5. Metodología
5.3 La muestra
In this chapter we sought to explain the sources of PSI variance across countries and time, testing six major explanations: social structure (H1), institutionalization over time (H2 and H3), institutional design (H4 and H5), party and party systems characteristics (H6 and H7), economic performance (H8) and electoral participation (H9). The results of the linear regressions with PCSEs (which explain between 46% and 56% of the variance found in the sample) provide some support for the theoretical explanations based on social structure, endurance of the polity and institutional design. In fact, countries with more durable polities, with the lowest levels of party system fragmentation, with extensive regulations for party funding/finance and with higher levels of ethnicity are associated to higher levels of PSI. The results also indicate, however, that the religious divide is the most prominent feature within ethnicity affecting the levels of PSI; with higher levels of religious fractionalization decreasing the level of PSI. Therefore, religion emerges as the cleavage with a larger potential of social differentiation and conflict in the sample. This makes sense if we consider that
137
religion is a singular social cleavage insofar as it is associated with differences among civilizations (hence implying different ways of understanding the world and social relationships) and it requires exclusivity (Montalvo and Reynal-Querol 2002).
Models 4 and 5 clearly show that, while the linguistic and the ethnic divides have a positive, even if not statistically significant, effect on PSI the religious divide has a negative and strong effect on PSI. Recent studies, which have explored the relationship between ethnicity and economic development in Africa, have also arrived to similar conclusions. Montalvo and Reynal-Querol (2002)75 found that religious polarization was better at explaining slow economic growth than ethnolinguistic fragmentation. Moreover, when both indexes entered the model together, ethnolinguistic fragmentation became insignificant, while the religious polarization remained a statistically significant predictor of poor economic performance. More recently, Kodila-Tedika and Agbor (2013) partly corroborated this result by finding out that religion had important effects on the development process through its effects on economic investment.
The fact that the religious divide emerges as crucial in our analysis, but also in the studies here quoted is particularly relevant since African societies are mostly known for their ethnic diversity, while religion, class and language tend to be more salient in the Western world (Horowitz 1985). Its relevance in our analysis opens this debate into considering the particular context within which these social cleavages are activated as well as their momentary and strategic mobilization by political actors. Macro-level changes also play a role here. Using time-series data from the World Christian Data base Kodila-Tedika and Agbor (2013, 2) insightfully note that Sub-Saharan Africa’s religious landscape «has undergone profound changes from a monolithic African traditional religious society to an increasingly polarized religious society». More precisely, since the 1950s Christian and Muslim population have shown a dramatic increase from about 50% to 85% (combined) while traditional religions lost their followers (down from about 60% to 13% within the same period). This is a significant change that can also account for the saliency of the religious divide in relation to the others, in contemporary African societies.
Overall, the models displayed in Tables 4.4 and 4.5 reveal that, ENEP, party funding/finance and polity durability are strong predictors of PSI as they exhibit consistently strong correlations with PSI, even when we control for the quality of the model (fitting a lagged dependent variable) or for alternative explanations, such as the type of electoral
138
system, president turnover and quality of institutionalization or when we run additional estimations with individual indexes for ethnic, linguistic and religious fractionalization (Table 4.5). In this sense H1, H3, H4 and H6 are confirmed.
In the group of the control variables we find that the type of electoral system is also relevant in explaining variance in PSI, with majoritarian formulas being more correlated to lower levels of institutionalization than proportional representation and mixed formulas. This is a general finding for the total of 83 elections*countries observed in this study, but as usual there are outliers. For instance, in the case of Botswana the finding does not apply, however if we consider elections held in Zambia, Lesotho (until 2002) and Malawi it does make sense. This result relevance is two-fold. Firstly, it indicates that the type of electoral system matters for PSI in Africa. Secondly, it encloses an interesting puzzle since majoritarian formulas are usually associated with higher levels of stability, at least at the government level76. Therefore, it will be interesting to explore the effects of electoral institutions in our two case studies since Zambia’s first-past-the-post elections create higher levels of fragmentation and weaker levels of institutionalization than Mozambique’s proportional representation system. This analysis will take place in Chapter VI where we additionally show that the structure of the party system and the salience and/or activation of territorial cleavages (ethnic, religious, linguistic and regional) are important to understand how the mechanical effects of the electoral institutions (proportionality and fragmentation) actually work; but before that, the next chapter presents the framework for the comparison of the case studies.
76 As Norris (1997) clearly puts it: «The aim of plurality systems is to create a “manufactured majority”, that is, to exaggerate the share of seats for the leading party in order to produce an effective working parliamentary majority for the government, while simultaneously penalizing minor parties, especially those whose support is spatially dispersed. In “winner take all”, the leading party boosts its legislative base, while the trailing parties get meager rewards. The focus is effective governance, not representation of all minority views.
139
CHAPTER V – THE FRAMEWORK FOR THE ANALYSIS OF PARTY SYSTEM