Introduction
Having the largest population of >200 million people in Africa and a diverse ethnicity, Nigeria is one of the countries in Africa that is most severely affected by armed conflict and terrorism. Albeit terrorist attacks by Boko Haram in the North-Eastern region of Nigeria have been claimed by the Nigerian government to have decreased in recent years, total yearly armed conflicts, including battles between armed groups, explosions/remote violence, riots, and violence against civilians, have stayed at a high level. For example, the Armed Conflict Location & Event Data (ACLED) Project recorded over 4500 conflicts in both 2021 and 2022 in Nigeria, with fatalities exceeding 10,000 each year. While these fatality numbers are close to the civilian deaths in Ukraine during the period of 24 February 2022 and 15 November 2023 in the Russia–Ukraine war, as estimated by the United NationsFootnote 1, the attention paid by the world to the fatalities in Nigeria due to armed conflicts and terrorism is however negligible to that paid to the casualties of Russia-Ukraine war.
Since the 11 September 2001, terrorist attacks on the United States of America, a great deal of effort has been made to study terrorism. Among the more interesting research in this effort include studies on the regional effects of terrorism on tourism (Drakos and Kutan, 2003), the variation of terrorism across political regimes (Findley and Young, 2011), the relation between terrorism and trust (Blomberg et al. 2011), modelling of terrorism (Brandt and Sandler, 2010; Clauset and Wiegel, 2010; D’Orsogna and Perc, 2015; Gao et al. 2017; Helbing et al. 2015; Python et al. 2019), and prediction of armed conflicts and terrorism (Blair et al. 2017; Cederman and Weidmann, 2017; Hegre et al. 2013, 2019, 2021a, b; Python et al. 2021; Weidmann and Ward, 2010; Witmer et al. 2017). More pertinent to the study of terrorism and armed conflicts in Africa in general and Nigeria in particular are the studies on the relationship between economics, governance, military expenditure, and terrorism (Abadie, 2006; Abid and Sekrafi, 2020; Blomberg et al. 2004, 2007; Emeka et al. 2024; Keefer and Norman, 2008; Ogbuabor et al. 2023), as well as causes of terrorism. So far, various causes of terrorism have been identified, including education and poverty (Krueger and Malečková, 2002), the failed states (Piazza, 2008), minority discrimination (Piazza, 2012), and religion (Jones, 2006). In particular, Dim (2017) has advocated an integrated theoretical approach to the study of the persistence of Boko Haram terrorism in Nigeria, by focusing on poverty theory, relative deprivation theory, and social identity theory. In an assessment of terrorism in Africa by Chatham HouseFootnote 2, it has been emphasized that a common factor in play in the Sahel region is corruption, which has led to unpaid troops mutinying or deserting in Nigeria. Worse than corruption, marginalized groups in the Sahel often view militaries as oppressors. This injustice has helped to fuel hostilities between national governments in the Sahel and their marginalized, poor, and neglected communities. Corruption and injustice thus lay a fertile ground for Jihadist insurgencies to thrive. Similar views have also been expressed by Speakers of the United Nations warning Security Council terrorism spreading across AfricaFootnote 3: “Despair, poverty, hunger, lack of basic services, unemployment, and unconstitutional changes in government continue to lay fertile ground for the expansion of terrorist groups and the flow of fighters, funds, and weapons. In addition, the online world provides a global platform to spread violent ideologies even further”.
While the above studies and assessments by Chatham House and the United Nations are all very insightful, they do not seem to be sufficient for guiding African nations and international organizations to design adequate and concrete measures to curb armed conflicts and terrorism in Africa in general and Nigeria in particular. To help achieve this goal, in this work, we propose to first systematically characterize the spatiotemporal evolution of armed conflicts and terrorism in Nigeria using the events and fatalities data from the ACLED Project by constructing a few new and readily computable indexes, then utilizing another media big data, the Global Database of events, language, and Tone (GDELT), to construct two general indices, to describe the collective national activity about cooperation and conflicts and examine how social perception on conflicts and terrorism in Nigeria has been changing.
Data and methods
Data
Considering that the violence led by the Boko Haram insurgency in Nigeria escalated dramatically in 2014 with over 10,000 deaths, while Boko Haram drastically expanded its territories, we use conflict events and fatalities data from January 2013 to December 2023 in Nigeria recorded by ACLED, which is a disaggregated data collection, analysis, and crisis mapping project, providing high-quality conflict event data. The data contains 31 fields, including event code, event date, event type, actors, latitude and longitude of the conflict event, notes detailing the nature of the event, and the number of fatalities, among others. The data have been widely used by journalists and academics and can be downloaded at https://acleddata.com/.
ACLED classifies conflicts into six types: battles, explosions/remote violence, riots, violence against civilians, strategic development, and protests. ACLED defines a Battles event as a violent interaction between two organized armed groups at a particular time and location. The battle event type may include ground clashes between different armed groups, ground clashes between armed groups supported by artillery fire or airstrikes, ambushes of on-duty soldiers or armed militants, exchanges of artillery fire, ground attacks against military or militant positions, air attacks where ground forces can effectively fire on the aircraft, and air-to-air combat. explosions/remote violence events are defined as incidents in which one side uses weapon types that, by their nature, are at range and widely destructive. The weapons used in explosions/remote violence events are explosive devices, including but not limited to bombs, grenades, improvised explosive devices, artillery fire or shelling, missile attacks, air or drone strikes, and other widely destructive heavy weapons or chemical weapons. Riots are violent events where demonstrators or mobs of three or more engage in violent or destructive acts, including but not limited to physical fights, rock throwing, property destruction, etc. Violence against civilians in ACLED is defined as violent events where an organized armed group inflicts violence upon unarmed non-combatants. Strategic development refers to the government’s repression of civilians, such as arrests, burning of houses, and looting of property, while protests refer to non-violent demonstrations, involving typically unorganized action by members of society. The data for the last two categories is much fewer than the other four types and, therefore, will not be analyzed here.
GDELT is another big media database used here. It includes more than 700 million distinct events across all countries, during the period from 1979 to the present, covering 20 categories and over 290 subcategories. GDELT events are drawn from a wide variety of news media, both in English and non-English, from across the world, ranging from local to international sources in nearly every country. It is based on the Conflict and Mediation Event Observations event coding ontology (Beieler et al. 2016; GDELT, 2020). Each event has two actors (Actor 1 and Actor 2), such as Country A and Country B. One of the most important and interesting attributes of the GDELT event data is that each event is assigned a set of attributes, including the conflict-cooperation scale value, called the Goldstein Scale (Goldstein, 1992), which is in the range of −10 and 10 and quantifies the degree of conflict or cooperation between the two actors of the event. GDELT has been used in several interesting research on crisis, disaster, risks, and peace (Keneshloo et al. 2014; Kwak and An, 2014; Qadir et al. 2016; Sun et al. 2021; Voukelatou et al. 2020). It can be downloaded from https://www.gdeltproject.org/.
Moreover, we also use the democracy index and the corruption perception index (CPI) as potential factors to analyze armed conflicts and terrorism in Nigeria. The democracy index published by the Economist Group is an index measuring the quality of democracy across the world. The index data are downloaded from https://www.eiu.com/n/campaigns/democracy-index-xxxx/, where xxxx is from 2016 to 2022. The CPI is published annually by the non-governmental organization Transparency International (TI) since 1995. The index value is on a scale from 100 (very clean) to 0 (very corrupt). The data are downloaded from https://www.transparency.org/en/cpi/xxxx, where xxxx is from 2016 to 2022.
Lognormal distribution
To characterize the daily number of events and fatalities of armed conflicts and terrorism, we will employ the lognormal distribution. In probability theory, a lognormal distribution is a continuous probability distribution of a random variable whose logarithm is normally distributed. Thus, if the random variable X is lognormally distributed, then Y = ln(X) has a normal distribution. Equivalently, if Y has a normal distribution, then the exponential function of Y, X = exp(Y), has a lognormal distribution. To examine whether the daily number of events and fatalities of armed conflicts and terrorism are lognormally distributed, we will take the logarithm of the non-zero data and check whether the distribution is normally distributed (i.e., bell-shaped).
Shannon entropy
A good approach to characterize the spreading of armed conflicts and terrorism across space in a country is to employ Shannon entropy, which takes the form of
$${ENT}=-\!\mathop{\sum }\limits_{k=1}^{N}{p}_{k}{\log }_{2}{p}_{k}$$
(1)
where pk is the probability of cell k being hit by armed conflicts and terrorism. It is computed as follows. First, a country is partitioned into cells (or regions) of size 0.5° × 0.5°. For a chosen period, such as one year, pk is obtained as the ratio between the total number of events (or fatalities) the cell k hit by armed conflicts and terrorism and the total number of events (or fatalities) the country hit by armed conflicts and terrorism. When a cell is free of armed conflicts and terrorism, its probability is then 0, and the term pklog2pk can be dropped. Therefore, the parameter N in Eq. 5 is the total number of cells of size 0.5° × 0.5° hit by at least one event (or one fatality, depending on whether the number of events or fatalities is focused on) of armed conflicts and terrorism. Given the total number of armed conflicts and terrorism, Shannon entropy becomes larger when armed conflicts and terrorism spreads more and more across a country.
Note that when a time series is not simply normally distributed, the spread of the data around the mean often may not be sufficiently well characterized by the variance of the data. In such situations, using Shannon entropy instead of the variance can be advantageous. This rationale motivates us to also compute Shannon entropy for the time series of the total fatalities. Concretely, the logarithm for the non-zero fatalities resulting from any armed conflict and terrorism in a month is taken, and then all such data are used to compute Shannon entropy for that month. To make comparison over different months feasible, a common set of intervals of equal length is used for all the logarithm of fatalities data.
Revealed comparative wealth (RCW)
In the study of armed conflicts and terrorism, economic condition is always an important factor to consider. In this work, we will use RCW, as it has overcome the problem of nonstationarity in GDP and GDP per capita time series, thus facilitating comparison over time and across different countries.
RCW is defined as (Gao et al. 2024; Wang et al. 2024)
$${RCW}\left(k,t\right)=\frac{\frac{W\left(k,t\right)}{{popul}\left(k,t\right)}}{\frac{{\sum }_{k=1}^{{Nt}}W\left(k,t\right)}{{\sum }_{k=1}^{{Nt}}{popul}\left(k,t\right)}}=\frac{\frac{W\left(k,t\right)}{{\sum }_{k=1}^{{Nt}}W\left(k,t\right)}}{\frac{{popul}\left(k,t\right)}{{\sum }_{k=1}^{{Nt}}{popul}\left(k,t\right)}}$$
(2)
where W (k, t) and popul(k, t) are the wealth and population of country k at time t, respectively, and Nt is the number of countries in the world at time t that have data. Following the argument of Gao et al (2024), W (k, t) is better approximated by the nominal GDP of a country. The importance of RCW lies in two aspects. First, when an individual in a country is concerned, an RCW much larger than the world average of 1 means a large number of the citizens in the country will be able to pursue arts, sciences, creative works, etc., instead of just struggling to make a living. Such activities are critical for the development of the economy in the long run. Second, when the state is concerned, a large RCW means the state will be able to provide a better welfare system, have a cleaner and less corrupt political system, have an economy that is able to better withstand external shocks, and be able to attract better talents in science and technology from abroad—in short, the society will very likely be more harmonious. On the other hand, if RCW in the country experiences a sharp drop in a short period of time, then many kinds of social discord, even social unrest and conflicts may ensue.
Statistical analysis
In statistics, the coefficient of determination, denoted as R2, measures the fitness of a model to the data under study. When a linear model is concerned, given two data sets {xi, i = 1, 2, …, n} and {yi, i = 1, 2, …, n}, with sample mean \(\overline{x}={\sum }_{i=1}^{n}{x}_{i}/n\) and \(\overline{{\rm{y}}}={\sum }_{i=1}^{n}{y}_{i}/n\), R2 measures the percentage of variation in y explained by x, or vice versa. In this case, R2 is simply the square of the Pearson correlation coefficient computed by the following formula
$$r=\frac{{\sum }_{i=1}^{n}({x}_{i}-\overline{x})({y}_{i}-\overline{y})}{\sqrt{{\sum }_{i=1}^{n}{({x}_{i}-\overline{x})}^{2}}\sqrt{{\sum }_{i=1}^{n}{({y}_{i}-\overline{y})}^{2}}}$$
(3)
Result
Temporal evolution of armed conflicts and terrorism in Nigeria
Figure 1 shows the fatality time series of the four different types, Battles, explosions/remote violence, Riots, and Violence against civilians. We observe that during 2014 and 2015, the fatalities caused by armed conflicts and terrorism in Nigeria were very significant, especially in the three types including Battles, Explosion/Remote violence, and Violence against civilians. In the Battles type, the highest number of deaths was on 1 February 2015, which exceeded 500. Besides, the 2015 Baga massacre, which was a series of mass killings carried out by the Boko Haram terrorist group in the north-eastern Nigerian town of Baga and its environs, in the state of Borno, between 3 January and 7 January 2015, 800 were killed on January 7 alone. Compared to these three types, the death toll from riots was much lower. Since 2016, the deaths from violence against civilians have greatly decreased.
a Battles, b Explosions/Remote violence, c Riots, and d Violence against civilians (Original data from ACLED).
As pioneered by Richardson (1944), war fatalities have often been modeled by power-law distributions (Friedman, 2015; Picoli et al. 2017; Spagat et al. 2020). While these 10+ year fatalities data do suggest heavy-tailed distributions (i.e., the tails of the distributions exhibit power-laws), unfortunately, we find that such a modeling is not very useful for uncovering the temporal evolutionary patterns of armed conflicts and terrorism in Nigeria. To better characterize the temporal evolution, we have computed the monthly mean and standard deviation of the fatalities resulting from any type of events of armed conflicts and terrorism. The result is shown in Fig. 2. We find that the mean is large in 2014 and 2015, and has been slowly but steadily decreasing since then. Before 2018, the plot for the standard deviation shows a similar pattern. However, after 2018, the values of the standard deviation do not change much (more precisely, fluctuate more after 2022). Figure 3 shows the monthly total number of events and fatalities in Nigeria. We observe that the monthly total number of armed conflicts and terrorist events has been steadily increasing from 2013 to the end of 2020. Then it peaked and has remained at a high level. A similar pattern is observed for the monthly total number of fatalities since 2016. Before that, however, the patterns for the total numbers and fatalities are quite different. This is because, in 2014 and 2015, a few armed conflicts and terrorist events caused large fatalities.
a monthly mean and b monthly standard deviation. (Original data from ACLED).
a monthly total number of conflict incidents and b monthly total number of fatalities in conflict events. (original data from ACLED).
We have also computed the probability density functions (PDFs) for the logarithm of the fatalities resulting from all different types of armed conflicts and terrorism for any given year. The temporal variation of such PDFs is shown in Fig. 4. We observe that the functional shapes of these PDFs are consistent with normal distributions. Therefore, the fatalities data in any given year largely follow a lognormal distribution. We also observe that since 2016, the PDF curves have slowly but steadily moved toward the y axis. This is consistent with what we have observed in Fig. 2.
Shown are kernal probability density functions (PDFs) in different years marked by different colors (Original data from ACLED).
To characterize the temporal evolutionary pattern of armed conflicts and terrorism in Nigeria, we have computed Shannon entropy for the fatalities resulting from any type of armed conflict and terrorism in a given month. The monthly variation of Shannon entropy is shown in Fig. 5. We observe that since 2016, Shannon entropy has been steadily decreasing, especially after COVID-19. Together with the total monthly number of events and fatalities of armed conflict and terrorism shown in Fig. 3, we can conclude that temporally, armed conflicts and terrorism have become more uniform characterized by a decreasing Shannon entropy together with a smaller daily mean of conflict events and fatality.
The grey and red curves represent monthly Shannon entropy time series and its trend, respectively (Original data from ACLED).
Spatial evolution of armed conflicts and terrorism in Nigeria
To illustrate the spatial evolution of armed conflicts and terrorism in Nigeria, we have chosen three years, 2014, 2016, and 2021, to represent three phases: (a) a phase of severe armed conflict and terrorism with very large fatalities, (b) a phase with not very active armed conflict and terrorism and thus not very large fatalities, and (c) post-COVID-19.
Figures 6–8 show the maps for the fatalities of 4 different types in Nigeria in these three years, where violence in the subplot (d) of the three plots denotes violence against civilians. We observe that in 2014, battles were widespread, with the Northeastern region of Nigeria being the most severe. Explosions/remote violence and riots were relatively few. However, violence against civilians was very serious, especially in the northeastern region and along the middle north-south line tilting to the west. In 2016, battles were more concentrated in the Northeastern region of Nigeria, while riots were scattered out in the southern and southwestern regions of Nigeria, where the major political and economic activities of Nigeria occurred. Violence against civilians was even more widespread, encompassing the Northeastern, southern, and southwestern regions of Nigeria. There were 1408 events and 4896 fatalities recorded in 2016. The maps were completely changed by 2021. We observe that battles and violence against civilians were all over Nigeria. Now the number of events recorded reached 4546, more than 3 times that of in 2016, and fatalities reached 10,880, more than twice that of 2016.
Shown are the maps for four different types: a Battles, b Explosions, c Riots, and d Violence (Original data from ACLED).
Shown are maps for fatalities of four different types: a Battles, b Explosions, c Riots, and d Violence (Original data from ACLED).
Shown are maps of four different types: a Battles, b Explosions, c Riots, and d Violence (Original data from ACLED).
Let us now focus on the characterization of the spatial spreading of armed conflicts and terrorism in Nigeria by using Shannon entropy defined by Eq. (5). To illustrate the change in the distribution of fatalities across Nigeria in different years, we have chosen to compare two years, 2016 and 2021, by ranking the computed probabilities for cells (or regions) of size 0.5° × 0.5° in descending order, and plotted them in Fig. 9. In the plot, the black and the grey curves are for 2016 and 2021, respectively. Consistent with our earlier observation, the number of regions hit by armed conflicts and terrorism in Nigeria is much larger for 2021 than for 2016, since the number of regions hit by armed conflicts and terrorism in 2016 was only a few hundred, while that for 2021 went up to almost 1000. Also, note that the first few probabilities for battles and violence against civilians were larger in 2016 than in 2021. That means a few battles as well as a few cases of violence against civilians were quite large in 2016. The Shannon entropies computed based on the number of events and fatalities are shown in Figs. 10 and 11, respectively. Irrespective of which data are used, events or fatalities, we observe that Shannon entropies have increased considerably about half a year after the COVID-19 pandemic started, especially for battles and violence against civilians. Therefore, Shannon entropy well captures the fast spreading of armed conflicts and terrorism in Nigeria due to the COVID-19 pandemic. Note also that Shannon entropies for battles and violence against civilians are larger than those for explosions/remote violence and riots. This observation is fully consistent with the maps shown in Figs. 6–8.
Shown are ranked probability sequence constructed from the maps shown in Figs. 7 and 8 for fatalities of four different types: a Battles, b Explosions, c Riots, and d Violence (Original data from ACLED).
Shown are monthly variations of the Shannon entropy for the recorded events of four different types: a Battles, b Explosions, c Riots, and d Violence (Original data from ACLED).
Shown are monthly variations of the Shannon entropy for the fatalities of four different kinds: a Battles, b Explosions, c Riots, and d Violence (Original data from ACLED).
Proxy of the evolution of societal perception on armed conflicts and terrorism
Media big data contain a lot of important information about the societal perception of armed conflicts and terrorism. When events of armed conflicts and terrorism are rare, they are immediately covered by the media whenever they occur. However, when armed conflicts and terrorism become too frequent, then only those severe events, such as large fatalities, would be seriously reported by the media. In some sense, this reflects the adaptation of the general public to armed conflicts and terrorism. When this happens, one may say the country has been in a state of new normal.
Based on the above rationale, we have employed GDELT to construct a few indices reflecting a country’s general activity. We start with the bilateral relation (BLR) index.
Consider two countries such as China and Japan. To make the proposed BLR index applicable to any pair of two countries, we first have to notice that the number of events of cooperation or conflict cannot represent bilateral political relations since the number of reported news events has been increasing rapidly over time. To overcome this problem, we use the proportion of the number of events of cooperation or conflict,
$$\frac{{{\boldsymbol{N}}}_{{ij}}^{{\boldsymbol{(}}{\boldsymbol{+}}{\boldsymbol{)}}}}{{{\boldsymbol{N}}}_{{ij}}^{{\boldsymbol{(}}{\boldsymbol{+}}{\boldsymbol{)}}}{\boldsymbol{+}}{{\boldsymbol{N}}}_{{ij}}^{{\boldsymbol{(}}{\boldsymbol{-}}{\boldsymbol{)}}}}{\boldsymbol{,}}\,\frac{{{\boldsymbol{N}}}_{{ij}}^{{\boldsymbol{(}}{\boldsymbol{-}}{\boldsymbol{)}}}}{{{\boldsymbol{N}}}_{{ij}}^{{\boldsymbol{(}}{\boldsymbol{+}}{\boldsymbol{)}}}{\boldsymbol{+}}{{\boldsymbol{N}}}_{{ij}}^{{\boldsymbol{(}}{\boldsymbol{-}}{\boldsymbol{)}}}}$$
in a given time period as a measure of cooperation or conflict, where \({N}_{{ij}}^{(+)}\) and \({N}_{{ij}}^{(-)}\) denote respectively, the number of events of cooperation and conflict between country i and country j (i ≠ j, i being the active actor while j the passive actor).
The above reasoning has treated all the events as the same. This can be improved, by multiplying the number of events by the average Goldstein scale:
$${R}_{\text{ij}}^{(\pm )}=\frac{{N}_{{ij}}^{(\pm )}* \left|{\overline{{GS}}}_{{ij}}^{(\pm )}\right|}{{N}_{{ij}}^{(+)}* {\overline{{GS}}}_{{ij}}^{(+)}+{N}_{{ij}}^{(-)}* \left|{\overline{{GS}}}_{{ij}}^{(-)}\right|}=\frac{\left|{\sum }_{k=1}^{{N}_{{ij}}^{(\pm )}}G{S}_{k}^{({ij},\pm )}\right|}{{\sum }_{k=1}^{{N}_{{ij}}^{(+)}}G{S}_{k}^{({ij},+)}+\left|{\sum }_{k=1}^{{N}_{{ij}}^{(-)}}G{S}_{k}^{({ij},-)}\right|}$$
where \({\overline{GS}}_{{\rm{ij}}}^{(+)}=\tfrac{{\sum }_{k=1}^{{N}_{ij}^{(+)}}G{S}_{k}^{(ij,+)}}{{N}_{ij}^{(+)}},{\overline{GS}}_{{\rm{ij}}}^{(-)}=\tfrac{{\sum }_{k=1}^{{N}_{ij}^{(-)}}G{S}_{k}^{(ij,-)}}{{N}_{ij}^{(-)}}\) denote average, and \(G{S}_{{\rm{k}}}^{({\rm{ij}},+)},G{S}_{{\rm{k}}}^{({\rm{ij}},-)}\) represent the degree of cooperation or conflict represented by the event k for the two countries. We emphasize that the superscripts ± in Eq. 8 should be matched (i.e., when + is chosen on the left side of Eq. 8, then + should also be chosen in the numerator of the right; similarly for the sign −).
Next, we introduce a weight \({W}_{\text{ij}}^{(\pm )}\) to measure the importance of country j among country i’s international relations (in other words, the index i is fixed). Again, this is achieved by utilizing Goldstein scales:
$${W}_{\text{ij}}^{(\pm )}=\frac{{\sum }_{k=1}^{{N}_{{ij}}^{(\pm )}}G{S}_{k}^{({ij},\pm )}}{{\sum }_{l=1,l\ne i}^{{N}_{c}}{\sum }_{k=1}^{{N}_{{ij}}^{(\pm )}}\left|G{S}_{k}^{({il},\pm )}\right|}$$
where Nc denotes the number of countries, and the superscripts ± on the left and the numerator of the right of Eq. 9 should be matched. Note that \({W}_{\text{ij}}^{(+)}\) is positive, and can be considered as a probability. However, \({W}_{\text{ij}}^{(-)}\) is negative. Nevertheless, after being taken absolute value, it can also be considered as a probability. These probabilities represent, respectively, the importance of country j’s cooperation or conflicts with country i among i’s overall international interactions.
Synthesizing the above considerations, we arrive at the positive and negative Bilateral Relationship indices \({BL}{R}_{{ij}}^{(\pm )}\):
$${BL}{R}_{\text{ij}}^{(\!\pm\! )}={W}_{\text{ij}}^{(\!\pm\! )}* {R}_{\text{ij}}^{(\!\pm\! )}=\frac{{\sum }_{k=1}^{{N}_{{ij}}^{(\!\pm\! )}}G{S}_{k}^{({ij},\!\pm\! )}}{{\sum }_{l=1,l\ne i}^{{N}_{c}}{\sum }_{k=1}^{{N}_{{il}}^{(\!\pm\! )}}\left|G{S}_{k}^{({il},\!\pm\! )}\right|}* \frac{\left|{\sum }_{k=1}^{{N}_{{ij}}^{(\!\pm\! )}}G{S}_{k}^{({ij},\!\pm\! )}\right|}{{\sum }_{k=1}^{{N}_{{ij}}^{(+)}}G{S}_{k}^{({ij},+)}+\left|{\sum }_{k=1}^{{N}_{{ij}}^{(-)}}G{S}_{k}^{({ij},-)}\right|}$$
(4)
where the superscripts ± in the left and numerator of the right side of Eq. 10 should be matched.
We further extend the formulation of BLR to a global scale to measure the positive and negative effects of a country in the world. We call this index the National Activity Index (NAI):
$${NA}{I}_{\text{i}}^{(\pm )}=\frac{{\sum }_{j=1}^{{N}_{c}}{\sum }_{k=1}^{{N}_{{ij}}^{(\pm )}}G{S}_{k}^{({ij},\pm )}}{{\sum }_{i=1}^{{N}_{c}}{\sum }_{j=1}^{{N}_{c}}{\sum }_{k=1}^{{N}_{{ij}}^{(\pm )}}G{S}_{k}^{({ij},\pm )}}* \frac{\left|{\sum }_{j=1}^{{N}_{c}}{\sum }_{k=1}^{{N}_{{ij}}^{(\pm )}}G{S}_{k}^{({ij},\pm )}\right|}{{\sum }_{j=1}^{{N}_{c}}{\sum }_{k=1}^{{N}_{{ij}}^{(+)}}G{S}_{k}^{({ij},+)}+\left|{\sum }_{j=1}^{{N}_{c}}{\sum }_{k=1}^{{N}_{{ij}}^{(-)}}G{S}_{k}^{({ij},-)}\right|}$$
(5)
The computed NAI indices for Nigeria are shown in Fig. 12, where the thick blue and red curves are the trend curves determined from the light green and black curves representing daily cooperative and conflicting behaviors in Nigeria, respectively. The determination of the trend curve is achieved by utilizing an adaptive filtering algorithm, which has been shown to be close to optimal (Gao et al. 2009, 2011). It is observed that both the blue and the red curves show an overall upward trend from 2012 onwards. Overall, this reflects increased armed conflicts and terrorism in Nigeria. It should be emphasized that besides the overall increase in the red curve, there are also short-term large increases in NAI−, including in early 2012, 2014–2015, and 2021–2022. This is due to high-impact conflicting events that had received a lot of attention from domestic as well as international news media in the time frame identified. To illustrate, we can focus on the period following the COVID-19 outbreak. Notable events include: (a) Starting from 2020-3-30, Nigeria entered into a two-week lockdown to cope with the COVID-19 pandemic, (b) On 2020-10-21 and 22, shots were fired at Nigerian protesters, (c) On 2021-1-24, Kidnapping of an orphanage was reported, (d) On 2021-4-20, kidnapping of students at Nigeria’s Greenfield University was reported, and (e) In 2021-5, the Islamic State West Africa Province launched an invasion of the Sambisa Forest in Borno State, Nigeria.
Their trend signals are also shown as blue and red curves (Original data from GDELT).
While overall, the thick red curve in Fig. 12 has an upward trend, it should be noted that starting from around the second half of 2021, the red curve has started to decrease considerably. Since armed conflicts and terrorism have not been decreasing in any sense, as we have discussed earlier, we have to conclude that Nigerian people and media have accepted the increased armed conflicts and terrorism as a new normal.
In order to gain more insights into how NAI− reflects societal perception of armed conflicts and terrorism, we have examined the Pearson correlation coefficients between the red trend signal shown in Fig. 12 and the trend signals of fatality time series of battles, explosions/remote violence, riots, and violence against civilians shown in Fig. 13. All these trend signals have a temporal resolution of half year. We find that NAI− and the trend signals of fatalities due to battles are highly positively correlated, with their correlation coefficient reaching 0.667, hence, R2 is as large as almost 0.45. The correlation between NAI− and the trend for the overall fatality time series is also positive and strong, with a correlation coefficient of ~0.634. However, NAI− and the other three trend signals shown in Fig. 13 are only weakly correlated or not correlated at all. Noticing that variations in NAI− are due to armed conflicts and terrorism of all kinds as well as other conflicting events, including international, with the R2 between the trends of NAI− and fatalities due to battles is already as large as 0.45, R2 for the correlations between the trends of NAI− and fatalities of other types have to be small or close to 0 since the summation of all R2 is 1. Overall, we can conclude that battles are covered in the media a lot more than other types of armed conflicts and terrorism. This is consistent with the general perception that battles have a wider-ranging impact than other types of armed conflicts and terrorism.
Shown are (a) NAI time series and its, and (b–e) trend signals of the fatality time series of four different types. (Original data downloads from GDELT (a) and ACLED (b–e).
Observability of causes of armed conflicts and terrorism
The many major causes of terrorism that have been identified by previous studies are not all observable. To have a more direct impact on the design of policies targeting to reduce armed conflicts and terrorism, here, we focus on the causes that are more readily observable. The causes explicitly mentioned by Speakers of the United Nations and quoted earlier include despair, poverty, hunger, lack of basic services, and unemployment. All these causes could be associated with the value of RCW we have mentioned earlier, in particular, the decrease of RCW. Other important factors of terrorism include corruption and lack of democracy. To examine how the economy, corruption, and democracy may be connected with terrorism, in Fig. 14, we have plotted the yearly time series of total fatality of armed conflicts and terrorism, RCW, democracy index, and CPI, starting from 2013. Since the large fatalities in 2014–2015 in Nigeria were autonomously instigated by the Boko Haram insurgency, not much can be said about the correlations or causalities among these variables. Therefore, to identify potential correlations or even causalities, we focus on recent years starting from 2016. In recent years, especially after COVID-19, clearly, we observe that a decrease in RCW, democracy, and CPI has been associated with large fatalities. In terms of percentage drop, the decrease in RCW is particularly large, since it decreases from around 0.3 in 2014 to ~0.17 in 2022. Therefore, GDP per capita dropped from ~0.3 of the world average level in 2014 to ~0.17 of the world average in 2022.
Shown are (a) yearly total fatalities of armed conflicts and terrorism, b RCW, c democracy index, and d corruption perception index (CPI) (original data used in (a–d) were from ACLED, the World Bank, the Economist Group, and Transparency International, respectively).
As these time series are too short to support formal regression analysis, we have plotted scatter plots between the total fatalities and events of armed conflicts and terrorism in Nigeria and RCW, democracy index, and CPI starting from 2016. The results are shown in Fig.15. We find that, indeed, there are strong positive correlations between the yearly total fatalities/events of armed conflicts and terrorism and the three factors examined here: RCW, democracy index, and CPI.
Shown are scatter plots between fatalities/events and RCW, Democracy Index, and CPI (original data source is the same as in Fig. 14).
It should be emphasized that COVID-19 has made the armed conflicts and terrorism in Nigeria more diffuse and diverse, with the total fatalities in 2021 and 2022 approaching the highest value, which happened in 2014. This has been well characterized by the large spatial and temporal Shannon entropies presented in Figs. 5, 10, and 11. The post-COVID-19 change in the spatial and temporal patterns of armed conflicts and terrorism in Nigeria may be attributed to the new strategies terrorist organizations have designed to better adapt to the challenges and exploit opportunities COVID-19 has created, including exploiting the disruption of counter-terrorist systems and utilizing the weakened security community to carry out smaller but many more conventional terrorist attacks.
Concluding discussions
Nigeria is one of the largest countries in Africa that has been severely affected by armed conflict and terrorism. To help African nations in general and Nigeria, in particular, to better tackle armed conflicts and terrorism, we have proposed to first systematically characterize the spatiotemporal evolution of armed conflicts and terrorism in Nigeria using the events and fatalities data of ACLED. We have found that spatially and temporally, armed conflicts and terrorism in Nigeria have become more widespread and uniform, respectively. While the former is characterized by an increasing Shannon entropy for the spatial distributions of conflicts and terrorism in recent years, the latter is characterized by a decreasing Shannon entropy for the fatalities time series together with a smaller daily mean of conflict fatality.
To delineate the most observable causes of armed conflicts and terrorism from the vast number of factors that have been identified by various scholars, we have resorted to a readily computable indicator, RCW, that quantifies the relative growth or decay of the Nigerian economy for the global average growth. RCW reflects well the degree of severity of conflicts and terrorism in Nigeria, especially, whenever RCW drops significantly, such as in 2014, 2015, and half a year after COVID-19 started, armed conflicts and terrorism rise considerably. Therefore, economic decay and associated decrease in citizen’s welfare are highly correlated with the intensification and spreading of armed conflicts and terrorism. Other important factors include democracy and corruption. While we can show using scatter plots that economic decay, decrease in democracy, and increase in corruption are strongly negatively correlated with the severity of armed conflicts and terrorism in Nigeria, unfortunately, due to the shortness of data points for RCW, democracy, and CPI, a formal regression analysis cannot be performed.
Finally, we have utilized another media big data, GDELT, to construct two general indices, NAI+ and NAI−, to describe the collective national activity pertaining to cooperative and conflicting behaviours, respectively. NAI− can be used as a proxy for social perception of conflicts and terrorism in Nigeria. NAI− has been changing over the years, especially has been increasing greatly about half a year since the COVID-19 pandemic started. This has to be attributed to the economic decay inflicted by the pandemic. This further corroborates that economic decay is a major factor responsible for the worsening of armed conflicts and terrorism in Nigeria. Associating NAI− with the general societal perception of armed conflicts and terrorism in Nigeria, we then have to conclude that the severe armed conflicts and terrorism after the COVID-19 pandemic in Nigeria have now been accepted as a new normal, since starting the second half of 2021, NAI− has been decreasing, while armed conflicts and terrorism have not been decreasing at all. Such a new normal has to be considered rather tragic, since the adverse effects of armed conflicts and terrorism on mental health are very severe, including inducing posttraumatic stress disorder, major depressive disorder, and anxiety disorders (Rigutto et al. 2021). Considering that prior research on the impact of terrorism on mental health took place in a Western, mainly American setting (Rigutto et al. 2021), where terrorism is far less serious than that in Nigeria and elsewhere in Africa, the actual effects of terrorism on mental health much be more severe and widespread in Nigeria and elsewhere in Africa than in western countries. This issue is very significant and warrants further research in the future. Here, we emphasize that this new normal will also exacerbate the economic crisis Nigeria is facing nowFootnote 4.
When the Speakers of the United Nations warned the Security Council that terrorism spreading across Africa at an alarming rate on March 28, 2023Footnote 5, greater support and enhanced international, and regional cooperation were called for. While this is of great importance, this alone will not suffice, however, since better economic development in Nigeria in particular and Africa in general cannot be achieved by such callings. Rather, there is a fundamental limit to the economic development in Africa and other developing and underdeveloped countries imposed by the global hierarchy recently characterized by Gao et al. (2024). To fundamentally address the armed conflicts and terrorism in Nigeria and other African nations, greater economic help from countries outside of Africa, especially from advanced economies, including upgrading the industrial chains in African nations, is imperative.
