The economic determinants of electoral volatility in Africa

Introduction

An institutionalized party system is considered essential for democratic governance. Many of the world’s third-wave democracies exhibit high levels of electoral volatility, which indicates that they have low levels of party system institutionalization (PSI). Nonetheless, the research literature has not reached consensus concerning the trends in electoral volatility in different regions. For example, Tavits (2005) finds that electoral volatility declined over time in postcommunist European countries. Similarly, Lane and Ersson (2007) find that volatility levels in Eastern Europe are declining, but increasing in Western Europe. By contrast, Bernhard and Karakoc (2011) note that the passage of time does not predict volatility in postcommunist Europe. In their cross-regional study, Mainwaring and Zoco (2007: 172) find that those party systems that emerged after 1978 not only have much higher levels of volatility than the advanced industrialized democracies, but they do not appear to be stabilizing. These findings are consistent with those of Bogaards (2008: 118), who sees no evidence of a trend toward declining volatility in his study of African party systems. Lindberg (2007: 238) also concludes that Africa’s “fluid party systems” have not stabilized over time. ¹

Many puzzles also remain in the literature about the determinants of PSI and volatility. Based on the literature on Western Europe, one might expect that the party systems in these countries have started to move toward institutionalization. This expectation is consistent with the notion that electoral experience provides citizens and elites with an opportunity to learn and develop partisan commitments. As many have noted, in the electoral periods immediately after a transition to electoral democracy, one would assume that volatility would increase (Bernhard and Karakoc, 2011). Over time, citizens of the third-wave democracies, including those in Africa, might begin voting based on party labels or the past performance of parties and party platforms, resulting in more consistent voting patterns.

Although many influences have been tested in the literature, several gaps remain, and particularly in the context of sub-Saharan Africa (hereafter, Africa), a region that has seen dramatic change in the past 25 years. Studies of volatility in Africa have examined economic factors such as economic growth (Ferree, 2010; Weghorst and Bernhard, 2014), but there has been little effort to examine external economic influences. Countless studies have examined the influence of foreign aid on various aspects of the economy and polity, but no study has yet examined how aid influences electoral volatility. Africa receives more aid per capita than any other region of the world (Asongu, 2015), and foreign donors often contribute moderate levels to total government revenues in some states. In this context, sudden “shocks” in foreign aid—due to aid withdrawal—should produce a budgetary constraint, with attendant effects on patronage and voter support of incumbent parties. Since the late 1970s, African governments have also adopted countless economic reforms as part of International Monetary Fund (IMF) and World Bank stabilization and structural adjustment programs (SAPs). To the extent that SAPs obligate governments to introduce reductions in social spending, such adjustment measures might lead to short- to medium-term volatility as voters punish the governing incumbent party (or coalition) introducing SAPs. Certainly, SAPs have been associated with the emergence of mass protest in various African regimes, at least in the initial phases of adjustment (Bratton and van de Walle, 1997; Tuman, 1994). Protest was perhaps the only pathway for citizens to express their discontent about SAPs prior to the democratic reforms that ushered in regular multiparty elections. Voting against incumbent parties is now an option in Africa’s multiparty electoral regimes. Surprisingly, the prior quantitative literature on PSI and volatility in Africa has neglected to model these external economic influences.

This study contributes to the debate on the factors behind electoral volatility by examining the effects of two previously unexamined variables, foreign aid and structural adjustment, while controlling for a number of other common influences. Our data set includes data on African elections held between 1972 and 2010, and our results point to the effects of both foreign aid and structural adjustment. In contrast, the more conventional measure of economic performance, economic growth, does not appear to be systematically associated with electoral volatility. Thus, the results of this study clarify and provide a more refined understanding of the effects of economic influences on voting patterns in Africa. Contrary to some recent findings, we also find that African party systems are not stabilizing.

Understanding the sources of electoral volatility can contribute to a broadly comparative assessment of democratic performance. For example, high levels of electoral volatility have been associated with political and democratic instability (Mainwaring and Scully, 1995). Political party systems must exhibit some stability for parties to perform important democratic functions, such as recruiting leaders, aggregating interests, and promoting accountability (Lindberg, 2007: 215). Democratic accountability is difficult without an institutionalized party system (Rose and Mishler, 2010: 802). If parties disappear from election to election, voters cannot punish or reward them based on their performance. Moreover, high levels of volatility undermine programmatic representation as “party labels provide weaker programmatic cues” (Mainwaring and Zoco, 2007: 157) and the supply of public goods (Robbins, 2010). Further, “political outsiders” who do not value democratic norms and institutions, such as the late Hugo Chavez of Venezuela, are more likely to come to power under conditions of high electoral volatility (Mainwaring and Zoco, 2007: 158). Clearly, then, studying the forces that drive electoral volatility is a valuable endeavor.

Prior studies disaggregate electoral volatility into two types of volatility (e.g. Mainwaring et al., 2016; Powell and Tucker, 2014; Weghorst and Bernhard, 2014). Type A volatility— or “extra-system volatility” (Mainwaring et al., 2016)—is produced when votes or seats are won by new parties. Type B volatility—or “within-system volatility” (Mainwaring et al., 2016)—is produced by the shifts in vote or seat shares of existing parties. Disaggregating electoral volatility in this manner addresses the concern that the level of electoral support going to new parties is an important but neglected indicator of PSI (Bogaards, 2008). Indeed, the habitual emergence of new parties may be detrimental to party system development (Tavits, 2008). Moreover, other scholars find differences in the correlates of type A volatility and type B volatility (e.g. Mainwaring et al., 2016; Weghorst and Bernhard, 2014); thus, we believe it is fruitful to examine these types of volatility in separate models. Because most studies have focused on overall volatility, we also analyze the correlates of overall volatility for the purposes of comparison.

In what follows, we review the prior literature on economic performance and electoral volatility and present our theoretical framework. After this, we discuss our research design and the statistical results of the analysis. In the conclusion, we review some of the broader implications of our findings.

Literature and theory

Economic explanations of volatility have found empirical support in several studies. The economic voting literature hypothesizes that voters punish parties in office when the economy fares poorly and reward incumbents when the economy does well (e.g. Lewis-Beck, 1986). An important stream of this literature argues that a country’s economic performance does not influence voter behavior in a symmetric manner. More specifically, voters are more likely to punish officeholders for a poor economic performance than they are to reward them for a positive performance (e.g. see Bloom and Price, 1975; Clagget, 1986; Nannestad and Paldam, 1997). This tendency seems to be especially strong in developing countries (Pacek and Radcliff, 1995) and the countries of Eastern and Central Europe (Roberts, 2008), where voters appear exclusively to punish incumbents for poor economic performance. These studies suggest that poor economic performance increases electoral volatility.

The preponderance of studies on volatility supports the tenet that economic factors have an asymmetric effect on voter behavior. Roberts and Wibbels (1999) found that economic growth depressed legislative volatility in Latin America, a finding that has been replicated in Africa and other regions. For example, Weghorst and Bernhard (2014) find that economic growth has an inverse relationship with volatility, although this finding only applied to type B volatility in Africa. In their study of 67 democracies around the world, Mainwaring et al. (2016) find that economic performance has a significant negative effect on type A and total volatility, but not type B volatility. Mainwaring and Zoco (2007), however, find only modest support for such a relationship. Ferree (2010) notes that economic performance does not influence electoral volatility in Africa. Bernhard and Karakoc (2011) do not find support for the conventional indicators of economic voting in their study of postcommunist countries, but they do find that inequality is positively associated with volatility.

Based on studies showing that African citizens’ vote choices are influenced by economic considerations (e.g. Bratton et al., 2012), as well as the provision of collective goods (Weghorst and Lindberg, 2011, 2013), we expect economic performance to influence electoral volatility. Like Bernhard and Karakoc (2011), we move beyond economic growth to test the relationship between economic performance and volatility. Given the importance of aid in African countries, we expect foreign aid to influence electoral outcomes. Structural adjustment has also been a prominent force. Given its salience, we think structural adjustment also has effects on voter behavior and electoral volatility in the era of multiparty elections.

Donor aid and electoral volatility

How is foreign aid likely to influence electoral volatility? We expect aid to have a negative relationship with electoral volatility of all types. Aid should have a negative effect on type A volatility because it strengthens the ruling party and therefore discourages new parties from forming and contesting elections. As Tavits (2008: 118) notes, the literature theorizes that new parties are more likely to enter the political arena when a country has poor economic performance as potential leaders of new political parties see an opportunity to attract disappointed voters. New parties do appear to succeed in taking votes from the ruling party when the ruling party has not performed well on economic matters (Mainwaring et al., 2016; Tavits, 2008). Moreover, citizens are more likely to continue to support the ruling party if it is successful in securing donor aid. They are therefore also not likely to switch their allegiance to existing opposition parties.

Several mechanisms are likely to link foreign aid with stronger ruling parties. Aid provides incumbents with additional resources to implement the activities needed to stay in power. Most African countries are highly dependent on foreign aid for their public expenditures. Aid allows the ruling party to provide more goods, services, and patronage to please its base. Aid also increases the resources incumbent parties can devote to campaign activities including electoral handouts. Moreover, commitments and disbursements of aid send a signal to citizens that the incumbent government’s promises to distribute resources and promote development are credible. In what follows, we elaborate on these mechanisms.

Most of Africa’s political systems are neopatrimonial and thus grounded in clientelism supported by state resources (Bratton and van de Walle, 1997; Englebert and Dunn, 2013). To maintain power and stability, governments distribute patronage such as licenses, jobs, targeted collective and private goods, and other resources in a “pyramid of patron–client relations” which descends from the presidency to the village level (Englebert and Dunn, 2013: 133). Opposition parties in Africa are weak partly because of their inability to access state resources (Rakner and van de Walle, 2009). Foreign aid provides the government with more resources to dole out to clients in exchange for continued electoral support. Neither domestic nor international forces are likely to prevent the incumbent government from channeling aid resources to supporters. Merit-based hiring in the civil service was abandoned soon after independence in many African countries (van de Walle, 2001: 134), and the civil service continues to be highly politicized, despite decades of reform (Yanguas and Bukenya, 2016). Moreover, decision-making in most African countries is highly centralized. Several studies provide evidence that African governments can distribute aid based on political considerations even when this aid is scrutinized (e.g. see Briggs, 2012, 2015; Jablonski 2014). In Kenya, Briggs (2014) found that President Moi directed heavily supervised project aid to his constituencies. Similarly, Jablonski (2014) finds that incumbent governments in Kenya can direct aid to the voters who will facilitate their electoral victories. Further, politicians are elected with the expectation that they will usher in development with aid-financed projects (Englebert and Dunn, 2013: 230). We therefore also expect voters to view the government’s ability to procure aid positively for this reason.

The role of aid-supported patronage and development projects in stabilizing party support is illustrated by Briggs’ (2012) study of Ghana. Briggs (2012) provides evidence that the unexpected and significant drop in aid prior to Ghana’s 2000 election contributed to the decline in electoral strength of the ruling National Democratic Congress (NDC) party. With the budget shortfalls stemming from the reduction, the NDC’s provision of public goods was insufficient to satisfy its base. Briggs’ (2012) examination of the disbursement of World Bank funds for an electrification project before the 2000 election shows the effects. Comparing the NDC’s vote shares in rural constituencies between 1996 and 2000, Briggs (2012: 617) found that the NDC lost on average five more percentage points in the constituencies that did not receive World Bank–funded electrification. On the other hand, quantitative (see Briggs, 2015) and qualitative evidence suggests that foreign aid can help incumbents win elections. For example, Briggs (2015: 220) presents evidence that education aid was diverted through sham construction contracts and “into campaign spending and cash handouts,” which helped President Muluzi and his party succeed in their bid for reelection in Malawi in the 1999 election. In a similar vein, Brazys et al. (2015) find that Malawi’s president was able distribute the benefits of an aid-supported (Food and Agriculture Organization of the United Nations) fertilizer program in such a way as to boost support for his newly created party in the 2009 legislative elections. These findings are consistent with those of Dionne and Horowitz (2016) who conclude that the agricultural subsidy program in Malawi strengthened support for the incumbent party between 2008 and 2010.

Foreign aid also signals to voters that the ruling party can fulfill its promises to direct resources to them. How do candidates convince voters that they will keep commitments? The logic employed in Kramon’s (2016) study of vote buying is pertinent here. Although vote buying is prevalent across Africa, parties are largely unable to enforce the agreements made with voters. Kramon (2016) explains why vote buying is still effective in some contexts. Kramon (2016: 462) observes that nearly all political candidates pledge to give voters resources once elected. Electoral handouts are used in campaigns to demonstrate candidates’ credibility (Kramon, 2016: 456). As a more stable financial resource, foreign aid helps to establish incumbent parties’ credibility. Evidence from Bangladesh suggests that informing study respondents that a development project was funded with foreign aid raised respondents’ confidence in local governments (Dietrich et al., 2015).

Based on the prior literature, and the nature of clientelism in Africa, we expect foreign aid to have an asymmetric effect on voter behavior. Supporters of opposition parties may not be in the incumbent party’s patronage network. If a voter does not expect to benefit from government resources, she has no incentive to switch her allegiance to the ruling party when it appears to have ample resources. In contrast, we expect potential beneficiaries of government largess to withdraw their support for the ruling party when the incumbent party cannot make credible commitments to provide resources. Thus, voters will be more likely to punish incumbent parties that fail to attract aid than to reward parties that do.

Structural adjustment and electoral volatility

In stark contrast to foreign aid, SAPs have been extremely unpopular in African countries (van de Walle, 2001: 168), and we hypothesize that agreeing to SAPs will destabilize party competition. Although SAP reforms were usually not fully implemented (van de Walle, 2001), SAPs ushered in policies to stabilize the economy and then longer-term structural reforms to liberalize the economy and reduce the size and influence of the state. IMF adjustment often included a bundle of reforms to reduce the public workforce and wages, eliminate subsidies on food and other basic items, incorporate or raise fees for various social services, and promote privatization and trade liberalization (Onimode, 1992; Tuman, 1994). Even the IMF acknowledged that adjustment policies would induce short-run costs and economic pain.

Because political power in Africa largely rests on the control of state resources, we would expect any measures that limit a government’s control of these resources to undermine the support for the ruling party. As Greene (2010) finds, the control of state resources is at the root of why dominant parties endure. The incumbent party can channel the resources of the state into patronage and campaign activities and raise the costs of aligning with the opposition party (Greene, 2010: 812). Greene (2010) argues that economic reforms that wrest control of a country’s resources from the government, such as privatization, bring about the end of single-party dominance. We would expect volatility to rise with the decline of party dominance, as former supporters of the incumbent party necessarily vote for different parties.

SAPs also serve as a signal to voters. SAPs not only signal rollbacks in patronage and government spending but also government feebleness. Bratton and van de Walle (1997) found that the number of SAPs a country had adopted, even if not implemented, helped to explain African protests occurring between 1985 and 1994. They argue that adopting SAPs made the government look weak and unable to safeguard national sovereignty (p. 133). Similarly, in their study of the relationship between SAPs and contentious collective action in Latin America, Ortiz and Bejar (2013: 492) argue that people protest “…mainly because they perceive a loss of legitimacy and question the sovereignty of their domestic governments.” Bleck and van de Walle (2011: 1135) find that the issue of sovereignty is so salient that “all politicians” in their West African countries of study invoke it in their public communication. Therefore, we would expect opposition leaders to seize upon this issue in elections following the government’s adoption of a SAP. We expect African citizens to express their discontent with governments that adopt SAPs by abandoning the incumbent parties at the polls.

The case of Zambia is illustrative. The Patriotic Front (PF) won nearly 30% of the parliamentary seats and the PF presidential candidate, Michael Sata, won a similar share in the presidential election in Zambia’s 2006 elections. Sata and the PF attracted electoral support by tapping into broad populist frustration, particularly in Zambia’s urban areas, and the country’s “long history of urban political mobilization and suspicion of international financial institutions” (Cheeseman and Hinfelaar, 2009: 65). Sata argued that regular Zambians had not benefited from the various structural adjustment reforms the Movement for Multiparty Democracy (MMD) government undertook or from foreign investment and participation in the Zambian economy (Larmer and Fraser, 2007: 626).

To summarize, we expect that incumbent parties that adopt SAPs will be punished at election time, although the effects are asymmetrical (i.e. incumbent parties who resist SAPs will not be rewarded). Given their unpopularity, SAPs are likely to especially increase type A volatility. Potential leaders of new parties are likely to see the government’s adoption of SAPs as increasing the likelihood new parties could fare well. SAPs may also lead voters to be so disillusioned that they abandon the status quo and vote for the new parties that emerge to capture the votes no longer going to the incumbent party, which should be reflected in type A volatility.

Data and methods

We employ a data set that includes election data for 35 African countries that held multiparty legislative elections between 1972 and 2010 with country election as the unit of analysis. Due to regular gaps in the years between legislative elections, and variation in electoral calendars among African states included for analysis, the data set has unbalanced panels without common periods. As such, the data set may be described as a cross-sectional time series with unbalanced panels. Table A1 in the Online Appendix lists the countries in our data set and the number of observed elections.

The data set summarizes election data for all multiparty elections held since independence in those countries where elections have continued uninterrupted or since elections were reintroduced in the early 1990s. We employed a stricter definition of multiparty elections, resulting in fewer observations than Weghorst and Bernhard (2014). Unlike Weghorst and Bernhard (2014), we exclude multiparty elections held before formal independence and multiparty elections held after independence if they occurred before a lengthy authoritarian period. We did this to avoid having two distinct periods of electoral experience for most of our cases. Thus, for most of the 35 countries included in our data set, we include elections held since the early 1990s. ² The mean number of elections in the data set is 2.94 elections.

Measurement

Trends in PSI in Africa

We begin the analysis with a discussion of the trends in PSI in Africa. Based on our analysis of volatility, the party systems of Africa generally do not appear to be institutionalizing. Weghorst and Bernhard (2014) claim to find evidence of institutionalization based on the trajectories of type A volatility and type B volatility over time. ⁷ According to them, type A volatility crested in the 1990s and has since declined while type B volatility has increased progressively since the mid-1980s (p. 1728). As they note, “The players are establishing themselves and competition for voters, rather than the institutional failure of parties, explains why total volatility has remained relatively high in the period of competitive politics” (pp. 1728–1729).

Although our analysis of the data is consistent with Weghorst and Bernhard’s (2014) finding concerning the increase in type B volatility, we find that during the period in question, the data do not suggest that type A volatility was decreasing. The mean level of type A volatility was 2.62 during the decade of the 1970s, but increased to 8.59 and 11.68, respectively, in the decades of the 1980s and 1990s. The upward trajectory in mean levels of type A volatility continued after 2000. Between 2000 and 2010, the mean level of type A volatility was 13.20. Thus, even if one accepts the proposition that increases in type B volatility with a contemporaneous fall in type A volatility indicates institutionalization, the party systems of Africa do not appear to be institutionalizing.

Results of multivariate analysis

We begin our examination of the determinants of electoral volatility with a discussion of the economic influences in the model. The results suggest that structural adjustment and foreign aid have significant influence on volatility. Table 1 displays the results of our analysis.

OPEN IN VIEWER

Table 1.

Determinants of electoral volatility in Africa.

	(1)	(2)	(3)
	Total volatility	Type A	Type B
Structural adjust. dummy	9.592* (4.213)	7.223* (3.523)	2.369 (2.669)
Net ODA per capita (ln, t −1)	−8.612* (3.603)	−3.460 (3.063)	−5.152** (1.993)
Party fractionalization, lag	19.056* (9.726)	23.368* (9.444)	−4.312 (4.855)
Mean district magnitude	−0.104*** (0.028)	−0.026 (0.024)	−0.078*** (0.018)
Real GDP per capita, % change ( t −1)	−0.015 (0.163)	−0.020 (0.131)	0.004 (0.094)
Real GDP per capita (ln, t −1)	5.564+ (3.151)	−2.375 (2.456)	7.939*** (1.740)
Ethnic fractionalization	−2.563 (20.204)	−24.815 (20.035)	22.252*** (4.547)
Democracy score	1.850* (0.902)	0.510 (0.782)	1.340* (0.552)
% Rural	0.476** (0.157)	0.039 (0.112)	0.437*** (0.087)
British colony	−4.605 (3.661)	−2.943 (3.070)	−1.662 (2.925)
Years since independence	0.563*** (0.161)	0.188 (0.133)	0.374*** (0.086)
Constant	−81.022+ (41.909)	23.044 (33.275)	−104.067*** (20.798)
R 2	0.304	0.294	0.264
N	103	103	103

Note: ODA: official development assistance; GDP: gross domestic product. Top entries are unstandardized OLS regression coefficients; panel-corrected standard errors in parentheses.

⁺p < 0.10, *p < 0.05, **p < 0.01, ***p < 0.001.

Disaggregating type A volatility and type B volatility provides a better understanding of the nature of the relationship that these variables have with volatility. As one can see in model 1 in Table 1, the coefficient for structural adjustment (lagged) is positive and significant, indicating that electoral volatility is higher in countries during periods when governments have implemented SAPs. Likewise, the coefficient for structural adjustment is also positive and highly significant in the model for type A volatility (model 2). However, for type B volatility, although the coefficient for structural adjustment is positive, it fails to achieve significance. Taken together, the results suggest that countries with governments that enter a structural adjustment agreement experience higher volatility stemming from new parties drawing support from existing parties, but not from the changes in seat shares of existing parties. This finding supports our hypothesis that entering an agreement with the IMF is not well viewed by African citizens and ultimately increases the probability that citizens will withdraw their support from the governing party and direct it to a new party. We theorized that political elites likely perceive that the government’s acquiescence to structural adjustment will result in voter dissatisfaction that can be transformed into support for a new party. Thus, new parties enter the political fray. The government’s adoption of SAPs can then be used as the basis for the new party’s campaign appeals which are likely to be effective given the perceived implications of SAPs for sovereignty and government spending. Citizens are so dissatisfied with the status quo, they embrace new options. Our results are consistent with this scenario.

The results reported in Table 1 provide support for our hypothesis that foreign aid is associated with lower volatility. As one can see, the coefficient for foreign aid is negative and significant in both models 1 and 3, indicating that total volatility and type B volatility are lower in countries that have received more foreign aid per capita. Foreign aid is not significantly associated with type A volatility. Aid’s negative effect on type B volatility implies that aid influences the changes in the seat shares of existing parties. In fact, the aid variable functions very much like an economic performance variable. Weghorst and Bernhard (2014: 1725) find that economic growth influences, “volatility due to vote switching between parties—considered a hallmark of accountability in well-functioning electoral regimes…” and note that this is consistent with the economic voting literature. Our finding about aid’s mitigating effect on volatility is also consistent with the economic voting literature. It indicates that African voters are concerned with the government’s performance on a critical indicator of the government’s ability to promote economic development and provide resources. We theorized that incumbent parties successful in procuring aid would be able to maintain the loyalty of their supporters at election time. The results are consistent with this scenario.

Other economic influences produced mixed results in the models. The results for structural adjustment and aid point to economic voting in Africa, but the more traditional measure of economic performance—growth—does not fully support the hypothesized relationship among economic performance, economic voting, and volatility. The coefficient for economic growth is insignificant in all the models. This suggests that after controlling for aid, adjustment, and other influences, growth had no effect, on average, on volatility. However, the level of development, as measured by the coefficient for real GDP per capita, is positive and significant in the model for total volatility and the model for type B volatility. Thus, the level of development appears to exhibit some influence on some forms of volatility. One possible reason for this finding is that after controlling for other factors, voters in more advanced economies might feel more comfortable experimenting with different parties. Certainly, additional research needs to be undertaken to address this finding.

As for the control variables, the results suggest that institutional influences have some impact. The coefficients for party fractionalization are significant and positive in models 1 and 2, indicating that highly fragmented party systems have higher total and type A volatility in comparison with other types of party systems. However, the coefficient for party fragmentation was not significant for type B volatility (model 3). Although Mozaffar and Scarritt (2005) describe a puzzling situation in which low levels of fragmentation coexist with high levels of electoral volatility in Africa’s party systems, our results indicate that fragmentation does generally increase electoral volatility, particularly that emanating from new parties winning seats.

Our results also suggest that electoral systems affect volatility, although not always in the way one might suppose. We find a negative relationship between district magnitude and volatility. In the models for total (model 1) and type B (model 3) volatility, the coefficients for magnitude are negative and significant. Our results complement Weghorst and Bernhard’s (2014) findings for total volatility and magnitude, while demonstrating that magnitude predicts type B volatility as well. Tavits (2005: 292) explains that in emerging democracies with low levels of voter loyalty, less restrictive political party systems enable voters to choose among several existing parties such that they are not obliged to switch en masse to a completely new party (as might be the case in restrictive majoritarian systems). District magnitude is not related to either of the measures of party fragmentation, however, so the mechanism through which magnitude depresses volatility is not clear.

More democratic countries appear to have higher levels of total volatility and type B volatility. However, for type A volatility, the coefficient for democracy is positive but not significant. These findings are consistent with the contention that PSI and democracy are integrally linked and would suggest a direct relationship between the two phenomena.

The models provide some support for the influence of social and demographic factors on volatility. Ethnic diversity is associated with an increase in type B volatility as demonstrated by the positive and significant coefficient for ethnic diversity in model 3. This finding is consistent with research pointing to the instability of ethnic coalitions (Ferree, 2010; Weghorst and Bernhard, 2014). The size of the rural population is significantly and positively related to total volatility in model 1 and to type B volatility in model 3, indicating larger rural populations are associated with higher volatility. Weghorst and Bernhard (2014) find the rural population size is only related to type B volatility. The tendency for citizens to vote in blocks in rural areas (Bratton et al., 2005) may produce large shifts when there are realignments in party affiliation and thus may explain what initially seems like a surprising result.

Colonial experience does not exhibit any influence on volatility. The coefficient for British colonialism fails to reach significance in any of the models. Similar to Weghorst and Bernhard (2014), we find that years since independence has a significant, positive relationship with total volatility and type B volatility. This finding is counterintuitive as one would expect that longer experience with statehood would depress volatility. The idea that party systems institutionalize over time is not supported by this result.

Conclusion

This study has sought to clarify the sources of electoral volatility in African party systems by examining influences that have not been analyzed in the prior research literature on Africa, including structural adjustment and foreign aid. The findings of the study suggest that IMF SAPs increase total volatility, although the disaggregated models indicate that the effect of adjustment is primarily concentrated on type A volatility. As noted, the most likely reason for this is that governments that enter into a structural adjustment agreement experience significant declines in support while new competitors emerge to replace them. Likewise, we find that aid flows are associated with lower levels of total and type B volatility. After controlling for the influence of structural adjustment and foreign aid, the other economic performance variable, economic growth, did not affect volatility.

More broadly, the results of this study contribute to the debate concerning electoral volatility in other regions of the world. Electoral volatility has been found to influence the political stability and quality of governance in a country. Many of the insights about the causal factors behind volatility that can be derived from the experience of Western Europe, and the classical comparative politics literature, may not offer us significant analytical leverage when it comes to explaining electoral volatility in third-wave democracies in different regions of the world. Moreover, even studies that focus on one region produce different findings. The purpose of this study has been to provide a more thorough examination of the sources of electoral volatility in Africa from 1970 to 2010. The results of our analysis demonstrate that economic performance matters for electoral volatility in Africa’s multiparty regimes. In particular, foreign aid depresses both total volatility and type B volatility. A government’s ability to secure aid from international donors would appear to be well viewed by the public and thus enable governing parties to maintain voter support at the polls. Structural adjustment agreements, on the other hand, tend to stimulate total volatility and type A volatility. The most likely reason for this relationship is that adjustment typically obligates governing parties to accede to budget reductions that are unpopular in the short run. More conventional indicators of economic performance, such as GDP growth, however, do not exhibit the hypothesized effects on volatility, although the level of development has a positive effect in some trials. By clarifying the varying effects of different economic influences, this study builds on the prior literature by providing a more thorough understanding of volatility. These results highlight mechanisms through which international actors may influence African elections. Research on the political economy of SAPs and foreign aid underscores the domestic political implications of both practices. Our findings contribute to research on the political economy of aid, illustrating the impact of these economic practices on election outcomes. The hesitation of African leaders to enter into SAPs is comprehensible given the electoral costs they are likely to incur for doing so. On the other hand, procuring aid would appear to strengthen a government’s image in the eyes of the public and therefore bolster its electoral support. Donors and international financial institutions would do well to be cognizant of the unintended effects their policies may have on the political development of recipient countries. SAPs seem to undermine the long-term development of party systems. Aid may contribute to party system stability but the most likely explanation for this effect is that aid benefits the ruling party.

As discussed, we also find support for approaches that focus on institutional influences and social demography. The mechanisms underlying some of the relationships revealed in the analysis are not evident. Clearly, the puzzles surrounding the origins of electoral volatility have not been resolved and require more investigation.

Supplemental material