections. Eventually, such gradual liberalization might lead the way for alternation. However, alternation in these cases might be an effect of democratization rather than the cause.

Third, the outcome of elections in authoritarian regimes and the coding of leading indices of democracy are highly correlated in election years. Freedom House, for instance, includes electoral turnovers and opposition electoral support as two of the criteria for determining its political rights score.Footnote ⁵ If coalition building were primarily a consequence of alternation, we would have to observe longer-term effects to really claim that coordination is important for democratization.

Accordingly, to understand the effect of opposition coalition building for democratization by elections, we need to understand how and why potential effects of coordination occur and whether these effects are short lived or survive past the election year. Three hypotheses are derived on the basis of the above discussion.

Is the relationship endogenous?

In a study discussing the creation of opposition coalitions, van de Walle (Reference Van de Walle2006) argued that the relationship between opposition coordination and electoral outcomes might be considered somewhat endogenous. He claimed that elections in authoritarian regimes could be described as a tipping game. When clear signs of regime instability appear, leading government politicians start to defect and join the opposition. These new opposition politicians do so because they realize that their chances of retaining power and access to state resources are higher if they contest the election on an opposition ticket. A prerequisite for opposition success is, however, that the opposition is able to create a credible alternative to the incumbent. When faced with realistic prospects of an electoral victory, former ethnic or ideological rivals would usually find it more attractive to create such an alternative.

This is a vital objection to the theory about opposition coordination and democratization by election. If this is indeed the case, what has previously been described as a positive relationship between these two variables could be partially endogenous, especially if it is shown that the effect of coalition building is intimately related to the cases where the opposition actually wins elections. Moreover, if the newly elected coalition governments are really nothing more than old wine in new bottles, this also gives an indication as to what kind of democratization process could be expected after the electoral turnover.

Although van de Walle's theoretical argument (Reference Van de Walle2006) is a critical objection to the theory of opposition coordination, it still needs to be tested empirically. It is impossible to factor the real perceived prospect for an opposition victory into a regression analysis. What could be done, however, is to take into account structural circumstances that are likely to affect the potential for incumbent defeat.

The most powerful predictor for incumbent defeat would probably be incumbent economic performance. Hence, we believe that opposition coordination would be especially common if the incumbent's economic performance was poor. The theory about economic voting is one of the most prominent explanations for incumbent electoral success in the Western context (see Duch and Stevenson Reference Duch and Stevenson2008). A study by Collier and Hoeffler (Reference Collier and Hoeffler2009) showed that economic performance also harms the prospects of incumbent re-election in authoritarian regimes. Bad economic performance sparks dissatisfaction among the electorate and increases the support for the opposition parties (see, for example, Geddes Reference Geddes1999; Teorell Reference Teorell2010). It also negatively affects the government's ability to distribute patronage (Greene Reference Greene2007).

What is a pre-electoral opposition coalition?

Pre-electoral opposition coalitions can be organized in different ways. In some cases, the coalitions are highly formalized and integrated. In such cases, the opposition decides to issue a joint list under one common name. The opposition in Kenya in 2002, for instance, used this strategy. In this election, all parties ran under one common National Rainbow Coalition (NARC) list and also nominated one joint presidential candidate. In other cases, the opposition parties have retained their original party labels and have run separate campaigns, but with a declared intention to form a coalition government if they obtained a parliamentary majority. The VMRO-DPMNE Party and the Democratic Alternative (DA), for example, used this strategy in the Macedonian 1998 parliamentary election. These are two different strategies, but in this article both are considered to be versions of a pre-electoral coalition. Whether to run as a joint or separate list is probably to a large extent determined by institutional arrangements. However, as argued by Golder (Reference Golder2006), pre-electoral coalitions are not only formed to circumvent disproportional voting systems, but is also used as a signalling mechanism to voters. In authoritarian regimes, both types of coalition (joint and separate lists) should signal a stronger opposition against the government and a clear indication that opposition parties do not intend to be co-opted by the incumbent after the election (Wahman Reference Wahman2011). To make sure that this decision does not affect the results substantially, robustness tests will also be run, excluding the 13 cases in the sample where the opposition ran on a separate list coalition.

It is hard to find any shortcuts when it comes to an adequate operationalization of the coalition variable. Van de Walle (Reference Van de Walle2006) used the share of the vote received by the leading opposition candidate in a presidential election as a ‘proxy’. He argued that a high share of the vote for one single opposition candidate implies a large degree of opposition cohesion. This operationalization does have serious drawbacks because it could just as well imply opposition strength. Accordingly, there would be a substantial correlation between this variable and opposition victory. Therefore, there is no other option than to code this variable manually (Howard and Roessler (Reference Howard and Roessler2006) adopt the same solution).

The opposition coalition variable in the analysis is dichotomous. An opposition coalition is considered present when all major opposition parties are included in one common pre-electoral coalition. That is, they have declared that they have the intention of governing together after a contingent electoral victory or they explicitly support a single presidential candidate. If important elements of the opposition are not included in the coalition, the variable is coded as 0. Consequently, cases such as Gambia in 2001 are not coded as a coalition. In this instance, three parties had united behind the main opposition competitor, Ousainou Darboe of the United Democratic Party. However, there were also three other opposition candidates running against Mr Darboe. The most successful of these alternative candidates was Hammat Bah of the National Reconciliation Party, who secured almost 8 per cent of the vote (BBC News 2001). In this case, opposition disunity clearly remained an obstacle for the opposition's challenge. Similarly, cases with multiple coalitions are not coded as one having a coalition. In these cases, the type of coordination that is hypothesized to have an impact on the electoral outcome was not present.

If the opposition is practically concentrated in one dominant party, this is also regarded as a coalition. An example of such a dominant party was the New Patriotic Party (NPP) in the Ghanaian elections of 1996 and 2000.Footnote ⁶ Information for the coding was taken from prominent secondary sources. Keesing's record of world events (various editions) and the Inter Parliamentary Union reports (various editions) were always consulted when available. If there were no explicit mention of coalitions in these sources, information was sought in Freedom House reports, scientific articles and articles from the press. A full list of elections and coding decisions is provided in Appendix 1.

Data and research design

The aim of this study is to keep the number of observations on as high a level as possible. Accordingly, there is no reason for restricting the population in time as long as there is accessible data. As will be explained later, the dependent variable of this study is dependent on Freedom House data; therefore it is not possible to go further back than 1973.Footnote ⁷ Because observations on many of the independent variables have decreased in the most recent years, the last observed year discussed is 2004.

Only authoritarian elections will be included in the population. There are different ways of identifying such regimes, but this article will use the classification from the work of Hadenius and Teorell (Reference Hadenius and Teorell2007) to identify this population. In their study, the authors utilized a category labelled ‘limited multiparty systems’ (LMS), which is equivalent to the concept of electoral authoritarianism. LMS codes all countries that, in a given year, have a government that has been elected in a popular and recurrent election where at least one real opposition candidate was allowed to participate, but where the democracy score is less than 7.5 on a 10-point (combined Freedom House and Polity) scale as an LMS state or, in this case, as ‘electoral authoritarian’.Footnote ⁸ This category is coherent with the definition of electoral authoritarianism used by Schedler (Reference Schedler2002b). Founding elections (that is, the country's first election under multipartyism) will also be excluded from the population. This, as Howard and Roessler (Reference Howard and Roessler2006) point out, is desirable because the outcome of these elections are seldom an effect of what happens in connection with the election but rather a consequence of power struggles that occurred in the process of setting up the electoral institutions.

The sample restriction outlined above leaves the study with a sample of 251 elections in 86 countries from 1973 to 2004. This is a significantly larger sample than all previous studies on this matter. The sample consists of 141 parliamentary and 110 presidential elections.