Malaria Prevalence and Civil Conflict Locations in Sub-Saharan Africa

• In Event: Climate and Health Challenges to Conflict Prevention

Abstract

As one of the most widespread and deadly pandemics, malaria has been considered a crucial factor that prolongs human and economic development in developing countries, most of which are located in the tropical and subtropical areas (Sachs & Malaney 2002). Previous studies argue that malaria can affect the duration of civil conflicts as a rough terrain factor that prevents effective counterinsurgency efforts (Bagozzi 2016). We theorize a non-linear relationship between civil conflict location and malaria prevalence, moderated by environmental and socio-economic factors.

To test the theory, we use spatially-explicit data to examine sub-national variations, going beyond existing empirical evidence based on country-level data. Specifically, we combine spatial-temporal data of conflict events, malaria distribution, and climate and land cover data in Sub-Saharan Africa. We measure conflicts with conflict event data from the Armed Conflict Location & Event Data Project (ACLED). Malaria severity is measured with data from the Malaria Atlas Project (Pfeffe et al 2018). To control for local geographic features, we use multiple spatial measurements of local climate and land cover/land use (e.g., forest, irrigated land) from multiple satellites, including TRMM, Landsat, MODIS, AVHRR, and SPOT. To control for local socioeconomic factors, we use grid night-light data from PRIO-GRID, population data from the Socioeconomic Data and Applications Center, and Gross Domestic Product (SEDAC) and Human Development Indices constructed by Kummu, Taka, and Guillaume (2018). To reveal regional patterns while maintaining local relevance, we aggregate the data into 0.5-degree grids (~55 km) following the common practice in continental analysis. This results in a dataset of 149,338 grid-year observations (repeated measures of each variable for 10,667 grids from 2000 to 2013).

We estimate varieties of GLM models. The outcome variable is the binary indicator of whether at least one incidence of conflict happens in the grid. The main explanatory variable is the Plasmodium falciparum parasite rate. Data of incidence rates and mortality rates are also used for as robustness checks. We control for various combinations of geographic and socioeconomic covariates. As a baseline, we use logistic regressions with two-way fixed effect. To account for the spatial-temporal autocorrelation, we fit the data to Bayesian spatio-temporal models (Finle, Benjerjee & Bradley 2007; Gelfand, Banerjee & Gamerman 2005).

Preliminary results with GLM with 2-way fixed effect indicates malaria prevalence and battle location have an inverted U shape correlation because of the nature of the malaria infection as well as the variations in state capacity. Militarized battles most likely occur in the middle infection interface where rebel forces and government presences came into direct clash with each other. Therefore, we can interpret the relationship between malaria infection and civil conflict as reflective of the state consolidation process as the state gradually expands its presence outside its stronghold in the urban areas. This paper thus makes a novel theoretical contribution to the relationship between tropical diseases and civil conflict processes.

References

Bagozzi, Benjamin E. “On Malaria and the Duration of Civil War.” The Journal of Conflict Resolution 60, no. 5 (2016): 813–39.

Finley, Andrew O., Sudipto Banerjee, and Bradley P. Carlin. "spBayes: an R package for univariate and multivariate hierarchical point-referenced spatial models." Journal of Statistical Software 19, no. 4 (2007).

Gelfand, Alan E., Sudipto Banerjee, and Dani Gamerman. "Spatial process modelling for univariate and multivariate dynamic spatial data." Environmetrics: The official journal of the International Environmetrics Society 16, no. 5 (2005): 465-479.

Kummu, Matti, Maija Taka, and Joseph HA Guillaume. "Gridded global datasets for gross domestic product and Human Development Index over 1990–2015." Scientific data 5, no. 1 (2018): 1-15.

Pfeffer, Daniel A., Timothy CD Lucas, Daniel May, Joseph Harris, Jennifer Rozier, Katherine A. Twohig, Ursula Dalrymple et al. "malariaAtlas: an R interface to global malariometric data hosted by the Malaria Atlas Project." Malaria journal 17, no. 1 (2018): 1-10.

Sachs, Jeffrey, and Pia Malaney. “The Economic and Social Burden of Malaria.” Nature 415, no. 6872 (2002): 680–85.

Authors

• Haohan Chen, New York University

• Enze Han, The University of Hong Kong