Magical Warfare Technologies and the Persistence of False Beliefs†

American Economic Review: Papers & Proceedings 2017, 107(5): 582–587 https://doi.org/10.1257/aer.p20171091

Why Being Wrong Can Be Right: Magical Warfare Technologies and the Persistence of False Beliefs†

By Nathan Nunn and Raul Sanchez de la Sierra*

Almost universally, societies have non-falsifi- In the face of this insecurity, beliefs in spells able beliefs about their origins, life after death, that protect villagers and militants against death and rituals that activate supernatural processes in combat have become widespread. We turn to help navigate life. Many such beliefs are now to a description of one of these beliefs, almost certainly incorrect, but are nonetheless bulletproofing. The information we present was ubiquitous. Some religious belief systems, such collected from interviews that were undertaken as monotheistic religions, are relatively well in the province of South Kivu in the DRC in known, but many others, such as superstitions, 2015. witchcraft, or sorcery, are much less well understood Bulbulia et al. 2013 . Yet, these belief I. Motivating Example: Ethnography of systems( are widespread within) developing coun- Bulambika tries in general, and in Africa in particular, and can have implications for behavior Gershman Although we have accumulated a large num- 2016 . Their prevalence raises an ( important ber of examples, we focus our discussion here question.) Given that they are often incorrect, on the village of Bulambika Bunyakiri in the why do they exist and how can they persist? province of Sud Kivu. The (village, which) is In this paper, we examine the role of magical shown in Figure 1, is typical of villages in the beliefs in warfare in the context of the Eastern area. Until 2012, the area had been exposed Democratic Republic of the Congo DRC . The to attacks by Hutu militias who operated with region has experienced persistent violence( ) and impunity in the absence of state forces to repel large-scale conflict since the Rwandan Genocide them the Front Démocratique de Libération du of 1994, when Hutu militia fled Rwanda into the Rwanda—henceforth,( FDLR . The Hutu mili- area. Eastern Congo became the epicenter of the tias, hiding in the forest and) the hills, would First Congo War 1996–1997 and the Second regularly engage in pillage, rape, and killings in Congo War 1998–2003( . Since) this time, con- Bulambika and neighboring villages. Villagers flict has persisted,( with dozens) of militant groups could not work in their fields far from the village still operating in the region today. It is common due to the high risk that they would be attacked. for villages to be under the de facto rule of mili- In January 2012, the FDLR killed 13 villagers tants who collect taxes and govern. For villages in the neighboring village of Lumendje, and that are not under the control and “protection” May 2012, they killed gruesomely 32 villagers of a militant group, it is common for them to be in the neighboring village of Kamananga, leav- frequently raided Sanchez de la Sierra 2016 . ing their mutilated bodies in public sight. Many ( ) of the villagers, including those we spoke with, had experienced the trauma of having friends * Nunn: Harvard University, Department of Economics, and family murdered, or of directly witnessing 1805 Cambridge Street, Cambridge, MA 02138 e-mail: violence. [email protected] ; Sanchez de la Sierra: University( of California Berkeley, Haas) School of Business, 2220 Piedmont One evening in 2012, an elder of the village Avenue, Berkeley, CA 94720 e-mail: rsanchezdelasierra@ had a dream. In his dream, the ancestors of his berkeley.edu . We thank Joseph( Henrich for helpful com- ) tribe taught him to use supernatural forces to ments. We thank Ariel Gomez, Lewis Dunia Butinda, and bulletproof the young men in the village and Samuel Leone for excellent research assistance. † Go to https://doi.org/10.1257/aer.p20171091 to visit the help them confront the source of the village article page for additional materials and author disclosure insecurity. The protection would make those statement s . who had been bestowed protection immune ( ) 582 VOL. 107 NO. 5 Why Being Wrong can be Right 583

men of the village. Now when the village or other villages in the area were raided by the( Hutu militia, rather than) fleeing, the young men would stay, undergo the ritual to activate the bulletproofing spell, and fight against the FDLR, and follow them deep in the forest. The gri-gri quickly spread throughout the area of Bunyakiri, where it became one of the adapta- tions of the gri-gri that a defense group called the Raia Mutomboki was using against the Hutu militia in other areas. Raia Mutomboki, which means “angry population” in Swahili, is a group Figure 1. Photograph Taken by the Authors in March that formed from multiple villages in the region of 2015 of the Village Bulambika in the Kalehe to defend the population against attacks by the Territory of the DRC FDLR. Initially, the defenders of the villages did not have guns. The villagers only had machetes to the bullets of the machine guns used by the and their gri-gri to fight the FDLR. Over time, FDLR. Any bullets fired at them would either they began to kill their enemies, from whom miss or bounce off. they also obtained firearms. This elder had a history of having similar Some of those from the villages that stayed to dreams and was believed by villagers to have the fight were shot and died. But, the cause of their ability to be informed about such powers.1 To death did not prove the spell to be false. Given obtain the necessary ingredients to implement the set of conditions that had to be respected, it the bulletproofing spell, the elder had to travel was logical that if they died, it must have been to a distant tropical forest. There, he searched because they did not follow some of the condi- for the roots, plants, and animal organs that the tions. By the time we arrived in the village in ancestors had instructed him to obtain. Upon his March of 2015, the village of Bulambika and return, the elder tested the newly produced pow- the others in the area had been freed and( had der. He asked other villagers to shoot at a goat to experienced peace for )the past two years. whom he had administered the powder. The goat This case provides one example for why false survived, and this proved to the villagers that the beliefs might persist. The bulletproofing ritual newly discovered “gri-gri” was effective.2 allowed the community to mobilize combatants To become bulletproof one had to participate by altering their beliefs about the likelihood in a ritual that was typically performed in the that they would die in combat i.e., about the forest. The bulletproofing protection requires costs of fighting back . While, at( the individual that certain conditions be followed in order for level, this false belief) is costly—it causes indi- the protection to be in place. Failing to respect viduals to underestimate the risk of combat—it the conditions would cause the protection nonetheless allowed the community to mobi- against bullets to stop working. The “gri-gri” lize against the aggressors, and to successfully only lasts for a short period of time, often hours eradicate them. Thus, although detrimental for or days. some, it was beneficial for the community as a After the discovery, the elder would perform whole. We now turn to a formal illustration of the protection rituals on those who volunteered this explanation. to protect the village and fight back against the Hutu militants. In general, these were the young II. Theoretical Structure

Consider a simultaneous move game where 1 While he was not the only person who had access to N players i.e., citizens who are each poten- supernatural powers to provide bulletproofing. He was one tial defenders( of the village each chooses the of a growing economy of emerging witch-doctors, who were amount of effort to allocate) toward protecting discovering secrets that helped the youth of the area mobi- their village, e, e, e, etc. Denote output i.e., lize against the threats and the attacks they were subject to. 1 2 3 ( 2 success in freeing the village by q e, … , e . Magical spells are also called “dawa,” which means ) ( 1 N) “the medicine.” Output occurs through team production. Thus, 584 AEA PAPERS AND PROCEEDINGS MAY 2017

Panel A. No bias, 1 Panel B. Mild bias, 0.75 Panel C. Signi cant bias, 0.5 γ = γ = γ = e2 e2 e2 * * e e * 1( 2) e1 e2 e e 1 1 ( ) 1 2( 1) e * e e * e 2( 1) 2( 1) 0.5 0.5 0.5 e * e 1( 2)

0 e1 0 e1 0 e1 0 0.5 1 0 0.5 1 0 0.5 1

Figure 2. Best Response Functions of Player 1, Player 2, and Nash Equilibria for Different Misunderstandings of the Cost of Fighting due to a Belief in Bulletproofing it is increasing in the effort levels of all players, belief about the cost of effort. Given that __ 1 e 2 is 2 i q ei 0 for all i, and by definition the true cost of effort, a value of i less than one ∂ q( · )/∂ e >e 0. Moreover, we assume γ ∂ (·)/∂ i ∂ j ≠ indicates that player’s beliefs underestimate the that greater effort by one villager increases true cost of effort. Beliefs about bulletproofing the marginal product of other villagers, serve to lower i. q e e 0. One is better able to defend γ ∂ (·)/∂ i ∂ j > In this setup, each player’s best response the village if others are fighting alongside him 2 3 / * ___1 / 1 3 her as well. function is e e / for player i j. i = 2 j ≠ Preferences are symmetric and are given [ γi] In the Nash equilibrium, e * ___1 for i 1, 2. by Ui V q e1, … , eN , ei for all i 1, … , N. i = 2 = = ( ( ) ) = γi That is, the payoff of every citizen depends on Both player’s best response functions and the the public good that is jointly produced, as well resulting Nash equilibria are shown in Figure( 2, as his her own effort. Each player’s utility is ) / panels A–C, for differing values of i. Without increasing in the amount of public good that is false beliefs Figure 2, panel A , eachγ player’s produced and decreasing in the amount of effort effort is equal( to 1 2 and each )player’s payoff exerted: U q 0; U e 0. / ∂ i/∂ > ∂ i/∂ i < is 3 8. It is straightforward to see that this is Given these assumptions, players’ efforts not /Pareto efficient. Instead, Pareto efficiency are strategic complements. One property of the is achieved when both players exert maximal Nash equilibrium in this type of game is well effort, e 1⁎ e 2⁎ 1. In this case, the payoff of understood: effort levels will be below the effi- each player= is 1=2, higher than 3 8. cient levels e.g., Alchian and Demsetz 1972; / / ( As shown in Figure 2, panels B and C, stron- Holmstrom 1982; Eaton and Eswaran 2002 . ger beliefs in bulletproofing a decrease in A belief in the efficacy of bulletproofing) ( i shifts the player’s best response functions serves to decrease the perceived costs of effort upward,γ ) so that for a given level of effort of the to an individual. If one believes that they are other player, the chosen level of effort is higher. protected from the enemy’s bullets, then the per- The case where the beliefs in bulletproofing ceived cost of bravery and effort is lower. This generates a value of i equal to 3 4 i.e., costs results in greater effort provision by all individ- are underestimated byγ both players/ by( 25 per- uals and a provision of the public good that is cent is shown in Figure 2, panel B. Thus, the closer to efficient. false) belief results in effort levels that are greater The following example provides a simple and closer to Pareto efficient levels. If bul- illustration of this point. Assume there are two letproofing induces i to be equal to 1 2 i.e., players, 1 and 2, and each chooses the level of costs are underestimatedγ by 50 percent/ , ( then effort to protect the village: e i 0, 1 . Total out- ) 3 ∈ ( ] e 1⁎ e 2⁎ 1 and Pareto efficiency is achieved. put is given by q ___e e , and the perceived = = = √ 1 2 __1 2 cost of effort is i e i , for i 1, 2. The γ [2 ] = 3 Note that it is necessary that both players have the false parameter indicates the bias of individual i’s γi belief for the Pareto efficient outcome to be achieved. VOL. 107 NO. 5 Why Being Wrong can be Right 585

Thus, false beliefs in bulletproofing, as long as of the enemies. This is done through explicit they are not too extreme, result in higher levels research and development by elders and indi- of effort being exerted and to higher payoffs to viduals who can communicate with dead ances- both players. tors of the tribe. For example, the anti-balle In the presence of group-level selection, vil- bulletproofing evolved from the anti-machete lages with the false belief of bulletproofing and( arrete-canon) aimed respectively at render- will be more likely to survive, and therefore we ing machetes and traditional guns ineffective. would expect this belief to spread throughout Yet, as machine guns became widespread, the the population. Thus, in an evolutionary setting speed with which they shoot bullets rendered the where groups compete for survival, such false arrete-canon obsolete, which relied on move- beliefs increase the average fitness of villages ments of the hand to stop bullets from leaving that hold them. Group-level imitation of suc- the gun. This motivated the development of the cessful groups would also have the same con- anti-balle, which protects fighters against bul- sequence Boyd and Richerson 2002 .4 Eastern lets, no matter their speed. Congo, like( many other war-torn parts) of the Bulletproofing spells are just one of many world, has characteristics that make it particu- spells, rituals, and superstitions that are present larly likely that group-level selection is strong in the Eastern DRC. Over the past few years, we see Henrich 2004 . The groups i.e., rural vil- have collected information about the universe of lages,( lineages, and) armed units (are small and rituals and spells that armed groups use in North homogeneous. They also have relatively) low lev- and South Kivu. We have documented the exis- els of migration between the groups. In militia tence of a total of 46 different military spell vari- groups, individuals are typically prevented from ants. These 46 spells arise from 17 recognizable leaving the group. In rural areas, there is little different major lineages of inventions. Each spell migration between villages. Most migration is to can serve multiple functions. One of these is bul- urban centers, while between village migration letproofing. Others include: scaring the enemy, of males is relatively limited. Lastly, because distracting the enemy, providing invisibility of the high rates of between-group conflict and during combat, preventing soldiers from leaving their contestation over a limited tax base, selec- the group, facilitating communication within the tion between groups is likely. With these char- group, increasing the compliance of the civil- acteristics, strong group-selection forces can ian populations, and even stopping helicopters induce false beliefs to spread and persist. in the air. These forms of magic are widespread among the militant groups in the region. Among III. Implications and Further Discussion 53 active armed groups for which we were able to obtain information on their magical warfare We have focused on one example of a situa- technology not including their factions , 46 tion where false beliefs were socially beneficial rely on gri-gri (. Interestingly, the 7 that do not) are and thus could arguably be sustained with group- of Rwandese origin or affinity Nyatura, M23, level competition. Although this is just one CNDP, FDLR, Mudundu 40, (Ngumino, and example, beliefs about bulletproofing and other Masunsu . Although further research is needed, similar beliefs about protection are widespread these spells) have consequences similar to bul- in Eastern DRC. The spells are continuously let-proofing. They reduce the perceived costs or fine-tuned and adapted to the changes in the increase the perceived efficacy of effort in con( - natural and supernatural warfare technology flict, helping to approach the) socially efficient ( ) levels of effort for the group. These magical beliefs emerge from a culture 4 For example, one could assume, as in Boyd and where spells are widespread. In addition to mili- Richerson 2002 , that an individual’s belief about bullet- ( ) tary spells of all armed groups, we collected the proofing i is influenced by a comparison of their payoff with that γof a randomly chosen person either from within universe of civilian spells that we could docu- the group or outside of the group . If( the payoff of the ) ment to exist, or have existed, in the territories other person is higher, that person’s bulletproofing belief of Walikale, Masisi, Rutshuru, Nyiragongo, and is adopted with some positive probability that is increasing in the difference in their payoffs. In such a setting, socially the city of Goma, comprising most of North Kivu excluding the “Grand Nord” . We docu- beneficial beliefs can spread, even if they are individually ( ) suboptimal. mented the existence of 39 well-known spells. 586 AEA PAPERS AND PROCEEDINGS MAY 2017

While the spells serve a range of functions, the common conditions have the feature of their many of them provide individuals with a greater adherence being difficult to observe, e.g., you sense of security and confidence, which can cannot drink rainwater, cannot eat cucumbers, serve to increase the effort provision of villagers etc. Second, conditions also result in the regu- for activities that are strategic complements and lation of behaviors by increasing the perceived therefore otherwise underprovided in equilib- costs of behaviors that are damaging for the rium, but also to reduce their anxiety and thus group. Common conditions for magical warfare improve their performance. For example, most spells to work are that the individual cannot steal of the spells provide protection, whether it be from civilians, cannot rape, cannot kill civilians, from drought, disease, attacks on the village, etc. Thus, through the conditions, such beliefs or even to harm potential thieves—and thieves serve to reduce the prevalence of undesired also believe in their efficacy, which acts as a actions, which are often socially inefficient. deterrent. Also common are spells of monitoring These conditions, especially for spells of armed i.e., to find lost items, to find out who is a thief, groups, evolved over the years together with the to( know when someone is coming, and to find objective of armed groups: initially, many popu- witches , and spells of production to increase lar militia had stringent conditions against abus- productivity,) and to decrease their productivity( ing the population, which eroded when armed of one’s business competitors . As in bullet- groups’ ties to the population weakened. Again, proofing, while these beliefs would) lead to effort the false belief results in individually subopti- levels that are individually suboptimal, if effort mal( actions) that are socially beneficial. levels are strategic complements, then they could lead to socially efficient effort levels and IV. Conclusions could persist as a result of group-level competition. The final type of spell that is common are The purpose of this paper is to explore the spells to harm others to make someone crazy, answer to a simple question: How can false to harm someone in order( to obtain money, to beliefs persist? We provided one answer to this render someone infertile, or to inflate the foot of question through a description of the traditional others . Although they can potentially be used belief of bulletproofing in the Eastern DRC. for coercive) purposes, these spells may have Using a case study and a simple theoretical benefits that also fit within the framework of framework, we have shown how such a false this paper. For example, the knowledge of such belief can be beneficial for the group, and( how) spells may induce individuals to behave in a intergroup competition can result in its per- more socially beneficial way, e.g., to not steal sistence. This can account for why beliefs about from or harm others. These spells effectively protection in combat are ubiquitous in the DRC increase the perceived cost of bad behavior. and other conflict ridden locations. Thus, individuals will exert less effort into activities that exert negative externalities on others. In equilibrium, theft levels, though individually References optimal, will be higher than is socially optimal. Beliefs in such spells will increase the perceived Alchian, Armen, and Harold Demsetz. 1972. cost of theft and similar bad behaviors , reduc- “Production, Information Costs, and Economic ing individual( effort on these activities,) moving Organization.” American Economic Review 62 the society closer to the social optimum. 5 : 777–95. As mentioned, a common characteristic of bul- Boyd,( ) Robert, and Peter J. Richerson. 2002. letproofing spells is their necessary behavioral “Group Beneficial Norms Can Spread Rapidly conditions that must be respected by the subjects in a Structured Population.” Journal of Theo- in order for the spells to be effective. We view retical Biology 215 3 : 287–96. these conditions as playing two functions. First, Bulbulia, Joseph, Armin( ) W. Geertz, Quentin D. they serve to make the belief harder to falsify. Atkinson, Emma Cohen, Nicholas Evans, Pieter For the example of the bulletproofing spell, the Francois, Herbert Gintis, et al. 2013. “The Cul- death of a fellow combatant is equally consistent tural Evolution of Religion.” In Cultural Evo- with the belief being false, as it is with the com- lution: Science, Technology, Language, and batant violating one of the conditions. Many of Religion, edited by Peter J. Richerson and VOL. 107 NO. 5 Why Being Wrong can be Right 587

Morten H. Christiansen, 381–404. Cambridge, Henrich, Joseph. 2004. “Cultural Group Selection, MA: MIT Press. Coevolutionary Processes and Large-Scale Eaton, B. Curtis, and Mukesh Eswaran. 2002. Cooperation.” Journal of Economic Behavior “Asymmetric Equilibria and Strategic & Organization 53 1 : 3–35. Behaviour in Oligopolistic Markets.” Applied Holmstrom, Bengt. 1982.( ) “Moral Hazard in Microeconomic Theory, edited by B. Curtis Teams.” Bell Journal of Economics 324–40. Eaton, Cheltenham: Edward Elgar. Sanchez de la Sierra, Raul. 2016. “On the Origin Gershman, Boris. 2016. “Witchcraft Beliefs and of States: Stationary Bandits and Taxation in the Erosion of Social Capital: Evidence from Eastern Congo.” https://raulsanchezdelasierra. Sub-Saharan Africa and Beyond.” Journal of files.wordpress.com/2013/09/revised- Development Economics 120: 182–208. manuscript.pdf accessed March 28, 2017 ( ).