Givers of Great Dinners Know Few Enemies: The Impact of Food Sufficiency and Food Sharing on Low-intensity Household Conflict in Eastern Democratic Republic of Congo

Abstract Our study establishes a linkage between household food sufficiency and food sharing behaviour with the reduction of low-intensity, micro level conflict using primary data from 1763 households of eastern Democratic Republic of Congo. We develop a theoretical explanation of such behaviour using the seminal theories of dissatisfaction originating from food insecurity and the reciprocity of gifts in economic anthropology. We first examine if food sufficient households are less likely to engage in low-intensity conflict. Following, we investigate possible heterogeneous effects of food sufficiency, conditional on food sharing behaviour. Using propensity score matching, we find that food sufficiency reduces household conflict risk by an average of around 10 percentage points. Upon conditioning on food sharing behaviour, we find that conflict risk in the subpopulation of food sufficient households is 13.8 percentage points lower for households that share their food while the effects disappear for households that do not share their food. Our results hold through a rigorous set of robustness checks including doubly robust estimator, placebo regression, matching quality tests and Rosenbaum bounds for hidden bias. We conclude that food sufficiency reduces low-intensity conflict for households only in the presence of food sharing behaviour and offer explanations and policy prescriptions.


Introduction
Historical accounts of food shortages causing conflict can be traced back to the Russian, English and French Revolutions of the seventeenth and eighteenth century. In modern times, prevalence of hunger has been documented to drive violent behaviour and conflict between and within communities through environmental, social, economic, and political channels (see for example, Bora, Ceccacci, Delgado, & Townsend, 2010). Due to the complexity of establishing a promote peace within fragile societies. Third, we provide empirical evidence of household level behaviour of engaging in petty conflict under food insufficiency as well as the reciprocal culture of gift and favours in the African peasant society previously absent from the literature. Thus, we contribute to the political anthropology of food sufficiency, food sharing and low impact community level conflict.
The remainder of the paper is organised as follows: section 2 establishes the theoretical basis and literature review; section 3 explains the data, variables and descriptive statistics; section 4 develops an empirical framework; section 5 presents the results and analysis while section 6 offers summary of results, discussion, and concluding remarks.

Theoretical framework and literature review
We commence our investigation by exploring the connection between low intensity, often nonviolent, micro-level conflict with food sufficiency in the eastern DRC society. We choose to study food sufficiency 2 over food security for two reasons. First, household food security is a multidimensional phenomenon that is difficult to capture without a detailed survey dedicated to that purpose. Second, food security can affect household conflict through multiple channels, thereby making causal exploration challenging and prone to multiple sources of bias. Instead, we focus on one aspect of household food securitywhether the household had sufficient food for the entire family over a six-month period. We draw motivation from FAO's Coping Strategy Index (Maxwell, Watkins, Wheeler, & Collins, 2003) which states, Clearly, food security is about much more than just how much people have to eat … Yet, having 'enough' food to eat is clearly the most important outcome of being food secure, and while physiological requirements differ, people largely know whether they have 'enough' or not.
The Social Science literature on peasant behaviour draws upon two sets of theories to explain disagreement, aggression, and rebellion expressed by citizens in a predominantly food insecure society. The first alludes to moral values while the second relies on rational choice. From a 'moral economy' perspective (Thompson, 1971), if pervasive food insufficiency persists households may get involved in disputes from the notion of subsistence ethics. Aggrieved households can also express dissatisfaction to resist the status-quo (Scott 1977(Scott , 1985(Scott , 1992 through unplanned disputes and confrontations. Recent contemporary studies in interdisciplinary development journals have confirmed the conjecture that food insecurity drives conflict through grievances. For example, food insecurity has been shown to initiate feelings of horizontal inequality, grievances and discontent (Humphreys & Weinstein, 2008, Østby, 2008Stewart, 2011). Nutrition and health studies have established that lack of food and hunger is related to poor mental health, depression, anger, and aggression (Bushman, DeWall, Pond, & Hanus, 2014;Carter, Kruse, Blakely, & Collings, 2011;Chilton & Booth, 2007;Heflin, Siefert, & Williams, 2005). On the other hand, evidence from Nepal and South Sudan suggest that food security can enhance a feeling of equality and harmony at a community level (McCandless, 2012). The second set of theories suggest a reflexive yet rational choice to engage in conflict in the process of acquiring food to sustain one's household. Deriving from Popkin's (1979) 'rational peasant', involvement in petty conflict can arise from an idea of serving self-interest in the context of rational decision-making. The process of acquiring food or ensuring necessary food sufficiency may lead a food insufficient household to daily disputes in a direct or indirect manner. Recent studies have also shown that food insecurity can provide individuals and households with both material and non-material incentives to engage in any form of anti-social behaviour (Justino, 2009;Martin-Shields & Stojetz, 2018).
Next, we turn our attention to channels through which households may be able to mitigate conflict by investigating the food sharing behaviours of households. This idea emanates from Polanyi's (1944) explanation of 'market morality' in the absence of a regulated market Impact of food sufficiency and sharing on conflict 1411 economy, which is also mentioned in moral economy (Scott & Bhatt, 2001). It can be further extended through the complexity of 'gift giving' behaviours in different cultures detailed by Mauss (1900Mauss ( /1925 and Malinowski (1922Malinowski ( /1984. Mauss (1900Mauss ( /1925 showed gifting in archaic cultures is driven by an obligatory sense of responsibility. Under this theory, it can be argued that food sharing behaviour by farming households originates from a moral community obligation which in turn diffuses a conflict situation. On the other hand, Malinowski's (1922Malinowski's ( /1984 argument of gift giving based on selfish motives appeals to the rational choice model. This no free gift without expectation model would predict that food is shared by households to avoid or diffuse conflict with individuals or members of the community. Thus, food sharing behaviour could be explained as a strategic choice of households as a form of bribe to avoid conflict with rebel groups and pastoralists. Distinguishing between such moral and rational motives for food sharing and conflict of households is beyond the scope of this paper. Instead, our intention is to document the nature of food sufficiency, food sharing, and low-intensity conflict that we observe among poor households in DRC and to establish a relationship between food sharing behaviour and interhousehold conflict. Thus, we argue that the poor can have a multitude of reasons, that may be based on morality, rationality, or both, to display aggressive behaviour in an environment of food insecurity. We refer to Rationality, Morality and Collective Action, where Elster (1985) highlights the existence of mixed motivations of participants in collective action, including a sense of duty, pleasure of participation, peer pressure, and so forth. Furthermore, Fafchamps (1992) lays out detailed arguments for acts of solidarity, which may include gift exchanges or food sharing, in preindustrial society being consistent with notions of both rationality where people pursue their long-term self-interest as well as the moral economy of peasants. Recent advances in behavioural and experimental economics also suggest that rationality and morality theories on peasant behaviour need not lie in contradiction. Contemporary research in development explains ways in which incentives and information, together with behavioural constraints, can shape human behaviour (see for example, Binswanger & Rosenzweig, 1986;Duflo, 2006;Duflo & Banerjee, 2011;Duflo, Kremer, & Robinson, 2011;Mullainathan & Thaler, 2000;Stiglitz, 2002). These studies explore the structural constraints that poverty and uncertainty impose on decision-making in an environment where information is incomplete. From a behavioural science perspective people are bounded by their ability to analyse and compute information; they do not always make choices that are in their best interest in the long run; and they are not purely motivated by self-interest (Mullainathan & Thaler, 2000). Bertrand, Mullainathan, and Shafir (2004) further argue that traditional economics has wrongly classified the poor as belonging to one of two extreme categories -either as fully rational beings who always act on self-interest or as having deviant values such as being misguided or confused. They argue that in reality, people are neither fully rational, nor fully deviant but rather act out of minor psychological reasons such as fear of humiliation, fear of isolation, and so forth.
Thus, drawing from these behavioural insights, it seems plausible that both theories of moral awakening and rational response in the process of obtaining food can help explain a household's altercations with fellow villagers, government and other groups that we observe in DRC. For example, food sufficient households may be less prone to grievances, greed, psychosocial frustration, anger and emotional stress compared to food insufficient households. By feeling content, such households would have lower motivation and aggravation for engaging in conflict. Furthermore, the opportunity cost of engaging in conflict may also be higher for food sufficient households, thus making it a rational choice for such households to avoid conflict. The food sharing behaviour of households can also be explained by both sets of theories on peasant behaviour. Households may share food with others out of a sense of justice, moral values, or social obligation. They may also share food as a rational strategy to avert conflict with others.
The reconciliation of seminal theories of peasant behaviour, reciprocation and obligation, through advances in game theoretic behavioural sciences, guides us to formulate two hypotheses -(i) food insufficiency can lead to aggressive societal behaviour through petty altercations, disputes and disagreements and (ii) the act of sharing food may help households that do so avert petty conflict with other households in the community.

Data, variables and descriptive statistics
During July 2014, The Howard G. Buffett Foundation funded and initiated the data collection for this research through Texas A&M University, as part of its Best Practices in Coffee and Cacao Production (BPCC) Project. The authors of this paper contributed to the survey design and data collection procedure. Pre-survey focus group discussions were conducted to identify key issues. Follow-up field work and three post-survey focus group discussions in Beni, Lubero, and Rutshuru areas were conducted to discuss and disseminate the knowledge from the surveys as part of the BPCC project. Details on the quantitative survey design and data can be found in Fatema (2019).
Our outcome variable of interest is the incidence of low-intensity disagreements and altercations experienced by households. Pre-survey focus group discussions with community members helped identify the most prevalent categories of petty conflict in our study areas. These included: (a) conflict with neighbours and fellow villagers; (b) disagreement involving Virunga National park; (c) landholder reclaimed occupied land; (d) border conflict with landholder; (e) dispute among non-dwelling family members; (f) occupied land granted to a new tenant; (g) disagreement with pastoralists; (h) conflict over community resources and agricultural inputs; (i) resource conflict with rebel forces; (j) land conflict with rebel forces; (k) land conflict with government; (l) resource conflict with government forces; (m) other kinds of conflict with government forces; and (n) any other kind of conflict that they were asked to specify. A binary variable was constructed to equal one if the household experienced at least one such instance of conflict in the six months prior to the survey.
The main explanatory variables include 'food sufficiency' and 'benevolence'. To measure food sufficiency households were asked, 'how often have you had difficulty feeding your entire family in the last six months?' Respondents could choose between 'often', 'sometimes' or 'never'. For our analysis, we categorise a household as food sufficient if it responded 'never' and food insufficient if it responded 'often' or 'sometimes'. To measure benevolence, we asked households if they had helped others with food in the past six months. Households that answered positively were classified as benevolent and households that responded negatively were categorised as non-benevolent.
To be circumspect about potential endogeneity biases, we employ a cautious research design and include thorough robustness measures. We argue that since households did not have much to gain by claiming to be food sufficient or insufficient, we can rule out the possibility of intentional misreporting. In addition, our summary statistics show that around 56 per cent of the households claim to be food insufficient, which is consistent with reported household surveys conducted by World Food Programme (2014) and United Nations (2010) in DRC and North Kivu. Based on the narrow definition of food sufficiency used and the nature of conflicts explored, it is unlikely that such low intensity, interhousehold conflict incidences would affect households' likelihood of having sufficient food over a sustained period. Citing some examples, conflict occurring over Virunga National Park resources 3 has a very limited probability to cause household food insufficiency. While violent conflict occurring from inheritance with immediate family may cause food shocks, we specifically inquire about disputes (alluding to a lower intensity of conflicts) over inheritance that is unlikely to cause food insufficiency within a six-month period. Similarly, for every other low-level conflict we explore, food insufficiency during a six-month period is Impact of food sufficiency and sharing on conflict 1413 improbable. Hence, we argue that our cautious approach and the categories of conflict considered abate reverse causality suspicions to a large extent. While it is impossible to rule out the presence of omitted variables from survey data, we include a large set of control variables from the relevant literature, as shown in Table 1. Community fixed effects are also included to capture inherent differences between communities.
The table shows that food sufficient households are systematically different from food insufficient households in terms of socioeconomic and demographic characteristics, thus justifying the need for a matching method as outlined in the next section.

Empirical framework
In an ideal world where food sufficiency was randomly assigned to households, estimating average treatment effects might provide the causal impact of food sufficiency on conflict. However, such an experiment that entails artificially ensuring food sufficiency for randomly assigned households is neither possible nor ethical. Since we cannot randomise an intervention to avoid selection bias, we are left with quasi-experimental techniques (see Cook, Shadish, & Wong, 2008) to improve (if not isolate) the estimates of the effect of food sufficiency on conflict. Two prominent approaches, instrumental variables and regression discontinuity, would be useful methods but are difficult to employ. Valid instruments are difficult to identify (Imbens & Wooldridge, 2009). Some possibilities exist, such as natural disasters, but require assumptions such as exogeneity of the instrument, that are particularly difficult to justify in this context. Regression discontinuity is another option but requires consistent decision-making around some arbitrary cut-off. In our case, food insufficiency is unlikely to be allocated in such manner. In addition, as Table 1 shows, the inherent differences between food sufficient and insufficient households in our sample points to potential selection bias which may lead to biased estimates if compared directly. Therefore, we employ a third quasi-experimental approach, propensity score matching (PSM), whereby observable differences between food sufficient and insufficient households that may confound the estimates are statistically balanced to neutralise any selection bias, thus allowing us to isolate the effects of food sufficiency on conflict. Propensity score for this study is the conditional probability that a household will be food sufficient, given its vector of observed covariates. A logit model is used to estimate the propensity score. PSM pairs each food sufficient household with food insufficient households with similar observable characteristics before estimating the average treatment effect on the treated (ATT) as the difference in mean outcomes between the two groups. This can be expressed as follows: where T is a binary 'treatment variable' equal to 1 if the household is food sufficient and 0 otherwise; Y 1 is the conflict outcome for a food sufficient household and Y 0 is the outcome for the same household had it not been food sufficient. Though Y 0 is the counterfactual, which is not observed in reality, given proper matching food insufficient households can serve as an appropriate proxy. Since PSM methods are sensitive to the exact specification and matching method (Caliendo & Kopeinig, 2008;Imbens, 2004), we employ three commonly used algorithms to ensure the robustness of PSM estimates: (i) nearest neighbour matching (NNM) using three neighbours with replacement; (ii) Kernel based matching using the Epanechnikov Kernel function with a bandwidth of 0.06 (Heckman, Ichimura, & Todd, 1997); and (iii) radius matching with a calliper of 0.001. The choice of variables included in the estimation is guided by economic theory, previous research and the literature on matching (see Abadie & Imbens, 2006;Caliendo & Kopeinig, 2008;Dehejia & Wahba, 2002;Heckman et al., 1997;Heckman, Ichimura, & Todd, 1998). We also employ heterogeneous treatment effect by observable characteristics (Crump, Hotz, Imbens, & Mitnik, 2008;Imbens & Wooldridge, 2009) to estimate whether the act of sharing food affects the conflict outcome of food sufficient and insufficient households differently. To do this, we divide our sample into two subsamples, based on whether or not the household shares food with others, before estimating a separate ATT for each subsample. The difference of the subsample ATTs provides the heterogeneous treatment effects (see Kibriya, Xu, & Zhang, 2017;Verhofstadt & Maertens, 2015;Xie, Brand, & Jann, 2012) and is expressed as follows: where B ¼ 1 if the household displays food sharing behaviour and 0 otherwise. Further details of PSM and ATT are provided in the Appendix. Table 2 presents the results from a logit model to determine the likelihood of being food sufficient, given observable characteristics of the household. It appears that certain traits are more likely to make a household food sufficient. For example, education enables a household to make informed decisions about agricultural practices (such as crop diversification or technology adoption); higher income allows households to not only purchase more food but to invest in agriculture; increased access to information and communication technologies may reduce information asymmetry as well as transaction cost for farmers; extension services from government or non-government organisations may make farming households more aware of new technologies and ways to use them. All these factors may further enhance agricultural income and productivity and thus ensure food sufficiency. Given the large fraction of rural households that use fuelwood for cooking it is not surprising that Impact of food sufficiency and sharing on conflict 1415 access to cooking fuel increases the probability of being food sufficient. Finally, holding important positions in the community can help households gain access to credit and services and increase social capital which can improve food sufficiency. Figure 1 shows the distribution of propensity scores between food sufficient and food insufficient households. Visual inspection of the density distributions of propensity scores for the two groups shows that there is much overlap between the estimated scores. Thus, the common support assumption is satisfied. Furthermore, there is sufficient difference in the distribution of propensity scores between the two groups to justify using matching. Table 3 summarises the main results using each of the three matching algorithms. Panel A shows the ATT estimates of conflict risk for all households regardless of food sharing behaviour. Panels B and C show the ATT estimates of conflict risk for benevolent and non-benevolent households respectively. It can be seen from Panel A that consistently across all three matching algorithms, food sufficient households are at an average of 7.6 to 10.1 percentage point lower risk of experiencing low-intensity conflict. Given that the average household conflict risk for the sample is around 50 per cent, a 10 percentage point reduction leads to an effect size of about 20 per cent lower conflict risk for a typical food sufficient household. The 90 per cent confidence interval shows that a value of zero is not included in any of the intervals which gives us reasonable confidence in our estimates.

Effect of food sufficiency on conflict
Information collected from focus group discussion reinforce and provide nuanced context to these findings. Participants unanimously agreed that households that did not have sufficient  Impact of food sufficiency and sharing on conflict 1417 Next, we investigate the heterogeneous treatment effects of food sufficiency on low-intensity conflict, conditional on food sharing or benevolent behaviour of households. First, we partition the data into two subsampleshouseholds that share their food with others, or benevolent households, and households that do not share their food with others, or non-benevolent households. Propensity scores are calculated for each subsample. Propensity score distributions are similar in the treated and control groups post-matching. An average treatment effect on the treated (ATT) is estimated for each subsample. This is analogous to including an interaction term in an OLS regression but is the more desirable approach when using PSM since it allows us to compare outcomes in two subsamples rather than having to repeat the computationally intensive process of balancing all covariates in each stratum of propensity scores every time a new dependent variable is introduced. Results are shown in Panels B and C of Table 3. Panel B shows that conditional on benevolence, the coefficient of food sufficiency on conflict risk is negative and statistically significant at 10 per cent significance level or lower, depending on the matching method used. Depending on the algorithm used, the expected conflict risk in the subpopulation of food sufficient households that share their food is 8.1 to 13.8 percentage point lower than that of food sufficient households that do not share their food with others. Confidence intervals at both 90 and 95 per cent are included to show that with nearest neighbour and kernel matching algorithms, the confidence intervals do not include zero even with a 95 per cent confidence level but does so when using a radius matching algorithm. However, at 90 per cent confidence level, the interval does not include zero even with radius matching giving some reasonable confidence in our point estimates. In contrast, the results in Panel C show that when households do not show benevolence through food sharing, the expected conflict risk in the subpopulation of food sufficient households is statistically no different from that of food insufficient households at 10 Source: Authors' calculations based on the survey data. Notes: Ã , ÃÃ , and ÃÃÃ indicate significance at 10 per cent, 5 per cent, and 1 per cent levels, respectively. All estimates shown are average treatment effect on the treated. Abadie and Imbens (2006) robust standard errors reported for nearest neighbour matching. Bootstrapped standard errors with 100 replications of the sample reported for kernel and radius matching. Kernel matching uses a bandwidth of 0.06. Radius matching uses a calliper of 0.001.
per cent level of significance or lower. The confidence intervals in Panel C leads to the conclusion that since zero lies within the 90 per cent confidence interval for all three matching methods for non-benevolent households, the current data does not confidently determine any effect of food sufficiency on reducing conflict risk if a household is non-benevolent with its food. Thus, with reasonable confidence we can conclude that this data suggests an effect of food sufficiency in reducing low-intensity conflict if and only if the household shows benevolence through its food sharing behaviour. These quantitative results were substantiated through our field notes and focus group reactions. Starting with a family's closest kin, households shared food with one another, including in some cases when they themselves did not have enough. It was clear that food sharing behaviour was not an uncommon practice and that it was not constrained by a household's wealth or food security status. The motives varied and were coined as generosity, reciprocity, and social obligation. Accordingly, food sharing is not only limited to special occasions. Neither is it limited to cooked food and may include raw food or food commodities. However, it was clear that it is culturally unacceptable for a household to get into disputes with a person/household that gifted them food. The unanimous statement was that 'no North Kivu resident will ever disrespect a household that fed them or harm a home that they dined in'. The ubiquitous norm of the region is to respect, preserve, and protect hosts and helpers.

Sensitivity analysis and selection on unobservables
We run a series of check and balance tests to ensure that the assumptions of propensity score matching hold and that the quality of matching is reliable. These tests include a 'Placebo' regression, covariate balance test for the matching process, performance comparison between nearest neighbour, radius, and Epanechnikov Kernel matching algorithms, covariate balance test for the matching process given benevolence and non-benevolence and comparison of the matching quality indicators in benevolent and non-benevolent sub-samples. Results for these sensitivity analysis and selection of unobservable are available in the Supplementary Material.
Next, we use robust measures of low-intensity conflict, such as the number of conflict incidents, conflict with individual households and conflict with groups to test the robustness of our estimates. In addition, we use alternate measures of food sufficiency, such as whether any member of the household had ever gone hungry in the past five days or whether any children in the household had ever gone hungry in the past five days. Our main results are unaffected by these alternate dependent and treatment variables. Furthermore, we investigate possible reverse causality issues by testing whether low-intensity household conflict affected food sufficiency using each of these alternate variables. Much to our assurance, this is not the case.
After that, we test the sensitivity of our results using a Doubly Robust Estimator (DRE) and the Rosenbaum bounds for hidden bias (Rosenbaum, 2002). DRE offers the advantage that by requiring specification of two separate models, one for treatment (food sufficiency) and one for outcome (conflict), it can provide unbiased estimates as long as either model is correctly specified. Results show that the DRE estimates are similar to the propensity score matching estimates in Table 3 thus assuring robustness of results.
Finally, the Rosenbaum bounds for hidden bias (Rosenbaum, 2002) test for potential selection on unobservable covariates. For example, if household members show aggressive behaviour both in pursuing measures to achieve food sufficiency as well as in their preference for violence, our estimates may be biased by the presence of unobserved aggression. The Rosenbaum bound measures how big this difference in unobservables needs to be to render the ATT estimates insignificant by presenting a worst-case scenario that assumes treatment assignment is influenced by the presence of unobservable covariates (Li, 2013

Impact of food sufficiency and sharing on conflict 1419
Overall, the results satisfy us that the assumptions and conditions of propensity score matching have been met, matching quality has been ensured, all point estimates are doubly-robust, and that any unobservable characteristics would have to cause matched households to differ substantially for the ATT estimates to be affected by potential hidden bias.

External validity and limitations
While PSM has its limitations (King & Nielsen, 2016), it still remains one of the most popular methods for simulating randomised experiments and making causal inference using observational data. Propensity scores have many productive uses besides matching, such as in regression adjustment, inverse weighting, and within other methods (King & Nielsen, 2016). We have applied some of these alternative estimation strategies to the analyses in this paper and our results remain robust. Furthermore, the effectiveness of propensity score methods depends on a carefully selected set of covariates that requires an understanding of the potential relationships between observed and unobserved covariates and how they may affect the assumptions required for estimation (Pearl, 2009). The choice of covariates included in this study are guided both by economic theory, previous research, our fieldwork and the literature on matching (see for example Abadie & Imbens, 2006;Caliendo & Kopeinig, 2008;Dehejia & Wahba, 2002;Heckman et al., 1997Heckman et al., , 1998.
We use seminal works conducted in England, Southeast Asia, and Oceania to develop a theoretical framework that lays a foundational linkage between interhousehold and community disputes and food insecurity in African societies. While our results are robust, and tested with different specifications, we cannot claim that they will hold universally. Eastern DRC is a resource rich conflict prone society with its unique challenges and surroundings. We designed our quasi-experimental setting and questionnaire based on the distinctive conflict situations and societal fabric of North Kivu. However, in general, the channel through which food security is connected to household/community disputes and the mitigating effects of food sharing behaviour is expected to hold in neighbouring Sub-Saharan African societies with similar conditions.
The Data section has previously discussed ways in which we address potential concerns of reporting errors and omitted variables. We address any pertinent concerns of reverse causality in three distinct ways. First, our theoretical framework stands on seminal theories of choice and morality in developing societies. Accordingly, we hypothesise that food insufficient households are more likely to get engaged in sudden disagreements and disputes. 4 We acknowledge that long lasting land cases with family members (settled in court and/or community councils) or uprisings against rebel or government groups (reported by conflict databases such as ACLED) may drive food insecurity. However, we specifically focus on low-intensity conflicts originating from sudden and petty altercations which should arise through food need, frustration or resistance and most likely do not contribute to food insufficiency. Second, as part of the previously mentioned Best Practices in Coffee and Cacao Production (BPCC) Project in North Kivu our team spent two years collecting anecdotal evidence, conducting pre and post survey focus group discussions and collecting individual excerpts of disagreements within the society. These transcripts and experiences clearly suggest that food insufficiency related frustration was one of the major harbingers of low-intensity conflict and petty disagreements within households and groups. Our surveys were carefully designed to frame questions that refer to such disagreements, frustrations and altercations. Third, we perform specific tests to check for potential reverse causality, as mentioned in the previous section.
6. Summary, discussion, and concluding remarks By exploiting survey data of 1763 households, we study the impact of food sufficiency and food sharing behaviour on low-intensity conflict experienced by farming households in the North Kivu province of eastern DRC. We employ a quasi-experimental propensity score matching approach to control for pre-existing differences between households. By exploiting heterogeneous treatment effects, we find empirical evidence to support that food sufficiency reduces household conflict risk only in the presence of food sharing behaviour. Though potential biases were accounted for through various econometric approaches including matching, placebo regression, doubly robust estimator, Rosenbaum bounds for hidden bias, we take extreme caution to claim causality. However, our checks and balance tests do not indicate concern for violations of the assumptions used suggesting that a causal impact is plausible at the least.
Our initial empirical explorations isolate food insufficient households and show that they are more prone to conflict with government and park officials, NGOs, neighbours, rebel groups, and fellow villagers. These results are embedded in both rational choice and moral economy schools of thought. While the moral economy and rational peasant theories may differ at an intellectual plane, the practical implications of both lead to similar effects in the context of conflict. Especially in a society such as North Kivu which is built on solidarity networks of kinship and village bonds, a 'repeated game scenario' would reconcile any inherent differences between these ideologies (see Fafchamps, 1992, for a detailed representation).
Extending from the theory of moral choices, the second part of our exploration contributes to the philosophical and social norms of gift exchange. Thompson and Scott's idea of moral economy is based on social justice, which is, ideally, built on fairness and mutual benefits. The act of gifting food protects food sufficient households from social aggression. While this act of altruism appears to be driven by moral economy, it is also analogous to rational deliberation. Such acts can be practiced in closed communities but is difficult to envision in modern market economies regulated by institutions as described by Polanyi (1944). North Kivu's failed market conditions with pervasive scarcity initiates an informal economy which depends on barter, social justice and benevolence (or fairness). Gifts in such contexts is tantamount to Mauss's (1900Mauss's ( /1925 Polynesian cultural interactions or Malinowski's (1922Malinowski's ( /1984) Papua New Guinean social exchanges. While Mauss and Malinowski differ in the motives of the giver, they both agree that social giving can gel kinship bonds which serves socio-economic and political functions. At a household level, gifting can promote social fairness and justice and ultimately serve mutual interests of both the food sufficient and insufficient. Food sharing thus promotes the magnanimity of the giver and establishes a higher social hierarchy for the household. On the contrary, a food sufficient household that does not share its food stands to lose its honour and social hierarchy. The givers of food can share food either out of kindness or a selfish motive to avoid conflict. The receivers of food are then bound to reciprocity (Malinowski 1922(Malinowski /1984 or obligation (Mauss, 1900(Mauss, /1925 to remain peaceful with the givers while seeking justice from others. The social norm of reciprocity and obligation thus avoids conflict for food sufficient households who share their food. While the existing literature mostly uses cross country or district level data for analyses of civil wars and conflicts, we shed light on facets of household conflict that are most often omitted from headlines and subsequently ignored. Our findings also advance the understanding of the intricate relationship between food sufficiency, kinship and conflict at the micro level. Food aid programs have been documented to have mixed effects on conflict (Barrett, 2001;Nunn & Qian, 2014). Our approach of analysing the relationship between household level food sufficiency and food sharing tendencies with low-intensity local conflict can offer new insights to program implementers and evaluators of international food policy. Our findings show that food sufficiency alone cannot reduce low-intensity, household conflict unless accompanied by the idea of kinship. As such, our results illuminate the need for food aid and other relevant food policies to incorporate communal affiliation. Thus, policymakers may find that encouraging food sharing practices within society is a useful and effective tool that can complement food security, poverty alleviation and conflict reduction initiatives.