Readout of the villages leader survey Annex B.pdf

Description and basic results of 2022 village leader survey Team N043 PIs

May 2023

Project Background The results described here are a component of a broader project under the DRG Learning, Evaluation, and Researcher (DRG-LER) II (Tasking N043 Contract No. GS-10F-0033M / Order No. 7200AA18M00016). The United States Agency for International Development (USAID) has commissioned this study to inform USAID/Mali programming in communities affected by conflict. The study also seeks to inform the wider community of practice working to address democracy, human rights, and governance issues in Mali.

Building on the findings from more than a dozen focus groups in rural Mali commissioned as part of a prior research tasking, this study proposes to experimentally evaluate two approaches to promoting reconciliation in communities affected by conflict. The first intervention will focus on trust-building within communities, with a particular emphasis on supporting community mechanisms or institutions that can mediate disputes between groups before they escalate into violence. The second intervention will focus on addressing basic needs through the provision of unconditional cash transfers. The motivation for this intervention is that livelihood support may blunt some of the competition and resentments that stand in the way of reconciliation. This study may also test both interventions in combination, motivated by the supposition that addressing basic needs is a prerequisite to successful reconciliation.

Specific research goals This survey was run to meet two research needs: (i) learning more about a candidate set of villages in order to help us choose which would be eligible for an intervention focused on dispute resolution and conflict prevention, and (ii) to obtain additional background information about facts on the ground relevant to the design of that intervention and the research tools to study its effectiveness.

The survey sample initially included 415 villages, determined on the basis of extensive analysis of secondary data sources in an effort to locate communes and villages within them that would be sufficiently safe and would have ethnic heterogeneity. (Rather than providing a list of these 415 villages here, in other materials we provide a list of the 329 villages from this set that will be selected for this project as described below). In 8 of these villages we did not receive even the most basic information (e.g. number of households), leaving a sample size of 407 villages described herein. In each village, the plan was to interview the village chief and one or more additional village level leaders, though exceptions to this are noted below.

Key language To clarify our purpose and the terminology used here, please note that we have adopted the use of the term dispute to refer to (often small) disagreements or transgressions between parties, which typically have only small consequences in terms of damage to property, and little or no physical violence. The primary concern is not the dispute itself, but the possibility that it escalates into a larger conflict in the village, in which different groups become recruited, presumably across pre-existing cleavages. We thus also use the terms groups or cleavages in reference to the sets of people that are anticipated to become engaged opposite of each other in these escalation events. The aim of the intervention will be to either (i) assist in building or strengthening the “dispute resolution” mechanisms that can prevent escalation of disputes into conflicts; or (ii) aid in establishing or enforcing rules that help to reduce the frequency of those initial disputes themselves. Most of our proposed interventions are more like (i) than (ii), but both are possible.

For many questions it will be easier to work with a village-level dataset with just one observation per village.

In reality, not all of these villages had a village chief present. The interview representing each village comes from the following distribution of respondent types:

## code num_responses ## Village Chief 1 385 ## Village Chief Representative 2 15 ## Village Chief Advisor 3 7

Number of households The number of households per village varies widely, and in fact quite a few are smaller than we’d like for purposes of deploying the intervention, and some are too large.

summary(dat_village$S1Q04A)

## Min. 1st Qu. Median Mean 3rd Qu. Max.

## 2.0 38.5 65.0 182.2 150.0 9000.0

In short: (i) half of villages below or at 65 households; (ii) when they start to get large, they get very large with an exponential distribution, representing large towns or cities. To better visualize this we can transform population size by log-base-2, meaning that every increase of one unit on the horizontal access below indicates a doubling of size. On this rescaling, the distribution looks more normal (bell-shaped).

ggplot(dat_village, aes(x=log2(S1Q04A))) + geom_histogram(bins=10) + xlab("Nu mber of households (log base 2)") + theme_bw()

(Lack of) Ethnic diversity The principal measure of ethnic diversity is the “proportion of the population that is the second largest group”. Here is the distribution:

dat_village = dat_village %>% mutate(S1Q13B = replace_na(S1Q13B,0)) ggplot(dat_village, aes(x=S1Q13B)) + geom_histogram(binwidth = 10) + xlab("Se cond largest group proportion") + theme_bw()

That is, a plurality of villages have effectively no second largest group. Only 196 villages have 10% (or more) of its population falling in the second largest group.

For purposes of our study this means that (i) ethnic heterogeneity within villages is much less common than expected, which will challenge our understanding of the problem and potential interventions; and (ii) ethnic diversity is not a suitable criterion to determine which villages to preferentially include. Consequently, we have refocused our understanding of “cleavages” to be more encompassing of livelihood-based and other possible cleavages, rather than using ethnicity as a proxy. Our research design consequently seeks to first determine what the appropriate form of cleavage is, and what the resulting groups are, in each village.

Note that most of the very smallest villages are homogeneous, but there are plenty of homogeneous ones going up into medium sizes.

Mobile money How many people use mobile money? In ths full data, not everybody answered, but among those who did:

response num_responses prop

Not common at all (Below 10% use it) 389 0.96

Sometimes common (10% - 25%) 10 0.02

Somewhat common (25 - 50% use it) 8 0.02

Disputes In this section, we see what we have learned about the nature and extent of disputes occurring in these villages.

In the leader survey analyzed here, for better or worse, the way we asked about “disputes” was broader than the meaning we give that term above, and could lead particpants to answer about “conflicts” instead, including armed clashes instigated either in the village or by outsiders or in their interactions. In this sense, not all of the event types described here are the “disputes” of interest to us, but we learn something about what is going on in these places regardless.

At the village level we have been working with above, response num_responses prop

Access and control of grazing areas 303 0.74

Land rights – land access rights 182 0.45

Family conflicts 92 0.23

Conflicts between neighbors (sometimes because of children) 42 0.10

Chieftaincy conflict 20 0.05

Gender-based violence (violence against women) 8 0.02

Access to water 6 0.01

Community vigilantes against radical groups 5 0.01

Excessive cutting of wood in common areas 3 0.01

Settling of personal accounts (taking justice into oneself) 2 0.00

Islamist groups impose sharia law 2 0.00

Robbery and armed banditry 1 0.00

Thus, out of the 404 villages counted here, in a large majority (74%) the village chief reported disputes over access to and control of grazing areas, with another 45% reporting disputes over land rights or access. We take these two to be overlapping categories. The next most common disputes were “family conflicts” (23%).

Next, we go back to the full data rather than focusing on the village chief/leader responses that generated just one response per village above. In this format, we see the following sorts of disputes:

response num_responses prop

Access and control of grazing areas 552 0.68

Land rights – land access rights 290 0.36

Family conflicts 175 0.21

Conflicts between neighbors (sometimes because of children) 89 0.11

Chieftaincy conflict 40 0.05

Access to water 19 0.02

Gender-based violence (violence against women) 14 0.02

Community vigilantes against radical groups 10 0.01

Settling of personal accounts (taking justice into oneself) 8 0.01

Islamist groups impose sharia law 4 0.00

Excessive cutting of wood in common areas 3 0.00

Robbery and armed banditry 2 0.00

Unequal distribution of NGO donations 2 0.00

Ethnic conflicts 1 0.00

The story here is largely the same – the top three types are the same, and the proportions are similar – though we note that it is complicated by the fact that different villages can have different numbers of observations in this case.

Escalation and violence

How often do these disputes escalate to violence? Counting all disputes, 11% reportedly escalated to violence.

Outside of our data, we see from the Armed Conflict Location and Event Data (ACLED) project that while only a minority of disputes lead to violence, the number of fatalities due to farmer-herder disputes alone has increased significantly since 2015, with 700 such fatalities documented in Mali since 2020. https://africacenter.org/publication/growing-complexity-farmer-herder-conflict-west-central-africa/ https://africacenter.org/publication/growing-complexity-farmer-herder-conflict-west-central-africa/ https://africacenter.org/publication/growing-complexity-farmer-herder-conflict-west-central-africa/

Disputes resolution success or desires

When asked about, with resolved disputes, “Why do you think these conflict management and resolution efforts were effective or ineffective?”, the answers were not very informative. Here we use all responses, and pool across multiple disputes when reported:

tab.all.resolution.why = tablewithlabels(c(dat_village$S4Q11B_D1, dat_village $S4Q11B_D2, dat_village$S4Q11B_D3)) prop = round(as.numeric(tab.all.resolution.why[,2])/nrow(data),2) tab.print = cbind(tab.all.resolution.why[-1],prop) flextable(tab.print %>% rownames_to_column("response")) %>% autofit() %>% theme_box() %>% theme_vanilla() response num_responses prop

Reaching a solution/consensus 174 0.21

Resolution of the problem/conflict 153 0.19

Establish social cohesion/reconciliation 76 0.09

Decrease tension 72 0.09

Decisions of leaders are respected 64 0.08

Good decision making by leaders/leaders 30 0.04

Paying fines to victims 22 0.03

Good management of the committee set up 19 0.02

The problem is not solved 10 0.01

The decisions are not fair 8 0.01

The problem has been taken to court 6 0.01

Tension continues 5 0.01

Decisions are not respected 5 0.01

Stop attacks 4 0.00

Fines paid 4 0.00

Resolution of the problem/conflict through justice 3 0.00

Successfully reframed the children 2 0.00

Don't know 2 0.00

Stop the rackets 1 0.00

The protagonists recognize their responsibility 1 0.00

Attacks continue 1 0.00

Create disagreement between neighbors/family 1 0.00

Not concerned 1 0.00

When asked “what efforts would you like to see to manage or resolve conflicts”, pooling across multiple disputes and in the full dataset, tab.all.resolution.hope = tablewithlabels(c(dat_village$S4Q12_D1, dat_village $S4Q12_D2, dat_village$S4Q12_D3)) prop = round(as.numeric(tab.all.resolution.hope[,2])/nrow(data),2) tab.print = cbind(tab.all.resolution.hope[-1],prop) flextable(tab.print %>% rownames_to_column("response")) %>% autofit() %>% theme_box() %>% theme_vanilla() response num_responses prop

Organize mediation/ talks 137 0.17

Involvement of the village chief and his advisors 117 0.14

Involvement of the notables 63 0.08

Involvement of local authorities 43 0.05

None 42 0.05

Involvement of the Imam 34 0.04

Raising awareness 28 0.03

Financial means for the management of conflicts 21 0.03

Support the chiefs in the mediation process 21 0.03

Compensation of victims 19 0.02

Forgiving each other 19 0.02

Involvement of the heads of families 18 0.02

Delimit plots/fields and grazing areas 18 0.02

Restoring peace and social cohesion 15 0.02

Involvement of ethnic leaders 13 0.02

Training for conflict management 13 0.02

Involvement of religious leaders 10 0.01

Involvement of women's leaders 8 0.01

Establish a system of monitoring and evaluation 8 0.01

Involvement of the police force 5 0.01

Set up a sanction system 4 0.00

Settle conflicts with transparency 3 0.00

Involvement of the justice system 2 0.00

Don't know 2 0.00

Involvement of landowners 1 0.00

Supervision of harvests by young people 1 0.00

Setting up a conflict management committee 1 0.00

Disputes/conflicts and exclusion criteria

Recalling that this set includes some events that are not “disputes” in the way we mean it, some of these events also hit on exclusion criteria we have for the villages to include in the study. For safety and appropriateness reasons, we can remove any village in which any person listed “Islamist groups impose Sharia law”, or “Community vigilants against radical groups”. These are coded as 8 and 11.

Other safety and violence indications We also asked separate questions designed to determine if a village was safe enough to propose working in. S2Q05 asks whether there have been attacks by armed groups or state security forces in the past year:

table(dat_village$S2Q05)

## 0 1 ## 391 16

Regarding what happened in these attacks, tablewithlabels(x=dat_village$S2Q05A)

## code num_responses ## Attaque islamistes avec des déplacés 8 4 ## Bracages des voitures 1 3 ## Bracages d'un jeune 2 2 ## Comba entre islamiste et miliste 3 2 ## Champs bruller 4 1 ## Route barrer par les islamistes 5 1 ## Sabotage du moteur d'irrigation 6 1 ## Attaque du poste de contrôle de police 7 1 ## Bracages des cultivateurs 9 1

Many of these are serious enough to suggest we ought to drop these 16 locations.

S2Q06 asks whether there have been attempts by armed groups to recruit village members in the past year, table(dat_village$S2Q06)

## 0 1 ## 382 25

And S2Q07 asks whether members of armed groups have stayed in the community overnight in the past year, table(dat_village$S2Q07)

## 0 1 ## 395 12

Village selection decisions With the analysis above we can try out a series of decisions about what villages to include.

Given budget projections we anticipate needing 300 villages. Given that some of the villages we choose to include at this stage may need to be dropped for unanticipated reasons, we can target here 315-330.

Most of the decision can be made through selecting places with suitable population sizes.

Let’s start with a filter of 25-800 households.

dat_village = dat_village %>% mutate(pop_okay = (dat_village$S1Q04A>=25 & dat_village$S1Q04A<=800)) dat_village_running = dat_village %>% filter(pop_okay)

The brings us to 353 already.

Second, we would remove those that reported armed attacks:

dat_village_running = dat_village %>% filter(pop_okay==TRUE & armedattack==0)

Which brings us to 340.

Third, the next priority is removing those additional places that reported Islamist attacks or vigilante attacks on radicals, when we asked about disputes.

dat_village_running = dat_village %>% filter(pop_okay==TRUE & armedattack==0 & exclude_village_conflict_type==FALSE)

This brings us to 329. This is what we propose, so we have the possibility of losing nearly 10% of these villages for reasons that emerge while being able to maintain the intervention in 300 villages.