Nigeria's 2015 election was the most closely contested since its 1999 return to democracy. Shortly after polls revealed that the top two candidates were nearly tied (Nengak, Mbaegbu, and Lewis 2015), the electoral commissioner delayed the election by six weeks. Declining global oil prices had hit the economy, allegations of corruption plagued the incumbent administration, and terrorist violence in the northeast raised doubts about holding elections there at all. Yet the defeat of the ruling People's Democratic Party (PDP) for the presidency, control of the bicameral National Assembly and a majority of the governorships still came as a surprise. The opposition All Progressives Congress (APC) was less than two years old, cobbled together from defecting PDP politicians and regional parties in a country with a long history of failed coalitions (Obafemi et al. 2014). What explains the PDP's stunning defeat?
Prevailing explanations for electoral outcomes in Nigeria emphasise patronage, electoral fraud and, more recently, observed rates of violence. Ethno-regional politics also feature prominently: ‘in advanced economies, cleavages are largely along social class lines … but in Nigeria cleavages are often along primordial lines, such as ethnicity, gender, and religion’ (Okafor 2013, 2). Drawing on field research before and during the 2015 elections and an original dataset, our cross-sectional statistical analysis of 36 states (plus the Federal Capital Territory) offers compelling quantitative evidence of ‘economic voting’. Specifically, weak economic conditions in the states (measured in terms of debt levels and internal revenue), negative views of the national economy and subjective reports of poor economic conditions at the local government level all systematically correlate with stronger electoral support for the opposition.
Our findings suggest that Nigerians, like a growing number of Africans, are voting based on assessments of policy performance and the economy in particular. Going further, we claim that the successful mobilisation of working-class voters by its presidential candidate, Muhammadu Buhari, significantly contributed to the APC's victory. We label these class-based appeals a ‘talakawa effect’, referencing a Hausa term for the ‘common people’ excluded from clientelist systems. While we do not systematically analyse issues discussed in the campaign, we do reference some of the APC's populist rhetoric in the south in order to justify our extension of talakawa beyond its classic meaning limited to northerners. Our tests are the only ones we know of that include 2015 presidential election results at the local government level – data that the government has still not fully published.
In the essay's first section we provide background about elections in the Fourth Republic, Nigeria's longest stretch of civilian rule and democracy since independence in 1960. We briefly elaborate on the conduct of the 2015 elections and common explanations for the PDP's defeat. Then, drawing on comparative research, we generate expectations for how patronage, subnational economic performance and violence could theoretically shape voting behaviour in Nigeria. In the next section we test three hypotheses: first, borrowing a popular campaign phrase used to describe patronage spending, our ‘stomach infrastructure’ hypothesis predicts that voters in states where the governor padded budgets ahead of elections were more likely to vote for the PDP. Such promises of future spending were particularly credible, given the PDP's dominance of governorships and billions of dollars in oil revenue grants controlled by the governors, who have broad constitutional latitude in how to spend them. Using original variables separately measuring capital and recurrent spending, none of our results is significant. Second, a public safety hypothesis predicts that states that experienced lower levels of violence were more likely to vote for the PDP. Even though national surveys and qualitative analyses highlight citizens’ concern with insecurity and terrorism, we find no statistical association between observed violence and voting behaviour at the state level. Third, an economic performance hypothesis predicts that positive economic performance corresponds with support for Goodluck Jonathan, the incumbent PDP president. We test this in three ways: first using states’ internally generated revenue as an instrument for subnational economic performance, second with perceptions of the national economy disaggregated at the state level, and finally using literacy as a proxy for income in local government areas. In all three tests, indicators of positive economic performance – whether subjective or objective – are highly correlated with support for President Jonathan, offering robust support for the economic performance hypothesis. Our results hold across a range of controls, and we interpret the consistently positive values on a literacy variable as evidence of the talakawa effect since comparative research and Nigerian studies link literacy rates to socioeconomic standing.
The essay's concluding section describes how our new explanation for the PDP's defeat advances our understanding of how dominant parties lose elections, especially in an era of ‘hybrid’ regimes that effectively limit political competitiveness (Levitsky and Way 2010). Africa's largest political party lost owing to more than just exogenous factors such as a global decline in oil prices or recurring terrorist attacks. Instead, our findings hint at a hopeful message about emerging limitations on national politicians’ ability to coordinate patronage for partisan purposes. Our results further emphasise the importance of economic conditions at the national and subnational levels, the electoral consequences of voters evaluating those conditions, and the qualities of effective economic populism. Like voters elsewhere in Africa, Nigerians are showing some signs of voting based on assessments of policy performance. Yet we caution that a recent growth in Nigeria's middle class combined with Buhari's unusual populist credentials provided a convergence in 2015 to break with historical patterns, and may not form the basis of a more stable class-based cleavage. Further, even though the APC won a more geographically dispersed distribution of votes than its predecessors did, we are careful not to suggest that the pull of regional politics has been rendered irrelevant.
Literature review
We begin with a brief background on elections in Nigeria since 1999, known as the ‘Fourth Republic’ after three previous attempts at democracy. Election quality and levels of political violence have varied tremendously during this period, as have citizens’ attitudes about institutions and electoral politics. Next, we review existing explanations for why the PDP lost in 2015, which generally point to voter concerns about the economy, terrorism and corruption. We go on to describe violence, patronage and other factors from the comparative literature that could constitute alternative explanations for the PDP's defeat and that inform our subsequent hypotheses. Finally, we elaborate on the idea of the talakawa historically, extending the concept beyond its conventional understanding. We also situate our theory of a ‘talakawa effect’ within contemporary findings about African support for opposition parties.
Elections in the fourth republic
Nigeria's post-1999 democratic trajectory has been parabolic. The 1998–99 elections were transitional elections: they passed muster insofar as they were peaceful, free from military interference, restored civilian rule and revived democratic institutions. The elections brought to power a president – Olusegun Obasanjo – and many governors who were ex-military or veterans of the Second (1979–83) and Third (1992–93) republics. Obasanjo's party, the PDP, rapidly consolidated its hold using ‘elite accommodation, the maintenance of its own electoral supremacy, and the distribution of patronage’ (Lewis and Kew 2015, 95). In the 2003 elections, the PDP expanded its influence amid ‘worrisome levels of disorganisation, fraud, and intimidation’ (Lewis and Kew 2015, 96).
The 2007 elections were the Fourth Republic's democratic nadir. Following a failed attempt to secure himself an unconstitutional third term, President Obasanjo sought to engineer the election of a malleable successor through which he could influence national affairs. Voting was so chaotic that it was often altogether not feasible (Rawlence and Albin-Lackey 2007; Obi 2011). Calling it an ‘election programmed to fail’, a coalition of hundreds of civil society organisations with 50,000 election monitors concluded, ‘We do not believe that any outcome of the elections can represent the will of the people. A democratic arrangement founded on such fraud can have no legitimacy’ (“An Election Programmed to Fail” 2007).
The 2011 elections marked an upswing in Nigeria's democratic trajectory and restored domestic and international confidence in the country's electoral institutions thanks in large part to the new chair of the Independent National Electoral Commission (INEC), Attahiru Jega. Although the elections were widely hailed as Nigeria's most free and fair in recent history, voter turnout was grossly exaggerated in ruling party strongholds and suppressed in opposition-held areas, while post-election violence left 800 dead (Akhaine 2011; Owen and Usman 2015). INEC subsequently instituted three key reforms aimed at enhancing the integrity and credibility of Nigerian elections: continuous voter registration, permanent voter identification cards and electronic voter accreditation on election day (Owen and Usman 2015). INEC's performance during off-cycle elections in several states in 2012 and 2013 varied (Transition Monitoring Group 2013). By 2014, however, INEC organised back-to-back elections in Ekiti and Osun states that were generally hailed as peaceful and credible (Ibrahim 2014).
The 2015 elections: prevailing narratives on the ruling party's loss
President Goodluck Jonathan and his ruling PDP lost control of the presidency, legislature and a plurality of state governorships to Muhammadu Buhari and his APC party in the 2015 elections. The electoral defeat of an incumbent and the peaceful handover of power to the opposition was a first for Nigeria. In many ways, Buhari's victory upended existing narratives about the advantages enjoyed by Nigerian incumbents in Nigeria, the fractiousness of the country's opposition and Buhari's own perennial ‘also-ran’ status after his electoral defeats in 2003, 2007 and 2011.
Lewis and Kew (2015) identify four factors that explain the defeat of Jonathan and the PDP: the emergence of a united, well-funded opposition; poor presidential performance and ruling party corruption; better INEC capacity and leverage; and increased assertiveness by civil society groups. Owen and Usman (2015) agree that a more capable INEC and poor executive performance were decisive, but they also argue that ruling party splits and the use of new campaign tactics helped propel Buhari to victory. They also suggest that a more credible primary process, well-timed engagement by international voices and the election postponement helped Buhari win.
In terms of Jonathan's poor performance, most evidence and commentary points to two specific shortfalls: his administration's lacklustre efforts to combat Boko Haram and its unwillingness to rein in corruption (Siollun 2015). Polls by Gallup showed that two-thirds (67%) of Nigerians said the government was ‘not doing enough’ to fight terrorism. Fully 95% called Boko Haram a ‘major threat’, with a virtually indistinguishable difference between Christians (who predominate in the south) and Muslims (who form a large majority in the north). Overall confidence in the national government dropped from 55% in 2010, when Jonathan succeeded his predecessor who died in office, to 29% in 2014 (Loschky 2014). On both corruption and security, Buhari arguably had a stronger track record than Jonathan. As an Army officer, he had seen combat experience during the civil war and later during border skirmishes with Chad. As military head of state (1983–85), he made the fight against corruption his signature issue. Buhari also touted his experience as petroleum minister (1976–78), contrasting it with the alleged corruption and mismanagement at the hands of Jonathan, failures that were even more glaring in light of historic lows in global oil prices.
Last but not least, Jonathan's re-election prospects were undercut by a major party split in the ruling PDP ahead of the election. In September 2013, five governors – including the governors of vote-rich Kano State and oil-rich Rivers State – broke ranks and joined the APC. A variety of factors triggered the split, including lingering resentment of Jonathan's 2011 presidential run among northern elements of the PDP, interference by Jonathan and his allies in state-level political disputes, and attempts by Jonathan to erode governors’ influence over party structures (Lewis and Kew 2015; Owen and Usman 2015).
Literature on African elections – and how today's talakawa are missing from it
The literature on African elections features several recurring themes that could generate alternative explanations for the defeat of a powerful party like the PDP. Patronage emerged as one such common theme in the decades after independence, with studies linking it to structural conditions shaped by colonialism and then later to neoliberal retrenchment (Chabal and Daloz 1999). Patronage fuelled ethnic polarisation as rulers steered government resources either to their kin or to groups allied with the ruling coalition of elites. Such analyses became so common that Gyimah-Boadi (2007) asked whether the compelling logic of patronage was applied too uncritically. Recent studies suggest a more revisionist view of patronage is warranted. For example in Kenya, ethnicity has long been associated with patronage, yet demand for public goods is on the rise (Gibson and Long 2009). Apparently, many ruling parties have come to realise that patronage is costly to maintain where the political risks of exclusion are high. Chang and Kerr (2017) find that those who lack access to patron–client relationships (like the talakawa in our study) are much more likely to support opposition candidates across 18 African countries.
Electoral violence constitutes a second common research theme, and with good reason: violence and intimidation occur in 58% of African elections, compared with 19% globally (Burchard 2015). One influential study differentiates between pre- and post-election violence, reporting that the risk of violence increases when the ruling party fears it will lose (Bekoe 2012). Economic stress and political uncertainty generated by democratic transitions are cited as common causes of electoral violence in Africa. ‘Even after third and fourth round rounds of competitive elections, Africa is still in a difficult process of transition’ (Laakso 2007, 226). In general, ‘triggers’ and causes of electoral violence are thoroughly explored by the literature but the broader political consequences of violence are less well known.
Class-based theories offer a third theme in electoral politics, though they have fallen out of favour since the post-colonial era. Sklar built a career by extending class-based analyses beyond classic Marxist conceptions, applying class to a broad range of ‘post-imperial’ phenomena (Sklar and Falola 2002). During the years of structural adjustment in the 1980s, worker coalitions in many African countries formed along class lines before they turned to mobilisation for democratisation (Kraus 2007). Well-known studies in the tradition of class analysis include Feinstein's study of the Northern People's Congress, which leveraged the rigid social stratification reinforced by indirect rule to mobilise voters based on socioeconomic distinctions (Feinstein 1973).
Nigeria has a strong tradition of class politics, with Aminu Kanu offering the clearest expression of an electoral coalition built along economic cleavages. His Northern Elements Progressive Union (NEPU) organised northern talakawa, plural of the Hausa word talaka, meaning the ordinary poor person in the street who holds no official title. Bagery describes talakawa as an ‘amorphous subaltern class, embracing small traders, workers, informal sector workers, the unemployed’ (cited in Watts 1992, 40). In other words, the talakawa stood apart from the office-holding class, masu sarauta, who wielded power over these largely illiterate masses. A stated aim of NEPU was therefore to place political power ‘in the hands of the talakawa’ (Reynolds 1999). In the decades that followed, the political power, awareness and tactical sophistication of the urban working class increased (Gambari 1992). According to Lubeck (1986), this produced ‘talakawa nationalism’, fuelled by a higher sense of class deprivation, surplus extraction by the local aristocracy and denial of access to state patronage. Class-based coalitions continued to shape politics through Nigeria's Second Republic (1979–83), as the aging Aminu Kano organised in the urban north (Forrest 1987). Buhari hardly shared Kano's progressive politics as military head of state (1983–85). But he did deliberately circumvent the emirate clientelist systems by building ties with the ‘Kaduna Mafia’, a group of northern elites with some economic and political autonomy from traditional Muslim authority (Othman 1984).
When democracy returned after 1999, conditions conducive to class-based organising continued or expanded, yet class analysis remained largely missing from research on Nigeria's five election cycles. By the African Development Bank's metrics, 23% of Nigerians are now ‘middle class’ (Robinson, Ndebele, and Mhango 2011). Anecdotal indicators of this economic transformation are equally evident. ‘In Nigeria, a person who is able to purchase a generator for personal use and run said generator every time power goes off is a member of the middle class,’ writes the novelist Elnathan John. That person often has a car with access to expensive black-market fuel and to ‘a person in government or authority who can change the course of events in their favor’ (John 2015).
However, as an oil boom from 2004–14 stimulated economic growth during the PDP's long reign, millions of modern talakawa were left out of the country's new wealth. By 2012, the media regularly contrasted an encouraging macroeconomic portrait with federal statistics documenting a rise in poverty. Even as middle-class Nigerians’ access to private-sector employment, consumer goods, digital communications technology and new political tools like social media grew (Orji 2016), a much larger and poorer segment of the population saw few socioeconomic gains (Ogunlesi 2014). As a result, the political interests of Nigeria's booming middle class and struggling underclasses began to diverge sharply, setting the stage for class-based interests to shape voting patterns during the 2015 elections.
The theory of the talakawa effect
Our study thus aims to bring class back into the conversation on electoral politics, considering its viability as an alternative to ethnic electoral coalitions and especially in light of recent economic growth. Exclusion from economic largesse under the PDP became a rallying cry for Buhari's APC, whose 2015 campaign specifically appealed to ‘the masses’ in Yoruba, Igbo and Hausa languages: ‘The APC is determined to give a voice to the youth, mekunu, the umu obenye, and the talakawa’ (All Progressives Congress 2015). By tapping into these constituencies, Buhari revived a political tradition that organised along class lines, extending it from its classic roots in the urban north into the south. We identify this as a broader ‘talakawa effect’, expanding the concept beyond its regionally specific (and Hausa) context, using it to broadly describe poor voters left behind by Nigeria's macroeconomic growth and without access to patron networks described above by John. Since the above research closely aligns talakawa social status with limited access to formal education, literacy should reliably predict patterns of voting. If Buhari's campaign appeals in Yoruba and Igbo successfully mobilised voters, we would expect these patterns to hold in the south as well. Literacy is a good proxy for socioeconomic status generally in Nigeria (Osili 2008).
In our theory, this low status implies more than just poverty or limited education. It also implies that disconnection from clientelist benefits generates frustration that manifests itself at the ballot box, like the disenfranchised African voters identified by Chang and Kerr (2017) who cast their lot with opposition parties. To explore this, we draw upon inferences about patronage spending at aggregated levels, across Nigeria's 36 governors, who control huge financial resources. Statutory allocations to states more than doubled with the rise in oil prices between 2006 and 2014 (Central Bank of Nigeria 2015). The absence of a systematic pattern between voting and state spending levels would lend some support to the idea that the PDP left many Nigerians behind during the economic boom. It would also generally contribute to the above research pointing to the importance of policy performance evaluations for African voters.
Finally, an electoral coalition along class lines would mobilise feelings of economic alienation more effectively than fears of (physical) insecurity. For the APC, this amounted to a gamble given post-election violence in 2011 and widespread worries of a repeat in 2015 (Nengak, Mbaegbu, and Lewis 2015). It also means that, unlike much of the above literature on electoral violence, we are interested in the general effects of observed violence on voter preferences as citizens evaluate the incumbent PDP's ability to provide public safety as a public good.
Hypothesis testing
In this section we test our theory of a talakawa effect through variables that measure economic, social and political conditions shaping the presidential election. Our dependent variable is the change in the PDP's vote share in the presidential elections of 2011 and 2015, measured in two ways: first, ΔPDPstatevote utilises official state-level results from INEC. Second, ΔPDPlocalvote is an original variable that combines the 288 presidential official election results reported by INEC with 331 local government area (LGA) results posted by accredited observers and civil society groups on social media at the time of the election.1 This yields N = 619 out of 774 LGAs. To our knowledge, this is the first time that 2015 presidential election results across LGAs have been published, and we share the data at carllevan.com/data/Elections2015-LeVan-Page-Ha.
We begin by statistically testing two null hypotheses characteristic of common explanations for electoral outcomes in Nigeria. A ‘stomach infrastructure’ hypothesis seeks to determine whether governors, who have considerable resources at their disposal owing to Nigeria's fiscal federalism described above, engaged in a patronage strategy by increasing state spending in order to buttress subnational support for their party. Next we test a ‘public safety’ hypothesis using violent deaths per state. Since Nigeria does not have state or local police, citizens usually point the finger at the federal government when insecurity increases, meaning that public safety is an especially appropriate policy to assess in a presidential campaign. Most importantly, a dramatic increase in terrorism under Jonathan's watch made public safety a prominent issue for both citizens and politicians. We find no systematic evidence supporting either null hypothesis.
Next, an economic performance hypothesis predicts that positive economic performance corresponds with support for President Jonathan. We test this first using an original variable measuring internally generated revenue (ΔIGR) to instrument for subnational economic performance. We expect an increase in a state’s internally generated revenue to correlate with increased support for the PDP. Second, the variable NatEcon measures perceptions of the national economy based on Afrobarometer Round 6 survey data disaggregated at the state level. We expect the PDP vote share to increase in states where citizens have positive views of the economy. Finally, using the dependent variable ΔPDPlocalvote and literacy as a proxy for income across Nigeria's local governments, we predict that the PDP will enjoy higher levels of support among voters with more formal education. These disaggregated data give us a good basis for making inferences about voting behaviour, rather than simply observing electoral patterns across states. All three tests offer support for the economic performance hypothesis and hold across a broad range of controls. States and local governments with strong economies were risk averse and overwhelmingly voted to keep the PDP, while areas struggling economically voted for the APC. We attribute this to a ‘talakawa effect’ as Buhari's populist appeals successfully mobilised poor, largely illiterate voters.
Control variables
All of our models include variables controlling for conditions that could interfere with the hypothesised relationship. Literacy measures the literacy rate in English and in the respective state's local language, expressed as a per cent (National Bureau of Statistics 2010; UNESCO 2012).2 We probed for the validity of literacy by testing it for collinearity first against household income measured by family income not spent on food, according to the government's Annual Abstract of Statistics. After logging household income, the Pearson's R with literacy is .7992. We also ran collinearity tests with infant mortality rate per 1000 people from the Multiple Indicator Cluster Survey, produced by the Nigerian government and UNICEF, which is based on a statistical sample of 29,600 respondents (National Bureau of Statistics, UNICEF, and United Nations Population Fund 2013). Pearson's R was −0.7389, further suggesting that literacy is a good proxy for broader socioeconomic conditions.3 Literacy is the better measure because of its statistical normality, and we are confident that it captures these broader socioeconomic conditions. This is an important control because voters in poorer states or LGAs could simply seek to replace the PDP in order to improve their own subjective economic conditions. They could also be risk averse, leading them to support the PDP.
Next, using data from the government's Debt Management Office (2016), we constructed the original variable statedebt, which measures subnational fiscal discipline calculated as a logged ratio of each state's debt (foreign and domestic) between 2011 and 2014 in real terms, using the World Bank's Gross Domestic Product GDP deflator (The World Bank 2017). In broad terms, fiscal discipline signals the willingness of politicians to borrow from the future, which is an especially common problem in Africa's monocultural economies (Collier and Gunning 2008). State borrowing increased during Jonathan's final years as federal grants decreased in line with global oil prices. In this sense, debt levels are not merely an indicator of governors’ time horizons, they also instrument for states’ desperation vis-à-vis the federal government. Borrowing money generates political vulnerability as states that do so have less leverage when bargaining with national party leaders or federal government officials.
Finally, we control for important political conditions in each state. The dummy variable govparty has a value of 1 for a PDP governor (and APGA for Anambra, which had adopted Goodluck Jonathan as its candidate) and 0 for APC. The PDP held 23 states after the 2011 elections, so the defection of five PDP governors to APC in 2013, along with 57 members of the 360-seat House of Representatives and 22 of 109 senators, deeply damaged the ruling party's credibility. The variable is coded after all of these defections. In addition, Δturnout controls for voter turnout, measured as per cent change in turnout between the 2011 and the 2015 presidential elections. This is important because apathy among ruling party supporters and/or intensity of support for the opposition could be driving the presidential results, resulting in a decline in turnout in states that the PDP lost and an increase in turnout in the states that the APC won.
Our base model is:
‘Stomach infrastructure’ hypothesis
Borrowing from a phrase popularised by Ibrahim (2014), we dub our tests for patronage the ‘stomach infrastructure’ hypothesis. When voters in Ekiti State were asked why a popular and educated governor lost a special election in Ekiti State in 2014, they responded with a Yoruba adage: Ona lo mi ko, ko ko inu! roughly meaning the governor built infrastructure – roads and bridges – but the people really want an infrastructure of the stomachs. As noted earlier, governors control massive budgets. They also face few political checks given the weak capacity of state legislatures. Moreover, since the 2015 budgets were passed in December 2014, barely three months before the presidential election, the timing makes patronage offers especially plausible.
Patronage could come in the form of capital spending, especially given Nigeria's tradition of large ‘mobilisation fees’ paid upfront for projects that often never materialise. Large increases in spending could also be visible in recurrent spending if governors and their allies attempt to put ‘ghost employees’ on the payroll. To accommodate both of these scenarios for patronage, we disaggregate state budgets into two original variables: capbudget measuring capital spending and curbudget measuring recurrent spending. Data were taken from Nigerian Muse's Star Information (accessed through AllAfrica.com) on proposed state budgets.4 We also searched 14 Nigerian newspapers accessed through Factiva and AllAfrica.com, for missing data. The hypothesis predicts that governors will increase spending prior to the 2015 election. The results are displayed in Table 1.
| Model 1 (Capital budget) | Model 2 (Recurrent budget) | Model 3 (Public safety) | |
|---|---|---|---|
| Capbudget | −.566 (2.403) | ||
| Curbudget | 2.259 (3.840) | ||
| LogViolent | −.702 (1.908) | ||
| Statedebt | 3.654* (2.123) | 3.305 (2.185) | .3.677* (2.145) |
| Literacy | .193** (.093) | .169* (.100) | .182* (.097) |
| govparty | 4.462 (4.291) | 4.690 (4.299) | 4.224 (4.301) |
| Δturnout | −.126 (.135) | −.118 (.131) | −.140 (.139) |
| Constant | −21.834 (22.654) | −47.999 (35.070) | −25.797*** (6.537) |
|
N R2 | 36 .245 | 36 .250 | 36 .246 |
Robust standard error in parentheses.
***p < .01.
**p < .05.
*p < .1.
Neither capital spending in Model 1, nor recurrent spending in Model 2, is statistically significant, meaning that we can neither confirm nor reject the ‘stomach infrastructure’ hypothesis. We take this to indicate that state-level patronage did not play a systematic role via the states – despite ample resources available to the governors. The control variable that stands out is the literacy rate, which is significant at the .05 level in Model 1 and barely significant in Model 2 (p = 0.101). Literate, better-off voters were consistently more likely to vote for the PDP.
Public safety hypothesis
As noted earlier, public safety was a huge issue facing the PDP. ‘What we will do in the immediate is to tackle Boko Haram,’ promised the opposition APC just prior to the election (Adeyemo, Soyinka, and Adeosun 2015). Buhari argued that the efforts were flagging, while Jonathan campaigned on a military ‘surge’ that led to the postponement of the elections (Idris, Mutum, and Doki 2015). In fact, trust in Nigeria's security services, which was low to begin with, saw a steady decline from 2011 to 2014 under Jonathan's tenure: as Boko Haram's attacks increased, confidence in the military went from 78 to 57%, and confidence in the police declined from 49 to 33% (Loschky and Sanders 2015).
Political economy literature often considers security (Zedner 2006) or public safety as a public good and a measure of government performance (Schwab and Oates 1991). To this end, the variable violentdeath measures overall levels of violence in each state, capturing both state and non-state violence, using data from Nigeria Watch (n.d.). We chose this database because it extends back to the beginning of Jonathan's term, allowing for a good baseline prior to a spike in violence later in his term, and also because the data collection methods seek to maintain high reliability even at disaggregated levels (Chouin, Reinert, and Apard 2014). The variable is constructed from five years’ average of relative death, compared with the population of each state as follows:
Its distribution is positively skewed, so we log the variable creating LogViolent. We predict that voters in states with higher levels of violence were more likely to vote against the PDP. Model 3 in Table 1 shows that LogViolent is not statistically significant. This suggests that proximity to violence did not systematically shape voting. The statedebt variable is weakly significant, and importantly for our argument about the talakawa, the coefficient on literacy is significant at the 0.1 level (p = 0.069). The tests offer no support for the public safety hypothesis. This is surprising not only because of the widespread concern about Boko Haram expressed in Gallup polls and elsewhere, but because the APC made security sector reform and its critique of Jonathan's counter-terrorism strategy a centrepiece of its campaign (Mohammed 2014). Though we do not directly analyse the effects of campaign rhetoric here, this seems plausible and ripe for future research.
The ‘economic performance’ hypothesis
In the tests rejecting our two null hypotheses, literacy stood out as statistically significant. Even though we noted that literacy correlates with infant mortality, disposable income and measures of social status in Nigeria, this is still only indirect evidence of a talakawa effect. Our economic performance hypothesis provides a more direct test. It predicts that areas with stronger economies are more likely to vote for the incumbent president. We test this first by instrumenting for states’ economic performance with a variable measuring internally generated revenue (IGR), then using perceptions of the national economy disaggregated at the state level, and finally using literacy as a proxy for income at the local government level. These local disaggregated data give us the strongest basis for inference about voting behaviour. In all three tests, positive economic performance is highly correlated with support for the PDP, while states and local governments with gloomier economies cast their lot with the APC. The local level tests are presented separately since the models differ owing to data limitations.
To gauge states’ economies, ideally we would use GDP per capita at the subnational level. Although such data are available for 2011 (Federal Government of Nigeria 2016), we opted not to use them owing to reliability concerns. For example, calculation methods are not described and in several instances different states have precisely the same GDP. We therefore use data on IGR collected from the Central Bank of Nigeria's Annual Reports for 2010–13 (Central Bank of Nigeria 2010, 239, 2011, 274, 2012, 300, 2013, 348). The variable ΔIGR is a difference of ratio calculated by subtracting the mean IGR as a share of state between 2010 and 2012 from IGR as a share of total revenue in 2013. We believe IGR offers a valid measure of subnational economic performance because it captures the ability of states to monetise their economic productivity. Increasing state revenue is difficult without economic growth and commerce that states can tax to generate revenue (Oseni 2013). IGR is also a good indicator of economic performance because it is distinct from the ‘bonus’ oil revenue that accrues to oil-producing states under the ‘derivation formula’, which might otherwise positively skew their economic performance relative to non-oil-producing states. Traditionally, entitlement to federal revenue allocations created disincentives for subnational revenue generation or economic diversification (Obi 2000). However, these patterns of dependence are rapidly changing as evidenced by surprising variation in tax collection across states (Elemo 2015). Hypothesis 3a, the subnational economy hypothesis, therefore predicts that high levels of IGR will correspond with support for the PDP, implying that voters in states with healthy economies were less likely to take a chance on an untested opposition party less than two years old. The results are displayed in Table 2.
| Model 4 (Subnational economy) | Model 5 (Perceptions of national economy) | |
|---|---|---|
| ΔIGR | .848** (.337) | |
| NatEcon | 14.601*** (4.463) | |
| statedebt | 3.114 (1.873) | 4.410** (1.941) |
| Literacy | .216** (.087) | 0.14 (0.116) |
| govparty | 5.049 (4.192) | 4.546 (4.528) |
| Δturnout | −.177 (.1424) | 0.132 (0.139) |
| Constant | −29.576*** (5.615) | −62.091*** (10.070) |
|
N R2 | 36 .351 | 33 0.454 |
Robust standard error in parentheses.
***p < .01.
**p < .05.
*p < .1.
The variable ΔIGR is statistically significant at the .05 level. Rather than generating political insulation from the federal government and inspiring voters to cast their lot with the APC, states that effectively monetised economic growth were actually risk averse and therefore more likely to support the PDP. By contrast, voters in states with weaker economies were more likely to vote for the APC. The only control variable that is significant is literacy.
Next, a ‘perceptions’ hypothesis tests to see whether subjective evaluations of the national economy influenced voting patterns across states. This is important because in March 2014, Goodluck Jonathan's administration rebased and recalculated the GDP for the first time in over two decades and announced that Nigeria had the largest economy on the continent. Soon thereafter he also effectively fired the Central Bank governor, prompting criticism from the opposition that his administration sought to manipulate fiscal policy. Moreover, the global price of crude oil dropped steeply in 2014, further upsetting the administration's economic portrait and lending urgency to the opposition APC's calls for economic diversification. We operationalise subjective perceptions of the economy with the variable NatEcon, based on data from the Afrobarometer Round 6 survey.5 Administered in late 2014, respondents were asked, ‘Looking back, how do you rate [economic conditions] compared to twelve months ago in this country?’ Respondents answered on a scale of 1 to 5, where 1 is ‘much worse’ (13%), 2 is ‘worse’ (32%), 3 is same (26%), 4 is ‘better’ (24%) and 5 is ‘much better’ (4%). Hypothesis 3b expects citizens with positive perceptions of the economy to vote for the ruling PDP, which would appear as a positive coefficient on NatEcon. The results in Table 2 show that positive perceptions of states’ economic performance are highly correlated with PDP support, with NatEcon statistically significant at the .01 level. This also means that in states where Nigerians had negative views of the economy, they were likely to vote for the opposition APC, lending strong support for the perceptions hypothesis.
Finally, a ‘local economy’ hypothesis predicts that LGAs with stronger economies are more likely to support the ruling party, as measured by ΔPDPlocalvote. Since data on internally generated revenue, debt or GDP do not exist at the LGA level, we again rely on literacy as a proxy for socioeconomic conditions. We add a control for language, which reflects the number of languages widely spoken in each LGA as recorded by the government (National Bureau of Statistics 2007). The results are reported in Table 3. The reduced N is due to loss of information in the control variable.
| Model 6 (LGA level) | |
|---|---|
| Literacy | 0.073** (0.034) |
| Language | −0.783 (0.604) |
| Constant | −17.131*** (3.093) |
|
N R2 | 583 0.013 |
Robust standard error in parentheses.
***p < .01
**p < .05
*p < .1.
The positive coefficient on literacy indicates that LGAs with higher socioeconomic levels were more likely to vote for the PDP. This is statistically significant at the .05 level. This support for the local economy hypothesis is especially important because by disaggregating electoral results to the local level – we believe for the first time – we are better situated to make inferences about voter preferences, rather than just observe voting patterns across states. It further means that our three sets of tests, using both objective measures of economic performance and subjective perceptions of the economy, offer compelling support that Nigerians in 2015 engaged in ‘economic voting’.
In sum, economic conditions and perceptions of those conditions systematically predicted partisan preferences better than patronage or proximity to violence. The analysis of election results across 36 states and 619 (out of 774) local governments is especially compelling because they increase the validity of one of our inferences about subnational conditions and voting behaviour. Moreover, the positive correlation between lower levels of literacy and support for the APC in Models 1, 2, 3, 4 and 6 provide evidence6 of what we called the talakawa effect: Buhari enjoyed widespread support among ‘commoners’, those often excluded from elite networks – such as those that could have flipped the results of the ‘stomach infrastructure’ hypothesis. We are also confident about the existence of a talakawa effect because lower literacy consistently aligns with support for the APC at a statistically significant level.7 Like his (undemocratic but popular) rise to power in 1983, class-based appeals formed a critical part of Buhari's campaign strategy that successfully defeated Nigeria's long-dominant political party. Our findings should offer encouraging news to Nigerians weary of ethnic voting and old modalities of political patronage, and ready for elections that assess economic performance.
Conclusion
No single factor accounts for Goodluck Jonathan's resounding defeat at the polls in 2015. His administration's failure to deter a rising tide of terrorism and its unwillingness to seriously combat corruption provided the APC with a coherent set of issues to run on. Fragmentation within the ruling PDP, and declining oil revenues that weakened patronage networks, further presented the upstart opposition with an historic opportunity for electoral challenge. Most importantly, though accounts of the PDP's defeat refer to economic performance and the Jonathan administration's corruption (Owen and Usman 2015), none have fully detailed the political interests Nigeria's new middle class or the consequences of excluding the traditional working class from macroeconomic expansion.
Our tests suggest that assessments of the economy, and the talakawa's continued marginalisation as the economy grew, played a larger role in 2015 than previously appreciated. First, to the extent that governors used patronage, we did not find evidence that it systematically influenced voting behaviour. This is not to say that politicians did not use patronage, and we hope that future research will be able to disentangle targeting of patronage. Second, the public safety hypothesis found no systematic relationship between proximity to violence and voting behaviour. This was a surprise given overwhelming public concerns about Boko Haram expressed in polls. Third, we found robust support that economic conditions mattered, with voters in poorly performing state and local economies more likely to vote for the opposition APC. At the state level, these results are especially compelling since rather than using GDP, which could mask vast inequalities in states that appear to be performing well, we used both internally generated revenue and budget deficits. These two variables capture different logics of political economy: debt implies a dependence on the centre while higher levels of IGR point to economic self-sufficiency. Since they both perform as hypothesised, this increases our confidence in the results. Voters in areas with healthy economies were risk averse and opted to keep the PDP in power. Our findings point to subtle but systematic effects of a class-based electoral coalition.
Since we only analysed a single election cycle, we remain a long way from arguing that economic cleavages will displace ethnic or regional allegiances. Also, regional politics still have an allure, and we note that under alternative specifications our statistical models suggest a stronger literacy effect across the north. Moreover, given Buhari's historic association with the talakawa and his populist credentials (Paden 2016), we remain cautious about claiming that future candidates will be able to mobilise talakawa with similar zeal. Similarly, we caution against interpreting our results to mean that APC effectively changed voting preferences. The 2015 presidential campaign shared some of the hollow promises of previous campaigns, and we leave it to future research to analyse the content of the APC's campaign rhetoric. We hope that our publicly available dataset, with original variables measuring state debt, internally generated revenue, state spending, gubernatorial turnover and presidential election results at the LGA level, will help answer some of these questions.
Despite these caveats, our results convincingly demonstrate that less educated people across the country were more likely to vote for the APC. Strong correlation with other socioeconomic characteristics, alongside the absence of systematic evidence of partisan effects of patronage spending, suggests that Nigeria's talakawa followed the example of the socioeconomically alienated voters identified by Chang and Kerr (2017), casting their lot with an untested opposition. Buhari's class coalition provided a missing electoral link in the political economy of the PDP's defeat. This also suggests that Nigerians do sometimes evaluate political candidates based on policy performance. This is hopeful news for its democratic development, and a powerful lesson for future presidential aspirants.