Abstract
How does political violence affect public health risks? While past research provides useful insights, it remains unclear how the effect of violence on health risks varies by perpetrator identity, target, and the intensity of attacks. We argue that indiscriminate attacks by militants will discourage civilians from accessing healthcare services. In contrast, clashes between militants and governments are likely to induce state actors to augment healthcare provision, while reinforced state control may facilitate civilians’ access. We test our arguments by employing multilevel modeling using subnational data on political violence events within Pakistan between 2005 and 2011, and immunization rates for under-5 infectious diseases.
A series of attacks on polio health workers across three towns in Pakistan in April 2019 (Saifi and Shah 2019) exacerbated existing concerns about the country's ability to tackle one of its most pressing health threats—the transmission of Poliomyelitis (“polio”). Pakistan, in addition to Afghanistan, is one of the few countries still struggling to eradicate the infectious disease. The Tehrik-i-Taliban Pakistan, a lethal insurgent group in Pakistan, has targeted and killed numerous health workers and security personnel accompanying them (RFE/RL, 2019) over the past decade. Similar to Pakistan, Afghanistan also experienced a spike in polio cases in 2019, with the highest levels believed to be in areas most affected by militancy. Afghan officials claim that many households refuse to immunize their children owing to fear of being targeted by the Afghan Taliban (Mohammad and Habibzada 2020). Since 2001, there has been a considerable rise in the number of cases of polio in Afghanistan; the number of reported cases of polio increased from 11 in 2001 to 80 in 2011 (Hachiya et al. 2012).
Targeting of healthcare workers and the general pervasiveness of political violence and terrorism in both Pakistan and Afghanistan, along with the existence of infectious viruses like polio, raises critical questions about how and when domestic political violence affects civilian populations’ health security. While there are numerous definitions of “health security” in the literature, we define it as all activities required to minimize threats to people's health, which is closely linked to the World Health Organization's definition of global health security (see Aldis 2008).
Findings in political science and health studies frequently link armed conflict and terrorism to a variety of detrimental effects on populations’ health and access to resources. In particular, armed conflict is widely perceived to be a driver of reduced immunization coverage, which is linked to the risk of vaccine preventable disease transmission, illness, and mortality (Heudtlass et al., 2016). However, most of these studies analyze the effects of intense inter-state wars, and/or use aggregated measures of violence to show the negative health outcomes for countries. Several questions remain unanswered in this literature: (1) how do the effects of violence on health vary by perpetrator identity and target type; (2) how do different strands of violence affect the risk factors associated with negative health outcomes such as access to childhood vaccinations; and finally (3) how do these effects vary temporally and by intensity? Exploring these questions can contribute to a more nuanced understanding of the relationship between violence and health security within states, which is essential to devise and implement targeted health campaigns.
In the context of internal conflict within states, and using the case study of Pakistan, we show variation in the effects of political violence on civilian populations' health security (captured by access to vaccines for infectious diseases), by taking into account the identity of perpetrators of violence and their targets and measuring both the magnitude and frequency of attacks. We develop and test several hypotheses with regards to perpetrators of violence (state actor or militant group) and their target types (civilians, state or militant group). First, we expect that targeted killings/assassinations and attacks against civilian targets by non-state actors will intimidate civilians and undermine trust in a government's ability to provide security, which will have a negative impact on civilians’ access to healthcare services. Second, where state actors directly target civilian populations, this is also likely to generate fear and distrust towards the government and dissuade access to healthcare services. Finally, in contrast to the above, where state and non-state actors primarily target each other, this is likely to maintain or increase civilians’ access to healthcare. We offer two parallel logics for this argument: first, where militants target state actors to undermine government legitimacy, it is likely that governments will proactively reinforce healthcare service provision, amongst other public services, to protect their own legitimacy and signal strength, and mitigate health risks within the country; second, a state's attacks on non-state actors are likely to undermine the domestic threat, at least in the short-term, which can result in overall constrained militant activity—encouraging civilians to access healthcare.
Employing district-level repeated cross-sectional data from Pakistan and using multilevel logit models, we measure and compare the effects of violent attacks perpetrated by the state and by non-state actors on civilians’ access to three different types infectious disease immunizations between 2005 and 2011. Our study uses the Pakistan Social and Living Standards Measurement (PSLM) survey data for the years 2006, 2009 and 2011 to gain vaccination data for districts across Pakistan in conjunction with the BFRS Dataset of Political Violence (Bueno de Mesquita et al., 2015), which records incident-level data on political violence in Pakistan for the same time period at the district level.
Overall, our results show that targeted killings of specific individuals are the most likely to hinder civilian populations’ access to vaccinations, as are frequent attacks against civilians by state actors. In contrast, where state and non-state actors primarily target each other, this leads to increased immunization rates. This suggests that direct conflict between a state and non-state actors does not necessarily directly threaten the health security of civilian populations. On the contrary, it may compel governments to make concerted efforts to provide health services to its civilians to retain or enhance its legitimacy as well as mitigate health risks. Targeting of militant groups may also diminish the actual or perceived threat posed by militant groups, which can facilitate healthcare access.
Our findings contribute to an improved understanding of how conflict shapes health security, especially in the context of low-intensity conflict. It is generally argued that large-scale conflict and violence within states can have a debilitating effect on civilian health outcomes because wars often result in significant infrastructure damage. However, low-intensity conflicts within states do not often necessarily result in similar wide-scale destruction of healthcare facilities since the use of conventional war weapons is limited to quelling domestic threats. In this light, our findings suggest that the effects of violence on health security can vary by the nature and magnitude of political violence, and operate through various mechanisms.
This study offers several contributions. First, it is one of the few studies, to the best of our knowledge, to conduct a quantitative analysis of the links between internal political violence and vaccination access at a subnational level. Examining district-level access to vaccinations, as opposed to the national level, allows a much more nuanced understanding of how violence affects a population's health security. Second, the majority of studies that examine conflict and health outcomes examine the effects on the outbreak of infectious diseases rather than on the risk of a disease. By focusing on the direct effects of violence on civilians’ access to three different types of vaccinations, we are able to isolate characteristics that may be specific to some health campaigns (e.g. polio) in Pakistan, as well as other environmental factors that could have an influence on the incidence of a disease. Finally, this study is one of the first to assess the effects of different types of violence on health security, taking into account the perpetrators of violence as well as the intended targets. This allows us to identify the type of political violence that has the most harmful effect on a country's health security, and the type of violence that compels governments to take steps to mitigate the risk of disease amongst its population for political and/or socioeconomic reasons. Indeed, as our study suggests, the desire to win over the goodwill of local populations where governments are engaged with domestic militant groups may actually increase access to some health services, such as vaccinations, rather than decrease it.
This article begins with a discussion of past research and theories on the links between conflict and health security, followed by the specific context of Pakistan, which has been engaged in a low-intensity civil war since at least 2001. The next sections discuss our hypotheses, the data and empirical approach for the study, followed by a discussion of the main results and conclusion.
Political violence and health security
In studying the effects of political violence, scholars have started to emphasize the need to explore the broader consequences of conflict beyond just the direct effects. While researchers have long been attentive to the direct impacts of war on mortality and battle deaths, there has been a push to study the less apparent effects—such as the impacts of war on public health (e.g. Murray et al. 2002). In general, the growing research on the links between conflict and public health has produced an overall consensus that political violence is associated with the prevalence and increase in disease (Kennedy, McKee & King 2015; Guarino et al., 2017) and is also linked to a wide range of negative mental health outcomes such as post-traumatic stress disorder and depression (Barber 2008; De Jong et al., 2008, Haj-Yahia, 2008).
Frequently, researchers point out that violence and instability disrupt service delivery and cause migration and destruction of infrastructure that supports the well-being and health of civilians. Studies conducted around the world in the Middle East, South and Central Asia as well as Africa show that political violence can cause widespread damage to a country's public utility systems and roads and bridges (e.g. Coward 2008; Gregory, 2008, Hamieh and Mac Ginty, 2010; Ozaras et al., 2016). Such destruction can have severe consequences on public health outcomes such as increased infectious disease (Reilley et al., 2002; Gayer et al., 2007; Wise 2017) and can make vaccination campaigns difficult to implement (Guarino et al., 2017; Herp et al., 2003). Several cases illustrate these effects; for example in Uganda during the mid 1980s, measles, tetanus and diphtheria increased drastically and infant mortality rates more than doubled in areas affected by violence (Urdinola, 2004). Ibrahim et al. (1996) argue that civilian conflict in Somalia between 1987 and 1989 resulted in a rise in infant mortality rates. Moreover, these effects may be even more pronounced where armed groups deliberately target health centers and medical personnel (Acerra et al., 2009; Farmer 2004; Flores et al., 2009; Yusufzai 2008). In another study of Pakistan and Nigeria, Steelfisher et al. (2015) show that beyond the difficulty of reaching households in conflict-affected areas, vaccination programs also face the challenges of civilians’ limited awareness and negative attitudes.
While earlier studies in this realm uncovered negative links between civil war deaths and measures such as healthy life expectancy (HALE) 1 at birth in a population (e.g. Ghobarah et al., 2004), more recent research has focused on the links between conflict and major communicable diseases like polio. The reemergence of outbreaks of polio has frequently occurred in places that have been beset by conflict and insecurity. An important global study by Guarino et al. (2017) finds that increases in a country's indicators of violence and instability, as measured by indices such as the Fragile States Index, are associated with incidences of polio cases, even after controlling for other risk factors. The authors recognize that conflict itself does not cause polio, but instead it does so via its effects on more proximal causes of the disease, such as vaccination rates. While their study indicates that instability negatively affects vaccination for polio, their broad measures, such as the Fragile States Index, do not provide a nuanced understanding of the specific type of violence that affects polio vaccination. Further, as their study highlights, the precise ways in which conflict will interfere with program implementation and spread of disease will be dependent on specific local contexts and the large number of variables which constitute their conflict/stability indicators make it impossible to discern the most important contributing factors. In a study of 16 countries, Grundy and Biggs (2019) analyze the association between conflict, immunization coverage and vaccine-preventable disease outbreaks between 2010 and 2015. They find that while these countries only accounted for 12% of the global population, they accounted for 67 and 29% of global polio and measles cases, respectively. Overall, their study finds that poor security conditions, impaired health infrastructure and a lack of human resources resulted in low levels of outreach services and immunization drives, as well as slow vaccine introductions—all of which contributed to low coverage and disease outbreaks in conflict-affected areas.
In another cross-national study, Kennedy, McKee and King (2015) explore the relationship between Islamist and non-Islamist insurgencies and the global distribution of polio. Their analysis is limited to cases where an insurgency (conflict between a state and an armed non-state actor) resulted in at least 25 or more battle deaths per year. Their findings indicate that internal armed conflict by Islamist insurgents (as opposed to non-Islamist groups) explains the global distribution of polio, but only in the post 9/11 era. In another polio-related study, Norris et al. (2016) examine the spatial relationship between Improvised Explosive Device (IED) violence and polio incidence in Afghanistan during 2010, comparing the average number of IED detonations in polio high-risk districts with non-polio high-risk districts between 2004 and 2009. The authors find that districts that have a high-risk for polio have statistically significantly higher mean numbers of IEDs than non polio high-risk districts—indicating that violence has hindered the effort to eradicate polio. The authors suggest that the correlation of polio incidents is probably due to both health workers being directly targeted by armed groups and the prevalence of conflict decreasing public trust in government actors.
This study, along with others (e.g. O’Reilly et al., 2012), suggests that the underlying mechanism appears to be a decrease in vaccination coverage. However, two points remain unclear from this body of research. First, how do geographically sporadic attacks by armed groups impact vaccination access? Second, and perhaps more importantly, does violence perpetrated by the state and non-state actors have the same effect on vaccination access?
In addition to elucidating the effects of violence on health outcomes, studies have increasingly focused on understanding the underlying mechanisms by which conflict can affect healthcare access. As highlighted by Penchansky and Thomas (1981), there are multiple categories through which violence can impact healthcare utilization, including availability, accessibility, accommodation and affordability. Where violence results in the destruction of healthcare facilities, this may affect the availability and accessibility of healthcare (Chi et al., 2015; Giacaman et al., 2005; Guenther et al., 2012). The prevalence of conflict can affect the accommodation of healthcare facilities if they are unable to operate normally owing to unstable conditions (Chi et al., 2015), and it can also decrease the affordability of healthcare provision if provision in conflict regions entails high costs. In Nigeria, for example, the Boko haram insurgency has resulted in the destruction of health facilities and displacement of health workers, which has severely impeded the availability of healthcare services to the civilian population (Patel et al., 2017). Other work has shown how militant groups tend to target aid workers intentionally, which often includes healthcare workers employed in conflict zones (Narang and Stanton 2017). In the case of countries like Afghanistan, such targeting of health or aid workers can compel non-governmental organizations, such as the Medecins Sans Frontiers (MSF) to exit the country. A recent example from Afghanistan includes the case of MSF, when it withdrew from Afghanistan's Dasht-e-Barchi hospital in Kabul after a brutal attack by Islamic State Khorasan (Islamic State's South Asian branch) killed newborns and mothers in May 2020 (MSF, 2020). Such attacks are often strategically and politically motivated as they can undermine a state's overall ability to provide adequate healthcare to its citizens.
Along similar lines, Yach (1988) shows that outbreaks of conflict and violence in Cape Town in 1986 impacted the accommodation of services as healthcare facilities such as preventive clinics and sexually transmitted diseases clinic services were unable to function at full capacity during times of violence. Using demographic and violence data from 1990 to 2000 in Colombia, Urdinola (2004) finds that in areas where there was a presence of violence perpetrated by either paramilitary forces or guerillas, the populations had higher levels of infant mortality. Urdinola attributes this to the destruction of physical infrastructure, the lack of state security in violence-affected areas, and the control of territories by militant groups. Moreover, all of the above mechanisms are relevant where individuals or entire communities are displaced as they flee areas of violence (Loretti 1997).
In contrast to the above studies, which focus on showing the direct and indirect effects of violence on health, Hilsenrath (2005) argues that militant organizations may use the provision of services—such as healthcare—to increase their own legitimacy and political capital. Using the example of Hamas, Hilsenrath argues that the group is actually a key contributor to the health sector in the West Bank and the Gaza strip. This piece fits into a larger body of research, which examines how violent extremist organizations provide public services to build popular support bases (for e.g. Flanigan 2008; Grynkewich 2008; Mampilly 2015; Arjona 2014; Arjona, 2016; Huang, 2016; Stewart 2018). Arguably, the need to amplify popular support also exists for government actors, especially where they are combating domestic insurgent groups with potential ties to local communities; in such cases a concerted effort by governments to increase healthcare provision in times of conflict may theoretically sway popular support towards government actors. However, past literature indicates that large-scale political violence can drain government resources, and redirect funds towards the security sector and away from the provision of social services, including healthcare (Sidel and Levy, 2008; Hamieh and Mac Ginty 2010).
Relatedly though, some recent research on the “guns verses butter” claim has been called into question; for example, examining the case of military spending and social spending in Egypt, Ali (2011) fails to find a crowding-out effect on social spending because of increased military spending. Besides, large-scale reallocations of military spending are likely to compromise the public sector budget if a state is engaged in an inter-state war, rather than a civil war and even then, it is not clear the extent to which the health sector specifically is affected. In this regard, if a state actor's engagement with domestic militant groups does not necessarily require significant shifts in its budget, then there is no reason to assume that a government may not intentionally increase social services to increase its legitimacy and undermine the popular support of violent extremist organizations.
Testing the three mechanisms: Fear, accessibility, and legitimacy
The discussion outlined above about the effects of conflict on health outcomes generally suggests a negative relationship. While different case studies attribute the negative outcomes to various reasons, two primary mechanisms emerge for the negative relationship between violence and health. First, violence can have a negative impact on healthcare outcomes owing to the fear and public distrust it generates amongst civilian populations. On the one hand, attacks by violent extremist organizations can deter civilians from accessing publicly provided services, especially where medical workers and health facilities are targeted. On the other hand, the inability of the state to curtail violence can undermine the confidence that civilian populations have in governmental and non-governmental-organization-run health campaigns, as in the case of polio vaccination campaigns in Afghanistan (Norris et al., 2016). Both of these factors undermine the willingness of civilian populations to receive or access healthcare services.
Second, the general prevalence of violence within a state can trigger a state of insecurity, which can lead to problems of accessibility, accommodation and affordability of healthcare (e.g. Chi et al., 2015). Where this mechanism is applicable, the negative effect of violence is not necessarily due to the unwillingness of civilians to access healthcare, but rather the ability of the state to provide adequate healthcare. While the literature above highlights the above two mechanisms, it does not necessarily tell us whether and how the effects of political violence on health outcomes vary by perpetrator and their target types, and how the intensity and frequency of violence affects these outcomes.
In addition to the above two mechanisms, we posit that there is also a potential third mechanism whereby, counterintuitively, incidents of political violence may be linked to a positive effect on health outcomes. Where governments are engaged in low-intensity conflict against domestic armed groups, and there is not necessarily a “butter versus guns” trade-off, governments may increase their healthcare provision to enhance both their domestic and international legitimacy, and contain the health risks to their populations.
The case of Pakistan provides us with an opportunity to test the above mechanisms using detailed, district-level violence and immunization status data for three different vaccinations over the period 2005–2011. More importantly, it provides us with the opportunity to explore whether the expected negative relationship identified in the literature between violence and health security holds when we account for the perpetrator's identity, attack type, intensity and target. While the specific characteristics of health campaigns, population demographics, and violence are unique to Pakistan, with some health drives being more politically charged than others (such as the polio vaccination campaign in Pakistan), the dynamics between different types of violence and health risks travel well to other countries experiencing low-intensity conflict and terrorism. In general, we expect violence to influence both the incentives and opportunities of citizens to access healthcare (to include vaccination access), as well as governments’ incentives to expand healthcare services in all conflict environments. Particularly, as developing countries struggle with internal conflict and domestic armed groups, these insights provide a much more nuanced understanding of the associated effects on health security. Taking into account the perpetrator identity (state actor or militant group) and target type (civilians, state or militant group), we test several hypotheses on the relationship based on the following mechanisms: (1) violence generates public fear and distrust in government, and makes civilians unwilling to access healthcare services; (2) governments targeted by non-state actors increase provision of healthcare to enhance legitimacy and mitigate health risks; and (3) the targeting of non-state actors by governments constrains domestic threats, or at least the associated threat perceptions, which can create a more permissible environment for civilians to access healthcare and other public services. We expect that different mechanisms will be at play depending on the violent perpetrator and their target.
Our first set of hypotheses relates to targeted assassinations, and attacks against civilians by non-state actors. For these, we expect that the first mechanism will be in effect whereby violence triggers fear amongst civilians and undermines their trust in the government's ability to protect them. Specifically, the use of terror tactics against civilians is very much a mechanism intended to induce fear into populations to dissuade them from supporting initiatives of the government, or to cease collaboration with government actors (Kydd and Walter 2006). This argument also generally aligns with research that indicates that militant groups will often target aid workers with the goal of expelling such welfare work out of particular areas to undermine support for the government (Narang and Stanton 2017). By reducing civilians’ access to essential services, such groups also seek to dismantle the government's control over specific regions. Additionally, much of the literature on foreign aid and militant violence also lends itself to the logic of our argument here, which indicates that aid-funded welfare projects can attract strategic insurgent violence to sabotage efforts by the government to win over civilians (Crost et al., 2014; Wood and Sullivan, 2015; Sexton, 2016; Khanna and Zimmermann 2017). As such, we test the hypotheses that targeted assassinations as well as indiscriminate attacks by militants against civilians will result in civilian populations becoming fearful of accessing state-provided services such as health vaccinations.
In parallel, where state actors directly target civilian populations, this is likely to generate fear and distrust towards the government. But why would governments target their own civilians? Warring parties often turn to violence against civilians to dismantle territorial control of an opponent, disrupt resource flows, and undermine civilian support (e.g Kalyvas and Kocher 2007). Democratic and authoritarian regimes have both been known to counter domestic threats, such as insurgencies, via coercion of civilian populations perceived to be supportive of rebel groups (e.g. Davenport 2007; Hazelton 2017; McCormick and Mitchell 1997). In such cases, we expect that there will be a negative effect on health security owing to the coercion and intimidation of civilians by state actors. We test the following hypothesis:
Where violence within a state constitutes conflict between a government and non-state actor and it is not intentionally directed towards civilians, we argue that this is likely to maintain or increase civilians’ access to healthcare for two main reasons. First, where militants target state actors in an attempt to undermine their legitimacy, it is likely that the government will proactively reinforce healthcare service provision, amongst other public services, to increase its own legitimacy (or to simply signal strength), and to mitigate health risks within the country. This logic runs parallel to why aid channeled through governments, especially humanitarian assistance, is often used by state actors to deter opposition, as it allows them to signal strength and limit rebel mobilization (Findley, 2018). Similarly, others have argued that small-scale and targeted projects can even motivate citizens to comply with the government and provide information on counterinsurgents (Berman et al. 2013). As such, governments will have many incentives to increase healthcare, whether funded by external aid or otherwise, in the areas where they target militants.
Second, a state's attacks on militants are likely to undermine the domestic threat, at least in the short-term, as this can result in militants being killed or captured, and even push them into hiding, which can subsequently result in overall constrained militant activity. Reduced militant activity or a lower threat perception of militants may encourage civilians to access healthcare services, such as acquiring vaccinations, and also reinforce government presence in affected areas.
In sum, conflict between a government and non-state actors provides governments with strong incentives to augment health services in order to buy its population's goodwill and protect it from health risks, while a mitigated domestic threat may create a permissible environment, which allows civilians to access healthcare services with more ease. Alternatively, it may be argued that violent clashes within states may create problems of accessibility and accommodation, where extant violence can make access to healthcare challenging. However, such circumstances are more likely to arise in large-scale conflicts between opponents of equal strength, as opposed to in low-intensity domestic conflicts, where most non-state actors are militarily inferior relative to state actors and are generally unable to cause widespread physical destruction. As such, we do not expect the accessibility and accommodation problem to arise where violence is linked to clashes between a state and non-state actor. As such, based on the aforementioned causal mechanism, we test the following two hypotheses:
Political violence and health infrastructure in Pakistan
While Pakistan has been suffering from high levels of internal conflict and terrorism since the early 2000s, it was around 2007 when active resistance to polio campaigns by militant groups emerged (Kennedy, McKee and King 2015). While vaccinations can often become a political issue within countries (with the recent example of mandated COVID-19 vaccinations), in Pakistan, polio vaccinations have emerged as an especially politically charged topic. Resistance to the polio health campaign emerged in the Khyber Pakhtunkhwa Province of Pakistan, where the boycott of polio campaigns included an assassination of the governmental head of the campaign in Bajaur Agency (Kennedy, McKee and King 2015). In 2012, the Tehrik-i-Taliban in North Waziristan banned polio vaccinations in its area of control in 2012 (Gostin 2014). More recently in 2019, a pamphlet distributed by Etahad ul Mujahideen, a local militant group in North Waziristan, called on civilians to oppose the polio vaccination (FATA Research Center, 2019). Such resistance to polio vaccinations has been fueled by Islamist insurgents’ narrative that polio workers are conducting surveillance activity for the United States or the Pakistani army to identify targets (Callaway 2012; Mohammadi, 2012).
Besides the targeted campaign against polio vaccinations, Pakistan has experienced high levels of internal conflict for almost two decades. While active insurgencies in Pakistan, most notably the Tehrik-i-Taliban Pakistan, have been based in the northern parts of the country, large cities, such as Karachi, Lahore and Rawalpindi, are frequently targeted by myriad militant groups. Additionally, Baluchistan, Pakistan's least populated and developed province, has been home to a long-running local insurgency, and frequently experiences confrontations between the state and militants, and violence against civilians. While the Pakistani Army has conducted wide-ranging military operations in northwestern Pakistan and Baluchistan, its law enforcement and intelligence units engage in counterterrorism operations across the country. Although these operations have dismantled many of the militant groups networks in their traditional home bases, it has also resulted in the dispersion of militants carrying out lethal attacks across the country (Jadoon, 2019).
Amidst dealing with such internal violence, and a targeted campaign against polio vaccination by militant groups, Pakistan has struggled with several other grave health challenges. Infectious diseases in particular are among the leading causes of disability and mortality in Pakistan, and the country is generally plagued with poor health indicators with high mortality rates for both communicable and non-communicable diseases (Khalil, et al. 2017).
Despite poor health indicators, Pakistan has an extensive health infrastructure, which comprises both public and private healthcare facilities (Islam, 2018). Under the public health delivery system, preventive, promotive, curative, and rehabilitative services are provided to citizens through primary, secondary and tertiary healthcare facilities. Primary healthcare facilities include basic health units, government dispensaries, and rural health centers. Secondary healthcare facilities span taluka headquarters, district headquarters, and maternal and child health centers. Tertiary-level facilities include hospitals (Kumar & Bano, 2017). In 2010, following a major constitutional amendment, the Ministry of Health, alongside 16 other ministries, was devolved to the provinces (Zaidi et al., 2019), which made the provinces responsible for health policy formation, planning, and implementation (Khalid & Ashley, 2019).
While the country maintains an extensive health infrastructure, widely accessible quality healthcare is sparse. In terms of health expenditures, Pakistan spends only 2.7% of its GDP on health, which is much less in comparison to other South Asian countries such as India (3.9%), Sri Lanka (2.97%), and Nepal (6.15%) (The World Bank, 2019). Consequently, out of pocket expenditures are high. In 2015, Pakistan's out of pocket expenditure, as a percentage of overall health expenditure, was an astounding 67% approximately (The World Bank, 2019). There is also an urban bias in the provision of health services, with out-of-pocket expenditures being much higher for rural households as compared with urban households (Khalid & Sattar, 2016).
Vaccination services in particular are provided to citizens through the Expanded Program on Immunization (EPI), a program initiated in 1978 (World Health Organization, 2020). It is funded through public sector development plans, the National Immunization Support Project, and development partners such as the WHO, UNICEF, The Japan International Cooperation Agency, The United States Agency for International Development and GAVI, which provide either technical or financial assistance to support the program (Government of Pakistan, 2020). Following devolution, different levels of the government (federal, provincial and district) support the activities and operations of the EPI. The federal government is actively involved with development partners and also in procuring vaccines, syringes and safety boxes. At the provincial level, the EPI cells fall under respective health departments and are managed by provincial EPI managers, while staff at the district level manage the affairs of the districts. At the union council level, services are provided through fixed centers and vaccination sites (Government of Pakistan, 2020). The EPI has a network of 6000 fixed centers and mobile health workers (female health workers and health visitors, and vaccinators; Hasan, Bosan, and Bile 2010). Vaccination for infectious diseases such as measles, polio, and DPT is provided across both public and private health facilities. In addition, Pakistan has one of the largest door-to-door immunization campaigns against polio. Through this campaign, routine polio vaccination is provided to households (Akhter, 2020). Vaccination campaigns for other infectious diseases (e.g. measles) are also held from time to time (UNICEF, 2021).
However, despite the extensive network and external support, immunization goals have remained elusive (World Health Organization, 2020); Pakistan is among the top three countries with the highest number of unvaccinated and under-vaccinated children (World Health Organization, 2020). In 2018, immunization coverage for measles (MCV1 and MCV2), was 66 and 58%, respectively. For polio it was 83% while for DPT1 and DPT3, 2 coverage was 84 and 72%, respectively (World Health Organization, 2018). Compared with other infectious diseases such as measles and tuberculosis, as alluded to above, polio vaccinations are an especially politicized issue in the country. Barriers identified include religious extremism and global political interests (Ali et al., 2019). Many have argued that increased political instability has contributed to Pakistan's inability to eradicate polio infections, even though until 2005, the incidences of polio infection were steadily declining (Khalil et al., 2017). Given that polio vaccinations have evolved to be a highly politicized issue in Pakistan, especially owing to accusations of the campaign being linked to Western intelligence-gathering efforts, we expect that the effects of political violence on polio vaccination uptake will be the most pronounced. Our hypotheses, however, are not specific to any particular type of vaccination, but rather are theoretically applicable to other types of healthcare services. Regardless of the degree of politicization of each type of vaccination, we expect violence to influence both the incentives and opportunities of citizens to access healthcare (to include vaccination access), as well as governments’ incentives to expand healthcare services in conflict environments. As such, regardless of the type of vaccination, which can be affected by “the availability of funding, political prioritization of vaccination,” (Burchett et al., 2012) and “unfamiliarity with vaccine-preventable diseases, and lack of trust in corporations and public health agencies” (Salmon et al., 2015), we expect all types of vaccinations to have the same relationship with political violence, although the more politicized a vaccination is, the more pronounced the effects are likely to be.
Overall, Pakistan's poor socioeconomic conditions, political instability, and religious extremism are some of the many factors that have placed the health security of its civilian populations in a precarious situation. By examining the variation in Pakistan's political violence at the district level and health outcomes, we seek to gain a more nuanced understanding of the relationship between different types of political violence and local populations access to various vaccinations for infectious diseases.
Methods
To test our hypotheses based on the three mechanisms regarding the relationship between political violence and health outcomes, we assess the impact of various violent events on different types of childhood vaccinations, via a pooled, repeated cross-sectional research design. Our dataset includes the number and intensity of violent incidents, perpetrated by state and non-state actors, and child health vaccinations from 114 districts across the four provinces of Pakistan (Punjab, Sindh, Khyber Pukhtoonkhwa, and Baluchistan) for the years 2005, 2007, 2009, and 2011 (see the Online Appendix for a list of districts).
While health outcomes may be measured in a number of ways, such as the outbreak of a disease, we opted to use child vaccination rates as this allows us to assess how political violence directly affects the risk of the spread of certain disease outbreaks. Within the main manuscript, we only include the results of the key variables of interest, with the complete models available in the Online Appendix (see Tables 1–5). In addition to the models discussed in the main manuscript, we run several other model specifications for sensitivity analysis, controlling for additional socioeconomic controls as well as using random intercepts for the four provinces in the country (see the Online Appendix Tables B1–D5).
Dataset, models and variables
We construct our dataset by combing incident-level data on violence with household level data on childhood vaccinations. Incident-level data is drawn from the BFRS dataset of Political Violence in Pakistan (Bueno de Mesquita et al., 2015). The BFRS dataset reports incident-level data on political violence in Pakistan from 1988 to 2011, and is constructed on the basis of open-source media reports of attacks and killings across Pakistan. The data is limited to incidents of violence where the perpetrator has a political, economic, religious, or social goal, and the act of violence is premeditated. As such, the data includes terrorist attacks, assassinations by militant groups, and military operations. Data on childhood vaccinations are drawn from the PSLM. The PSLM survey is administered by the Government of Pakistan and collects socio-economic and demographic data on households, and individuals within those households, across districts in Pakistan (Pakistan Bureau of Statistics, 2020). We restrict our study time frame (2005, 2007, 2009, and 2011) based on the availability of corresponding data in both the BFRS dataset and PSLM surveys. We retain only those districts for which data is available in the PSLM for all four years. Our repeated cross-sectional estimation dataset has 130,518 children in 78,159 households nested within 99 districts across the four provinces of Pakistan. As this is repeated cross-sectional data, a unique set of households in districts is included across different waves. Given the hierarchical structure of our dataset, we estimate multilevel logit models. More specifically, multilevel models were used since our outcome variable (childhood immunization) is nested within higher-level units (districts), and this modeling technique helps in addressing the possibility of correlated errors (Kreft & Leeuw, 1998). We add a dummy variable for time to control for the effect of time across waves; however, we do not hypothesize specific effects of time on childhood immunization. We cluster our results at the household level and log our targeting variables to smooth out fluctuations. Because the same households do not complete the surveys in each wave, we are unable to track changes per household over time.
The dependent variable, childhood vaccinations, is assessed through three separate indicators: vaccinations for polio, vaccinations for measles, and vaccinations for DPT. 3 These binary variables take on a value of 1 if the child has received all of the vaccinations for the disease in question, and 0 otherwise. The use of three different vaccinations serves as a check on whether any observed effects may be rooted in factors that happen to be specific to a certain health campaign.
We have five main independent variables of interest, which we draw from the BFRS dataset. The first set of violence variables relate to incidents perpetrated by militant groups, which include the following: (1) targeted assassinations of specific individuals (civilian, government or military state target); (2) attacks against civilians; and (3) attacks against the state (which includes attacks against the military, police, paramilitary and intelligence personnel). The second set of violence-related variables includes incidents where the perpetrator is a state actor and the violence was directed against: (1) civilians and (2) militant groups (i.e. non-state combatants). Within these two sets of variables (state and non-state perpetrated violence), we construct a series of variables to capture the magnitude (number of incidents) and intensity (lethality) of each type of violent incident. Summary statistics for each variable are provided in Tables E1 and E2 in the Appendix. In sum, for each of the five variables, we calculate the following:
Total attacks—aggregate number of attacks (events) in a year. Total attacks (two-year average)—total number of attacks (events) during the previous year and current year. Total attacks (five year sum)—running sum of attacks in the current year and last four years. Total lethality—total fatalities in a year. Total lethality (two-year average)—total fatalities during the previous year and current year. Total lethality (five-year sum)—total fatalities in the current year and last four years.
We create the magnitude and lethality variables for all of our violence variables to assess whether the frequency and intensity of violence lead to different outcomes when it comes to health security—i.e. is it the repeated exposure of violence or its intensity that matters more when it comes to influencing civilians’ willingness or ability to access health vaccinations for their children? Additionally, by using a five-year cumulative variable, we are able to account for the temporal component of the effects of violence on health outcomes. For example, repeated incidents of violence by either a state or non-state actor over an extended period of time may be more likely to create fear or difficulties in accessing healthcare than a single year of heightened violence which may be seen an aberration. In addition to our key independent variables, we also include several controls in our models which may affect households access to vaccinations; these include the literacy of mothers (1 if literate, 0 otherwise) literacy of household head (1 if literate, 0 otherwise), age and gender of the child (1 if female, 0 otherwise), material of roof of house (base bamboo), distance from health clinic or hospital in minutes (base 0–14 min), and whether the household is in a rural or urban area (1 if rural, 0 otherwise). As data for district-level indicators in Pakistan is largely absent, we do not control for confounders at the district level. However, we address district-level variation by accounting for the nested data structure (using multilevel models).
To provide a visual overview of the distribution of vaccinations across Pakistan's districts, and violence, we mapped the number of vaccinated and unvaccinated individuals for polio per district (according to the PSLM data), for the year 2005 (see Figures 1 & 2). In Figures 3 and 4, we map the total numbers killed in a five-year period, via assassinations, and indiscriminate violence by militants (Figures 3 and 4). While this study does not conduct spatial analysis, overall, a high-level view of the distribution of individuals surveyed shows that specific districts which had high levels of lethality owing to assassinations (over the previous five years) also had higher numbers of individuals who were either not vaccinated at all or received less than three doses (primarily in the northeast and southeast regions of Pakistan). The total number of individuals who received all three doses (Figure 2) can be observed in many districts where there was minimal violence by militants in the form of either assassinations or indiscriminate violence.
Total number of individuals unvaccinated for polio in 2005 (all three doses).
Total number of individuals vaccinated for polio in 2005 (all three doses).
Number of individuals killed in assassinations by militants (2005: five-year sum).
Number of individuals killed in indiscriminate attacks by militants (2005: five-year sum).
Results and discussion
Assassinations by non-state actors
Table 1 reports the results of the relationship between vaccinations and targeted assassinations by militants for polio and DPT. Overall, we find strong support for hypothesis 1(a). We find that households are less likely to get their children vaccinated for polio as the total number of assassination attacks by militants increase in a year (model 1), suggesting that targeted assassinations may trigger fear in the population in the short term. We observe the negative effect for both polio and DPT vaccinations as the total number of assassinations increases over a five-year period (Table 1, models 2 and 8), indicating the long-term effects of sustained targeted killings. This decrease in vaccinations for polio is visually illustrated in Figure 5: as the total number of assassinations over a five-year period increases from minimum to maximum, the predicted probability of childhood vaccinations decreases from 0.945 to 0.931. Interestingly, we do not find intensity of attacks (fatalities) to have any impact on polio vaccination, but the intensity variables do exert a negative effect on DPT vaccination (Table 1, models 10–12). Collectively, these findings provide support for the hypothesis that targeted assassinations by militants will be associated with lower levels of access to healthcare services as they can generate fear and deter civilians from getting their children vaccinated. The effects are also consistent for vaccinations for measles over a five-year period (see Table A1 in the Supplementary Materials).
Predicted probability of polio vaccinations and non-state assassination (total attacks—five-year sum).
Vaccinations and targeted assassinations by militants.
Standard errors in parentheses: *** p < 0.001, ** p < 0.01, * p < 0.05 (all variables are logged values).
Attacks against civilians by militants
While the results for the relationship between vaccinations and targeted assassinations by militants are in line with our hypothesis, we find mixed results for civilians targeted by militants as reported in Table 2. While we do not find any significant relationship between attacks against civilians and polio vaccination, for DPT vaccinations, we find that as the total number of attacks against civilians by militants increases over a five-year period, civilians are less likely to get their children vaccinated (Table 2, model 4). Contrary to expectations, we find the opposite effect for our short-term variables: total yearly and two-year average number of attacks, and two-year average number of killings against civilians have a positive effect on civilians’ vaccination for DPT (Table 2, models 1, 3 and 5). On average, the results are consistent for vaccinations for measles as well (see the Online Appendix Table A2). Thus, overall, we do not observe a consistent effect of attacks against civilians by militants across the vaccinations, and as such, do not find strong support for hypothesis 1(b).
Vaccinations and civilians targeted by militants.
Standard errors in parentheses: *** p < 0.001, ** p < 0.01, * p < 0.05 (all variables are logged values).
Attacks against civilians by state actors
Table 3 reports the results of the relationship between vaccinations and targeting of civilians by state actors. In general, we find that the effects of state perpetrated violence against civilians differ by magnitude and intensity. As the total number of attacks increases over a five-year period, civilians are less likely to get their children vaccinated for both polio and DPT (Table 3, models 3 and 9). For DPT, we see a negative effect even in the short term (Table 3, models 7 and 8). This suggests that distrust in the government may be triggered both over the short and long term as government attacks against civilians increase. Contrary to expectation though, we see effects in the opposite direction when it comes to the intensity of attacks. As the number of civilians killed by the state in a year increases (Table 3, model 10) and as the average number of civilians killed by the state over a two-year period increases (Table 3, model 5), civilians are more likely to get their children vaccinated for DPT and polio, respectively. The results are consistent for vaccinations for measles (see Table A3, Supplementary Materials). Overall, the results suggest that a higher number of attacks by the government may trigger distrust in the government, or even cause problems of accessibility, which lends support for hypothesis 1(c). However, the results also show that more intense attacks by the government in the short term may increase access to vaccinations. As mentioned earlier, governments often turn to one-sided violence (violence against civilians) as a tactic to dismantle territorial control of an opponent, disrupt resource flows, and undermine civilian support (e.g. Kalyvas and Kocher 2007). The short-term positive effects of attacks’ intensity on polio and DPT warrant a closer examination to understand the extent to which state actors may engage in compensatory behavior by increasing the provision of social services (health services in this case) when they use civilian repression as a tactic to quell domestic threats in the short run. In sum, although we do find some support for hypothesis 1(c), the mixed results suggest that the effects of state-perpetrated violence on health security are complex and need a more nuanced analysis via case study research.
Vaccinations and attacks against civilians by the state.
Standard errors in parentheses: *** p < 0.001, ** p < 0.01, * p < 0.05 (all variables are logged values).
Conflict between state and militant actors
Table 4 reports the results of the association between vaccinations and militants’ attacks on state actors. We find that as the number and intensity of attacks against the state by militants increase, civilians are more likely to get their children vaccinated for polio and DPT (Table 4, models 1–8, Figure 6). Referring to Figure 6, we find that as total lethality (five-year sum) by non-state actors against the state increases, from minimum to maximum, the predicted probability of childhood vaccinations for polio increases from 0.941 to 0.95. This finding provides support for hypothesis 2(a), and suggests that the state may intentionally step in to enhance its legitimacy through enhancing public services such as healthcare when it is attacked directly. The results are consistent for vaccinations for measles (see the Online Appendix, Table A4).
Predicted probability of vaccinations for polio and non-state actors targeting the state: total lethality (five-year sum).
Vaccinations and militants’ attacks against the state.
Standard errors in parentheses *** p < 0.001, ** p < 0.01, * p < 0.05 (All variables are logged values).
Table 5 reports the results of the relationship between vaccinations and state actors’ targeting of militants. We find that as the total number of attacks by state actors against militants increases in a year, civilians are more likely to get their children vaccinated for polio and DPT (Table 5, models 1 and 7). For all other variables related to attacks and killings of militants, civilians are more likely to get their children vaccinated for DPT (Table 5, models 8–12). The results are consistent for vaccinations for measles (see Table A5, Supplementary Materials), providing support for hypothesis 2(b). Collectively, these results strongly suggest that governments are more likely to expand healthcare service provision, amongst other public services, when they are engaged in a direct clash with domestic threats. This may potentially be due to a combination of two factors: first, states have incentives to increase such services to increase their own legitimacy in unstable conditions, and mitigate health risks faced by their populations; second, a state's targeting of militants extends the writ of the state in unstable regions, which creates an environment where civilians can access health services with more ease.
Vaccinations and state’s attacks against militants.
Standard errors in parentheses: *** p < 0.001, ** p < 0.01, * p < 0.05.
The results of Tables 4 and 5 in general test our hypotheses regarding conflict between the state and non-state actors, and allow us to test the hypotheses 2(a) and 2(b). Overall, the findings presented in Tables 4 and 5 show that rather than reducing healthcare provision owing to problems of accessibility and accommodation, direct conflict between the state and militants can result in increased health service provision. Of course, this is not to say that direct conflict between state and non-state actors may not create other negative health outcomes; our findings here directly relate to the provision of healthcare in the form of vaccinations, which is an important factor to consider to assess risk factors associated with diminished health security of civilian populations. In Table 6, we provide a summary of our results across all the different variables. Overall, the table shows that the models yield the most consistent results with regards to assassinations by non-state actors and clashes between the state and non-state actors.
Summary of the effects of violence on health outcomes (vaccinations for polio and DPT).
We conducted several robustness checks to test our results sensitivity. First, we included socio-economic control variables such as household earnings, sanitation facilities, and province to ensure that important/alternative intervening variables were also included in the model. These variables ensure that factors at the household and provincial levels are accounted for. Second, we estimated models that included a random intercept for provinces to understand residual variance in immunization within the provinces and if it might be driving the results. However, our results were consistent across these various specifications (see the Online Appendix Tables B1–D5).
Conclusion
Overall, this study furthers our understanding of the effects of political violence within states on civilian populations’ health security. The past literature on the topic overwhelmingly indicates a negative relationship between war and conflict, and health outcomes. However, our study suggests that the effects may be more nuanced depending on the perpetrator and target of violence, as well as the frequency and intensity of attacks. Moreover, our study examines how domestic political violence and terrorism affect the risk factors associated with health outcomes rather than overall outcomes themselves, which may be impacted by a number of other confounding factors.
In this study, we tested several hypotheses related to domestic political violence and health vaccinations in the context of Pakistan. Using insights from the past literature, we posited that targeted assassinations and attacks against civilians by militants or state actors will result in civilian populations becoming fearful of accessing state-provided healthcare services. With regards to direct conflict between state actors and militant groups, we argued that governments are likely to make a concerted effort to increase provision of health services to bolster governments’ legitimacy amongst civilian populations and protect them from health risks. Additionally state targeting of militant groups may also allow civilians to access healthcare with more ease.
Overall, the results indicate that where the engagements or attacks are between the state and government, there is a strong positive effect across the different types of vaccinations. This positive effect in both the short and long term is likely explained by governments’ desire to maintain their legitimacy, and to constrain the risk of disease in a conflict environment while also reducing militants’ ability to threaten healthcare access. This finding is contrary to what most literature has assumed is the primary link between political violence and human security owing to the focus on how large-scale violence can interrupt service provision linked to problems of accessibility and accommodation. We also find evidence that targeted assassinations by non-state actors have a strong negative effect on vaccinations. As can be expected, the most robust effects of assassinations are associated with their frequency—the higher the number of incidents of targeted killings, the more fearful civilian populations are likely to become. Finally, attacks against civilians by state or non-state actors yield mixed results, and we do not find any strong support for the hypothesis that terrorism or state repression against civilians in general is more or less likely to constrain civilians’ access to vaccinations. Overall, this suggests that the effects of low-intensity conflict can differ substantially from the effects of large-scale wars that can have a debilitating effect on civilian health outcomes owing to significant infrastructure damage. However, it is likely that the links between domestic terrorism, state repression, and health security are context-dependent, and that the effects vary temporally. Overall, this study is a first cut at disaggregating the complex relationship between health security and low-intensity violence, paving the way for further investigation via in-depth case studies, global analysis, and field experiments.
Supplemental Material
sj-pdf-1-cmp-10.1177_07388942221094006 - Supplemental material for Fear, accessibility, and legitimacy: An examination of the effects of political violence on health security in Pakistan
Supplemental material, sj-pdf-1-cmp-10.1177_07388942221094006 for Fear, accessibility, and legitimacy: An examination of the effects of political violence on health security in Pakistan by Amira Jadoon and Hina Khalid in Conflict Management and Peace Science
Footnotes
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Supplemental material
Supplemental material for this article is available online.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
