Political influence on the media

Freedom House’s Freedom of the Press rating relies on regional and country experts who assign scores to a country based on the legal environment, including the protection of journalists and how media laws are implemented; the political environment, focusing on political pressures on content through, among others, ownership structures and sourcing; and the economic environment, including production costs, advertisement dependence, and sensitivity to corruption. Most countries in Africa have a “partly free” or “not free” rating. For the year 2017, Nigeria had an overall score of 51/100, with higher scores indicating less freedom. In particular, the political environment received a high score, similar to scores in previous years.^{Footnote 1}

Indeed, while Nigeria has one of the most vibrant press media landscapes in Africa today, its news media continue to experience political dependencies. On the one hand, the press performs a strong watchdog role in the Fourth Republic by revealing political corruption scandals and promoting accountability (Freedom House 2016; Olukoyun Reference Olukoyun2004; Ojo Reference Ojo2003). This role can be seen as a remnant of their opposition campaigns to previous military dictatorships. On the other hand, the media remain sensitive to political and ethno-regional pressures in their reporting. A much-cited example to support this claim (e.g., Jibo & Okoosi-Simbine Reference Jibo and Okoosi-Simbine2003:183–84; Olukoyun Reference Olukoyun2004:83; Yusha'u Reference Yusha'u and Mano2015:144–45) is that southern newspapers were virulent in their critiques of the northern Speaker of the House when they found out that he had lied about his age, making him ineligible to be elected to the House, and about his University of Toronto degree. A southern Yoruba senator, Bola Tinubu, faced much less criticism, however, when it turned out he had lied about his University of Chicago degree.

Historically, independence leaders such as Nnamdi Azikiwe and Obafemi Awolowo set up a number of newspapers not only to promote the independence cause, but also to increase their political influence (e.g., Adebanwi Reference Adebanwi2016:109–10; Coleman Reference Coleman1963; Jibo & Okoosi-Simbine Reference Jibo and Okoosi-Simbine2003:182–83; Sklar Reference Sklar2004). Moshood Abiola, the winner of the aborted 1993 presidential elections, also saw the ownership of a newspaper as an important tool for a political leader (Ojo Reference Ojo2003:836). Yet private newspapers as well remain sensitive to political pressures, as they are generally dependent on advertisement from politicians and affiliated businesses. Low salaries for journalists, along with persistent delays in payments, also encourage “the brown envelope system” (Olukoyun Reference Olukoyun2004:78) by which politicians bribe journalists to write favorable reports about them (see also Freedom House 2016). Furthermore, journalists are primarily dependent on political sources for information, both government officials and political party representatives (Ciboh Reference Ciboh2017). Finally, press media and journalists also face harassment and violence from disgruntled politicians and their supporters (Freedom House 2016). All these dynamics encourage self-censorship among the Nigerian press media. This is at times accompanied by actual censorship, for instance to prevent the circulation of reports that might be damaging to influential people or the government.

The media and elections

Since the return to democracy in 1999, Nigeria has held multiple federal, state, and local legislative and executive elections. Presidential elections have been held five times: in 1999, 2003, 2007, 2011, and 2015. These elections have often engendered violence. The inaugurating elections of 1999, which brought former military ruler Olusegun Obasanjo to power, still passed in relative calm, although they were characterized by widespread irregularities and fraud (Kew Reference Kew1999). The 2003 elections were again won by Obasanjo and also saw manipulations such as voter card fraud and ballot theft as well as voter intimidation (Lewis Reference Lewis2003). Relatively few deaths occurred during the electoral process, yet severe and lethal violence in Rivers state could be linked to electoral politics (Human Rights Watch 2007).

The 2007 general elections in Nigeria were regarded as seriously flawed by international and domestic observers, and a step back in Nigeria’s democratization process (Egwu et al. Reference Egwu, Leonard and Matlosa2009; Rawlence & Albin-Lackey Reference Rawlence and Albin-Lackey2007; Obi Reference Obi2008). The legitimacy of the elections was curtailed not only by clear cases of fraud—in some areas almost no actual voting took place—but also by the level of political violence witnessed. The number of deaths due to electoral violence is estimated at between 55 and 300. As Obasanjo’s attempt at a third presidential term had failed (Suberu Reference Suberu2007), PDP candidate Umaru Yar’Adua gained the presidency.

Yar’Adua died of natural causes in 2010 and was succeeded by vice-president Jonathan. Jonathan himself was elected president in 2011. The 2011 electoral process was generally regarded as free and fair and an improvement after the 2007 elections (Akhaine Reference Akhaine2011), but severe violence erupted after Buhari lost to Jonathan. While electoral violence in 2007 was most pronounced in the southwest, southeast and south-south, in particular in Ondo, Osun, and Ekiti states (Omotola Reference Omotola2010), severe electoral violence in 2011 took place in the north of Nigeria, specifically the states of Adamawa, Bauchi, Borno, Gombe, Jigawa, Kaduna, Kano, Katsina, Niger, Sokoto, Yobe, and Zamfara. This violence is estimated to have claimed around 800 lives (Human Rights Watch 2011). Part of the reason the violence was so severe is that it took on an ethno-religious dimension, with patterns of violence overlapping with prior fault lines between Christian and Muslim groupings (Angerbrandt Reference Angerbrandt2018).

The 2015 elections did not bring a repetition of the severe violence of 2011 (Lewis & Kew Reference Lewis and Kew2015; Orji Reference Orji2017). Still, around fifty deaths were reported. Rivers state in particular remains a hotbed of electoral violence. Many results of the 2015 national legislative elections have been declared void because of malpractices in Rivers (Oluokun Reference Oluokun2015). The 2015 presidential elections brought the APC candidate, former military ruler Buhari, to power. Jonathan quickly accepted defeat and no serious forms of post-electoral violence occurred.

In general, the media have played an important role in election monitoring and violence deterrence (e.g., EU election observer mission 2015:22–25; Lewis Reference Lewis2003:142; NDI 2011:42–43; Rawlence & Albin-Lackey Reference Rawlence and Albin-Lackey2007:499). Nonetheless, political influences have also arisen during election times. This has already been shown for media coverage of parties and candidates. While the Electoral Act requires equal media coverage for all political parties, there is generally an incumbent advantage in coverage. In the run-up to the 2011 elections, for instance, about forty broadcasting stations were fined for violating the Electoral Act (NDI 2011:42–43). Government-owned media clearly represented the incumbent party in a favorable light, but even private media gave greatest coverage to the ruling party. The same patterns were observable in the run-up to the 2015 elections (EU election observer mission 2015:22–25). In some cases, the PDP received up to 80 percent of electoral coverage. While disproportionate media attention for the incumbent President and his party may be a sign of political pressure on media houses, it may also be a direct result of the widespread interest most people have in the President’s actions. Similar incumbent dominance was found for political advertisements in news media, however, indicating an economic advantage as well (Alawode & Adesanya Reference Alawode and Olufunke2016).

The literature suggests that Nigerian news media are sensitive to political pressures during election times. While previous research has highlighted differences in media attention and advertisement between incumbent and opposition, this study focuses in particular on electoral violence. This focus allows an analysis of the content of newspaper reporting more in depth, focusing on how reporters actually engage with politically sensitive forms of violence. The following section explains the methodological framework for the newspaper content analyses, beginning with the newspaper selection and then addressing the coding process.

Newspaper selection

As mentioned previously, Nigeria has a vibrant media landscape. Print media, our primary data source in this article, are widely available, and the market is competitive. A large number of newspapers is distributed on a (relatively) national scale. Furthermore, Nigerian newspapers often publish issues on a daily basis, with newspapers commonly counting more than sixty pages. As papers are sold at low prices (200 to 250 Naira), most income stems from advertisements. This also leads to the high page counts. Newspapers are primarily targeted at an upper middle class and elite audience. Publishing houses are predominantly located in the south of the country. The most important northern newspaper is The Daily Trust.

Our selection of newspapers was guided by two principles. Firstly, we aimed to select newspapers which were affiliated with either the PDP or the APC as the main parties contesting the elections. The second criterion was that newspapers be relatively national in scope and have a wide readership. This excluded local, small-scale newspapers and party outlets. The major challenge was determining affiliations or “sympathy” toward a political party. For one of the newspapers selected, The Nation, this was more obvious. The Nation is a widely-read newspaper in Nigeria and is hence also collected in national and university library archives. Furthermore, its ownership is linked to Tinubu, who is the major APC strongman in the southwest region (see also the section “Political influence on the media”). The newspaper can hence be considered an APC (and previously Action Congress Nigeria) outlet (Ademilokun & Taiwo Reference Ademilokun and Taiwo2013:452). The northern newspaper, The Daily Trust is privately owned, but could also have been considered as sympathetic towards the APC, based on its status as voice of the north, which is Buhari’s regional origin. At the time of our data collection, however, we did not have access to the Daily Trust archives in Abuja due to travel restrictions.

The selection of newspapers which are favorable to the PDP was less obvious, but after consultations with Nigerian communication scholars at the University of Lagos (UNILAG), we decided to select two newspapers, The Guardian and This Day. Both are privately owned, but as mentioned in the previous section, this does not preclude political sympathies. The Guardian is sometimes considered to be representing the regional interests of the south-south zone (Ette Reference Ette, Mutsvairo and Karam2018:177; Yusha'u Reference Yusha'u and Mano2015), the region former President Jonathan hails from. Furthermore, Sunday O. Alawode and Olufunke O. Adesanya (2016) find that with regard to electoral advertisement there was more inclusion of PDP-sponsored messages than APC sponsored messages in The Guardian in the run-up to the elections, in contrast to the Daily Trust. Our own observations also indicated that PDP advertisement was substantially more prominent in The Guardian and This Day than APC advertisement, while PDP advertisement in The Nation was extremely rare. Nonetheless, the EU election observer mission report (2015:25) states that election coverage in The Guardian was relatively balanced. Overall, whether The Guardian is an independent outlet or leans more towards the PDP can be debated. Nevertheless, Guardian editorship also states that many people consider them pro-PDP (Interview AS, Lagos, December 18, 2017).

According to the EU election observation report (2015:25), coverage in This Day was also balanced. Yet This Day owner Nduka Obaigbena has also been implicated in a major arms deal scandal under the Jonathan administration (Premium Times 2017). Moreover, This Day salary payments to journalists are extremely irregular and low, indicating a potential for brown envelope journalism (Udo Reference Udo2015). The extent to which reporting on party violence is biased remains, of course, the empirical question of this study.

Coding process

For the purposes of this study we established a dataset on conflict events in Nigeria based on events recorded in The Guardian, This Day, and The Nation. The dataset covers a wide range of events, including peaceful demonstrations, riots, violent clashes, battles between the national army and Boko Haram, violence against civilians, and terrorist attacks. As several other datasets also focus on these (sub-)types of events in Nigeria, it is worth highlighting some of the differences between our dataset and other data projects such as ACLED (Raleigh et al. Reference Raleigh, Linke, Hegre and Karlsen2010) and the Nigeria Watch project.^{Footnote 2}

The ACLED dataset has recorded events of political violence in most sub-Saharan African countries since 1997, with monthly updates. Besides international news wires and NGO reports, ACLED relies on online information from a range of Nigerian newspapers. Nigeria Watch, maintained at the University of Ibadan, has recorded violent events since June 2006 and draws from ten national daily newspapers. These datasets do not allow us to study the effects of political orientation on the selection of events into the news (“selection bias”) or on how events are described (“description bias”) (Earl et al. Reference Earl, Martin, McCarthy and Soule2004), however. Firstly, datasets record the news source on which an event was based, but they do not provide information on whether events were only included in that source or also reported by other newspapers (selection bias). Secondly, the event data do not take into account differences in how an event is reported between news reports and newspapers (description bias). Thirdly, the coding scheme does not allow for possible contestation over the responsibility for the use of violence in events. The coding scheme developed for our dataset does allow these distinctions. Finally, while ACLED captures different forms of violence, including forms of vandalism, Nigeria Watch only includes lethal events, including accidents.

To sample the events, we consulted the hard-copy versions of the newspapers at the University of Lagos (UNILAG) library as well as additional data archives based in the city.^{Footnote 3} Relevant news reports were scanned and saved to a hard drive. To select reports, coders were instructed to check for relevant content by reading through the newspapers. A list of key words aided coders in the scanning process, although the presence of a key word was no condition for selection if the event was clearly relevant for inclusion. Opinion pieces, interviews, and editorials were explicitly excluded, as the focus lay on news reporting practices. The period of study was April 2014 to March 2015. The 2015 presidential elections took place on March 28-29, 2015. Hence, the data cover the one-year period before the general elections. Events which fall outside this date range but were reported in the newspapers within this period of study were also excluded.

Events were coded by the principal investigators and research assistants following the codebook developed by the authors. The coding process followed the research strategy proposed by Weidmann and Rød (Reference Weidmann and Espen2015), which consists of coding each report separately in order to investigate differences between them, rather than coding only one report of an event and leaving other reports aside. Each time coders came across a report of an event, they coded the necessary information based on that report. If the same event was spotted again, this report was also coded. Different event reports were then aggregated in a final event dataset. The dataset covers 2191 events and 7288 corresponding event reports. All newspapers were scanned a second time for missed events. Two independent coders have a congruence of around 85 percent for the selection of events after double-checking (tested on a sub-sample). Inter-coder reliability scores (Krippendorff alphas) for the variables used in the analyses are higher than 0.77 (Krippendorff Reference Krippendorff2013).

To aggregate events, we combined information on the actors involved, types of violence mentioned, and the dynamics of violence or who started/committed the violence. The aggregation of events relied on time, location, and actor information but was done conservatively: if there was a good probability that two reports concerned the same event, they were ultimately coded as one event in the aggregated dataset. The aggregation of the violence dynamics variable is particularly important in the purview of this study. The variable on who started the violence includes five answer categories: Actor A, Actor B, Actors accuse each other (disputed statements on who was responsible for the start of the violence), unknown (e.g., a PDP-APC “clash” occurred), or Not applicable (in case of a peaceful protest). When there is one report stating Actor A started the violence, or multiple reports which do not deny this finding, the aggregate event will state Actor A committed the violence. If, however, multiple reports in the same newspaper disagree (e.g. due to new party statements dismissing allegations), the coding will be “accuse each other.” As such, our analyses focus on the aggregate message newspapers disseminate on electoral violence events.

Selection bias in The Guardian, This Day, and The Nation

The occurrence of a range of events can be related to the proximity to elections. In the quantitative literature, “electoral violence” or “violence related to elections” has been conceptualized in different ways. Staffan I. Lindberg (Reference Lindberg2006), for instance, focuses on events that are explicitly linked to the electoral process, while Scott Straus and Charlie Taylor (Reference Straus, Taylor and Bekoe2012) as well as Arthur A. Goldsmith (Reference Goldsmith2015) associate different forms of political violence witnessed in a country with the electoral process if the violence takes place within a specific time frame. Furthermore, the occurrence of elections has also been used as a key independent variable in studies focusing on more general forms of violence. Ole M. Theisen (Reference Theisen2012), for example, finds that land scarcity and population pressure do not systematically affect the occurrence of armed conflict events in Kenya, but that election years systematically see more violence. Lars-Erik Cederman et al. (Reference Cederman and Kristian2013) in turn, find that competitive elections increase the chance of ethnic civil war onset.

As a range of events is potentially affected by proximity to the electoral calendar, we first inspect how many events the three newspapers have included by making use of different conceptualizations of conflict. We first focus on the general event type categories included in the dataset: protests (demonstrations, sit-ins), riots (violent protests or mob justice), kidnappings (kidnapping for non-criminal purposes), armed attacks (assassination attempts, battles between government troops and rebels, violence against civilian populations etc.), and terrorist attacks (IED, suicide bomber, car bomb etc.). These categories are mutually exclusive. Secondly, we use the Actor variable to distinguish between different types of events which have a political nature. A first type, “political party event,” includes all general event types, including protests, in which a political party is mentioned as an actor. This also includes parties other than the PDP and APC. The second and third sub-type only concern violent events in which either the APC or the PDP were victims of violence (e.g., secretariat bombing, assassination). As the conflict and actor coding is done separately for the three newspapers, the categorizations are newspaper-specific.

We also focus on the reporting of the Boko Haram conflict, as the handling of that crisis has been a major source of criticism of the Jonathan administration. Moreover, agitation against the government to “Bring Back Our Girls” has also been seen as APC-directed, most famously by Patience Jonathan, the president’s wife. Including events of Boko Haram killings and excluding events related to Nigerian army successes might undermine the PDP’s position for the 2015 elections. Again, we use the Actor variable to categorize events that include Boko Haram. However, we also make two sub-distinctions: events in which Boko Haram attacks civilians and events in which the Army is engaged in fighting the insurgents. The former could be more prone to overreporting by The Nation (APC), for instance, than the latter. The event types are not mutually exclusive, however, as some Boko Haram events involve civilians as well as troops at the same time.

Table 1 gives an overview of the event frequencies per newspaper. This Table shows that in terms of conflict reporting, The Nation includes the most events, while This Day includes the least. All newspapers are more or less similar in the number of pages they count, however. Interestingly, The Nation includes more conflict events involving political parties, and clearly more events that depict the APC as a victim of violence. In contrast to what could have been expected, however, The Guardian and This Day also report more events in which the APC is a victim than events in which the PDP is the victim, even though the discrepancies are less wide. This Day devotes most attention to events involving Boko Haram, yet other than this observation there does not appear to be an immediate bias in the reporting of Boko Haram-army or -civilian events.

Table 1.

Event type per newspaper

^a The ‘All’ category also contains events which involved other parties than the PDP and APC, as well as events which mention a “political party” without specifying the party. The category also includes a limited number of protests.

^b Boko Haram events can involve the army and civilians at the same time, hence the categories are not to be seen as mutually exclusive, or adding up to 100 percent. Furthermore, for a limited number of events, Boko Haram is linked to actors other than the army and civilians.

Event reporting and the electoral calendar

We now investigate to what extent the different event types included in Table 1 are affected by the electoral calendar. We aim to determine, on the one hand, whether events occurred more frequently as the elections of March 28–29, 2015, drew closer, and, on the other hand, whether there are differences between the newspapers analyzed. Firstly, we look at the overall relationship between time and conflict in Nigeria. Figure 1 shows the cumulative conflict count line as the March 2015 elections approached.^{Footnote 4} As the dataset starts at April 1, 2014, and ends at March 30, 2015, the x-axis runs from 362 to -2. The number of conflict counts at -2 represent the total count per newspaper in the dataset (see Table 1). As Figure 1 demonstrates, there is no clear relationship between the elections and conflict events in general in the country, nor are there differences between newspapers. The cumulative lines steadily progress without clear jumps, which would indicate an acceleration of conflict. A slight jump occurs on the election days themselves and is shown by all three newspapers. A broad conceptualization of conflict hence does not reveal a relationship with the electoral calendar (Goldsmith Reference Goldsmith2015; Theisen Reference Theisen2012). We find similar results for the Protest, Armed Attack, and Boko Haram events.^{Footnote 5}

Figure 1.

Proximity to the 2015 elections and conflict in Nigeria

Turning more in depth to the Boko Haram crisis, one can see a relationship with the electoral calendar. At the beginning of the crisis, the slow response of the federal government led to much criticism from civil society as well as from opposition parties. Eventually, more troops were deployed to the northeast, especially right before the elections, which were delayed for six weeks to allow for the operation. As a result, while attacks against civilians were more prominent between April and October/November 2014, they declined somewhat thereafter. Interestingly, this decline is only visible in The Guardian (PDP) and The Nation (APC) newspapers. Clashes between Boko Haram and the army in their turn become more prominent after September 2014, in particular for The Nation and This Day (PDP) papers. It appears that reporting on Boko Haram events is hence also not affected by the individual newspapers’ political affiliations.

Figure 2.

The 2015 elections and Boko Haram attacks against civilians

Figure 3.

The 2015 elections and Boko Haram battles with the army

As could already be derived from Table 1, more differences are expected for events involving political parties. Figures 4 to 6 show how each newspaper constructs electoral violence occurrence in the run-up to the 2015 general elections. For the broad category of conflict events which involve political parties, there is a clear connection with proximity to elections, as these events accelerate rapidly as the elections draw nearer. Another interesting feature is that in all three newspapers, reporting of violence victimhood becomes more prevalent over time. Indeed, it is not the case that The Guardian and This Day will start overreporting violence toward the PDP, as compared to The Nation, in the campaign period. The cumulative lines for PDP and APC victimhood do not cross or only slightly so (This Day). The Nation devotes more attention to APC victimhood in the campaign period, however, with the cumulative line accelerating stronger. PDP victimhood is largely overlooked.

Figure 4.

Political violence and proximity to elections (The Guardian)

Figure 5.

Political violence and proximity to elections (This Day)

Figure 6.

Political violence and proximity to elections (The Nation)

It is worth emphasizing again that all three newspapers show that the APC was more likely to be the victim of violence than the PDP in the pre-electoral period, regardless of the paper’s political affiliation. Most of this violence was perpetrated by PDP actors. This result supports earlier findings by Scott Straus and Charlie Taylor (2012), who argue that the opposition is most likely to be the victim of electoral violence in the pre-electoral period, in contrast to the view of Paul Collier and Pedro C. Vicente (2014), who argue the opposite.

The spatial distribution of political party conflict and electoral competitiveness

Previous analyses have focused on general trends on the national level. Here, we investigate how conflict relates to electoral competitiveness on the sub-national level. We only consider conflict events which involved political actors. Table 2 gives the top five states per newspaper for political party conflict. There is clear agreement that Rivers state and Ekiti states were most prone to conflict in the run-up to the 2015 presidential elections, while more disagreement arises over other states’ relative position in the ranking. This could perhaps be due to differences in regional foci or sales numbers for each newspaper.

Table 2.

Top 5 states most prone to political conflict per newspaper (number of events in brackets)

Two election moments took place in Ekiti: the governorship election in June 2014 and the presidential and legislative elections in March 2015. The 2015 vote margins—the differences in vote share between the winner and runner-up—for the listed states are (Figure 7): 91 percent for Rivers (PDP victory), 17 percent for Ekiti (PDP victory), 21 percent for Osun (APC victory), 11 percent for Lagos (APC victory), 16 percent for Edo (PDP victory), 3.5 percent for FCT (PDP victory), 88 percent for Akwa Ibom (PDP victory), 39 percent for Kwara (PDP victory), 26 percent for Oyo (APC victory), 10 percent for Benue (APC victory), 86 percent for Cross River (PDP victory), and 84 percent for Ebonyi (PDP victory).

Figure 7.

Vote margin by state, 2015 presidential elections

Source: INEC (2015). Map made with Indiemapper (http://indiemapper.io/app/)

Interestingly, there is no immediate relationship between party conflict and competitiveness. Nonetheless, in Rivers, many results were cancelled, and bye-elections were eventually won by the APC, indicating more competition than first assumed. There is also no relationship between events reported and the eventual winner of the elections at the state level.

Differences in reporting on the same events

While each newspaper reports over 150 conflict events involving political parties (see Table 1), only 70 of these events are shared by all newspapers.^{Footnote 6} Lethal events have a higher probability of being shared events. For about two-thirds of these events, there are no notable discrepancies between newspapers in how they report violence dynamics, including who is the victim and who is the perpetrator.

Some of the differences that do occur in reporting are in line with the expectation that political ties influence the reporting of violence. For instance, The Nation reports that a March rally in Lagos by the O’odua People’s Congress (OPC) is associated with the PDP, and that violence occurred against civilians and APC campaign materials. This description is not corroborated by The Guardian, however. This Day reports that APC could have also started the violence. A widows’ protest in Imo in February 2015 degenerates into violence. According to The Guardian, the violence was caused by APC thugs, while The Nation accuses the PDP of being behind the violence. This Day reflects both views, resulting in an “accuse each other” code. An attack on Borno governor Shettima from the APC is described as committed by the PDP in The Nation, while This Day speaks of “irate youth” and The Guardian of “attackers.” A burning of a PDP campaign office is attributed to the APC by The Guardian and This Day, while The Nation speaks of “angry youth.”

Yet there are also examples which are not in line with the expectations. For example, while This Day and The Nation state that unknown gunmen attacked a PDP senator in Akwa Ibom, The Guardian includes allegations that PDP thugs themselves committed the attack. Similarly, an attack on the brother of Rivers governor Amaechi, who by that time had switched to the APC, is specified as perpetrated by the PDP in The Guardian. An attack by the APC on another political party is reported in The Nation, while The Guardian and This Day do not include statements on who started the violence. The comparison of event descriptions generally supports our above findings that political conflict reporting cannot be directly associated with political preferences of newspapers.

In the following section, we contextualize these findings further through interviews with Nigerian journalists and editors. The interviews targeted various media channels, not restricted to the three newspapers investigated in the quantitative analyses. The interviews specifically aimed to elucidate more clearly how and to what extent reporters allow political affiliations to influence their reporting practices, in particular with regard to inter-party conflict.