## Appendix A32 — "Fraying Threads: Exclusion and Conflict in Sub-Saharan Africa"

## Source details

**Canonical URL:** [Appendix A32 — "Fraying Threads: Exclusion and Conflict in Sub-Saharan Africa"](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024004-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024004-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024004-print-pdf.pdf.json)

---

### Introduction and research focus
- Exclusion defined as a condition or process through which certain individuals or groups are systematically denied access to rights, opportunities, or resources available to other segments of the population.
- Research objective: identify the role of social, political, and economic exclusion in sparking conflicts in Sub-Saharan Africa (SSA) and identify micro-level correlates to guide policy interventions.
- Empirical analysis conducted at two levels:
  - National level: how perceptions of exclusion (economic, social, political) contribute to conflict and instability; claim that intra-state marginalization exerts a more substantial influence on conflict likelihood than poverty or economic underdevelopment alone.
  - Individual level: use of survey data to identify socioeconomic and governance factors that cultivate feelings of exclusion.
- Measurement approach: construct Perception of Exclusion Index (PEI) and sub-components using factor analysis rather than ad hoc weighted averages.
- Endogeneity approach: empirical model employs lagged explanatory variables to address potential reverse causality (conflicts at time t unlikely to have impacted previous levels of exclusion).

### Conflict in Sub-Saharan Africa: context and quantitative evidence
- SSA characterized by persistently high conflict incidence and displacement; conflicts often begin with extremist operations and expand regionally.
- Food insecurity and climate impacts:
  - Forecasts for 2023 indicate nearly 142 million individuals in the region will confront acute food insecurity, marking a 12 percent elevation from the preceding year.
  - Over one-third of these individuals have succumbed to acute food insecurity following the outbreak of the COVID-19 pandemic.
  - Recent research: fragile states may incur cumulative output losses amounting to four percent in the wake of severe climatic events, compared to one percent loss in other nations.
  - Sahel subregion: proportion of people facing acute food insecurity surged threefold in the preceding three years, rising from 3.6 million to 10.5 million.
  - Global Hunger Index classifications highlighted: Chad "alarming"; other Sahelian nations "serious".
- Recent trends in conflict incidents and fatalities (ACLED-based reporting; latest data as of 19-08-2023 unless otherwise noted):
  - Fatalities: 42,653 in 2022, 11,686 in 2012, and 6,755 in 2010; stated increase of 531.8 percent (2010 to 2022).
  - Conflict incidents: 27,330 in 2022, 5,497 in 2012, and 2,483 in 2010; stated increase of 1000.4 percent (2010 to 2022).
  - Sahel subregion 2022: incidents jumped from 4,170 in 2021 to 6,120 in 2022 (nearly 20 percent increase); recorded fatalities increased from 6,610 in 2021 to 10,740 in 2022 (34 percent increase); incidents concentrated in Burkina Faso and Mali.
  - Geographic concentration: conflict largely near national borders in often under-provisioned "ungoverned" spaces; hotspots include Liptako-Gourma tri-border area and parts of northern Mozambique, South Sudan, Central African Republic, Ethiopia.
  - Coup activity: post-2020 the region has witnessed 10 successful or attempted coups, signaling an escalation after two decades of relative stability.
  - Overall: forty percent of Sub-Saharan Africa is categorized as either fragile or conflict-ridden.
- Conflict type dynamics:
  - Violence against civilians has increased particularly strongly; rebel group involvement rose markedly since 2018.
  - Latest spatial/conflict mapping data noted as of 28-04-2023 for location-based figures.

### Literature review: exclusion and conflict
- Prior studies emphasize ethnic/political marginalization, horizontal inequalities, polarization, and minority exclusion as conflict drivers (Cederman et al. (2010); Stewart (2000); Østby (2008); Gurr (2000); Murshed and Gates (2005); Esteban et al. (2012)).
- SSA-focused findings and gaps:
  - Literature on exclusion’s impact in SSA is limited.
  - Boutellis and Zahar (2017): political and social exclusion significant in Mali conflict.
  - Nyadera and Massaoud (2019): prolonged marginalization and ineffective governance produce ungoverned territories exploited by armed groups; recommend multidimensional reforms to enhance inclusiveness in the Sahel.
  - Other studies emphasize resource capture and climate change as determinants (e.g., Mbaye (2020); Vesco et al. (2020)).

### 4. A New Exclusion Index — methodology and normalization
- Index constructed via confirmatory factor analysis (CFA) on selected V-Dem variables to produce:
  - An overall exclusion index (Perception of Exclusion Index, PEI).
  - Three subindices: social, economic, political.
- Single-factor CFA setup notation (as presented):
  - ExclVarij represent the jth V-Dem variable related to exclusion for country i; ExclIndexi denotes the latent overall exclusion measure for country i.
  - Single-factor model: ExclVari1 − μ1 = λ1 · ExclIndexi + εi1 … ExclVarin − μn = λn · ExclIndexi + εin.
  - Factor loadings λ1, λ2, ..., λn measure relationships between the latent exclusion variable and observed variables.
- Normalization and interpretation:
  - Applied min-max normalization to scale the exclusion index from zero to one.
  - Index range: 0 to 1.
  - Interpretation: higher values indicate greater exclusion; lower values indicate lesser exclusion.
- Subindices constructed by applying CFA separately to variables classified as social, political, or economic.

### Exclusion index empirical patterns and country-level insights (Figures 5–7)
- Figure 5: Perception of Exclusion by Country, 2022
  - Index scaled 0 (optimal) to 1 (most severe); World Bank 2023 FCS classification overlaid.
  - Fragile states tend to manifest elevated exclusion values, with exceptions.
  - Country-level examples:
    - Mauritania and Chad: heightened levels of exclusion relative to West African peers.
    - Niger and Burkina Faso: low values on the exclusion metric despite being classified as fragile states.
    - Equatorial Guinea (GNQ), Madagascar (MDG), Rwanda (RWA), Angola (AGO): not classified as fragile by the World Bank but score high values (above 0.5) on the exclusion measure.
- Figure 6: Exclusion and State Fragility, 2012-2022
  - Positive correlation between average PEI and Fragile State Index (FSI) over 2012-2022.
  - Examples: South Sudan, Chad, Ethiopia (fragile with elevated FSI and high PEI); Botswana, Seychelles (non-fragile with low PEI).
- Figure 7: Exclusion Multidimensions, 2022 (values averaged between 2012 and 2022)
  - Relative prominence of social, economic, political exclusion by country:
    - Burkina Faso, Niger, Mali: escalated social and political exclusion (fragile states).
    - Tanzania and Burkina Faso: exclusion driven substantially by social factors.
    - Seychelles and Mauritius: social exclusion minimal.
    - Nigeria and Mali: economic exclusion consequential.
    - Ghana and Botswana: economic exclusion less intense.
  - Use: indicates where dimension-specific interventions may be most impactful.

### 6. Empirical results — 6.1 The effects of exclusion on conflict (national-level)
- Estimation approaches: pooled OLS, fixed effects, random effects. Dependent variables: Fragility (normalized 0–1), Fatalities (logged), and RSUI (normalized 0–1). Regressors include Exclusion index, Sahel G5 dummy, Uncertainty index, Government Effectiveness, and Exclusion×Sahel G5 interaction.
- Key baseline coefficient estimates (Table 1; specification notes and standard errors in parentheses):
  - Pooled model (Fragility, Model 1): Exclusion = 0.818 ∗∗∗ (0.027).
    - Prose interpretation: "a 1% increase in exclusion results in 0.8% increase in fragility."
  - Random effects (Fragility, Model 3): Exclusion = 0.088 ∗ (0.052).
    - Prose interpretation: "a 1% increase in exclusion results in 0.1% increase in fragility."
  - Pooled model (Fatalities, Model 4): Exclusion = 5.636 ∗∗∗ (0.378).
  - Fixed effects (Log fatalities, Model 5): Exclusion = 7.285 ∗∗∗ (1.812).
  - Sahel G5 dummy (Fragility, pooled): Sahel.G5 = 0.482 ∗∗∗ (0.034).
  - Interaction Exclusion:Sahel.G5 (Fragility, pooled): 0.659 ∗∗∗ (0.059).
    - Prose interpretation: "a 1% increase in exclusion is associated with an additional 0.7% increase in fragility in Sahel G5 compared to other countries."
  - Uncertainty (Fragility pooled): Uncertainty = 0.234 ∗∗∗ (0.019).
  - Government Effectiveness (Fragility pooled): Gov.Effectiveness = −0.150 ∗∗∗ (0.014).
- Observations and R2:
  - Observations: 623 (Fragility models), 722 (Fatalities models), 242 (RSUI models).
  - Example R2: Fragility pooled R2 = 0.546.
- Dynamics from Local Projections (five-year impulse responses; Figure 9):
  - Median response estimates to a 1% rise in the Exclusion index over five years:
    - Fragile State Index (FSI): approximately 0.1 percent.
    - log fatalities: approximately 0.4 percent.
    - Reported Social Unrest Index (RSUI): approximately 0.6 percent.
  - RSUI shows the most significant and immediate response to exclusion, followed by fatalities (longer lag) and then FSI.
- Subindex IRF patterns (Figures 10–12):
  - Social exclusion:
    - Strong positive correlation with FSI that intensifies over time.
    - Most persistent and largest impulse response on RSUI.
    - Initial effect on fatalities is insignificant (delayed response).
  - Economic exclusion:
    - Positively correlated with FSI but less pronounced.
    - Effect on fatalities is inconsistent and occasionally positive.
    - Effect on RSUI similar but slightly smaller than social exclusion.
  - Political exclusion:
    - No clear impact on FSI.
    - Muted or insignificant effect on fatalities.
    - Smallest effect overall; some influence on RSUI but weaker.
- Synthesis:
  - Exclusion has a positive, statistically significant, and persistent effect on conflict measures.
  - Social exclusion is the most critical driver of increased fragility and social unrest.
  - Economic exclusion plays a role but is less pivotal for fatalities.
  - Political exclusion is the least consistent and generally the weakest predictor of conflict outcomes.

### 6. Empirical results — 6.2 The main drivers of exclusion (individual-level)
- Micro-level regressions model positive measures of inclusion: trust in government (ConfNatGvt), perception that country is good for ethnic/racial minorities (GdPlEthMin), and perception that country is good for religious minorities (GdPlRelMin).
- Full-sample key correlates (Table 2; coefficients with standard errors in parentheses; adjusted R2 and observations reported):
  - HealthSatisf = 0.03 ∗∗∗ (0.01) — positive for ConfNatGvt; also positive for GdPlEthMin and GdPlRelMin.
  - EduSatisf = 0.11 ∗∗∗ (0.01) — positive for ConfNatGvt.
  - AirSatisf = 0.04 ∗∗∗ (0.01) — positive for ConfNatGvt.
  - WaterSatisf = 0.04 ∗∗∗ (0.01) — positive for ConfNatGvt.
  - Food Insecure = −0.02 ∗∗∗ (0.01) — negative for ConfNatGvt.
  - ShelterInsecure = −0.03 ∗∗ (0.01) for GdPlEthMin (negative).
  - Corruptionindx = −0.25 ∗∗∗ (0.02) — negative for ConfNatGvt.
  - LawandOrderindx = 0.28 ∗∗∗ (0.02) — positive for ConfNatGvt.
  - IncomeQuint generally no notable effect (e.g., IncomeQuint = −0.003 (0.004) for ConfNatGvt).
  - Social, economic, political indexes included (example: economicindex = 2.69 ∗ (1.59) in ConfNatGvt; politicalindex = −0.93 (1.29) in ConfNatGvt).
  - Adjusted R2: ConfNatGvt = 0.18; GdPlEthMin = 0.08; GdPlRelMin = 0.09.
  - Observations: ConfNatGvt 49,686; GdPlEthMin 49,713; GdPlRelMin 33,328.
- Sahel G5 subsample (Table 3) highlights:
  - HealthSatisf = 0.02 ∗∗∗ (0.01) (ConfNatGvt) and = 0.06 ∗∗∗ (0.01) (GdPlEthMin).
  - EduSatisf = 0.15 ∗∗∗ (0.01) (ConfNatGvt).
  - AirSatisf = 0.07 ∗∗∗ (0.01) (ConfNatGvt) and = 0.14 ∗∗∗ (0.01) (GdPlEthMin).
  - WaterSatisf = 0.06 ∗∗∗ (0.01) (ConfNatGvt).
  - Food Insecure = −0.02 ∗∗∗ (0.01) (ConfNatGvt) and = 0.02 ∗ (0.01) (GdPlEthMin).
  - ShelterInsecure = −0.02 ∗∗∗ (0.01) (ConfNatGvt) and = −0.04 ∗∗∗ (0.01) (GdPlEthMin).
  - Corruptionindx = −0.24 ∗∗∗ (0.02) (ConfNatGvt); LawandOrderindx = 0.25 ∗∗∗ (0.02) (ConfNatGvt).
  - Notable index coefficients in Sahel G5: socialindex = −36.00 ∗∗∗ (3.46) for ConfNatGvt; politicalindex = −2.95 ∗∗∗ (1.29) for ConfNatGvt.
  - Observations: 10,104 and 10,139 for the two Sahel G5 regressions shown. Adjusted R2: 0.21 and 0.13.
- Interpretive summary:
  - Higher satisfaction with public goods and services (health, education, air and water quality) contributes positively to inclusion perceptions.
  - Food insecurity and shelter insecurity contribute negatively to inclusion measures.
  - Institutional quality matters: higher law and order scores positively correlate with inclusion; higher corruption correlates negatively.
  - Individual employment status and income quintile generally have no effect on inclusion measures in these regressions.
  - Results hold in the Sahel G5 subsample, indicating consistent policy relevance.

### Policy implications and priorities (synthesized)
- Addressing perceptions of exclusion is important for conflict prevention and mitigation alongside efforts to manage climate shocks and natural-resource disputes.
- Suggested policy orientation:
  - Target inclusion-enhancing interventions across social, economic, and political domains to reduce alienation that fuels armed mobilization.
  - Strengthen equitable provision of basic public services (clean water, food security, healthcare, education, economic opportunities), especially in border and peripheral regions prone to being "ungoverned".
  - Adopt multidimensional reforms (social and economic) to build resilience in fragile regions such as the Sahel.
  - Prioritize conflict prevention policies that consider both macro-level exclusion indices (PEI) and micro-level drivers revealed by survey data.
  - Emphasize improving institutional quality (reduce corruption, improve law and order) and public-good provision over focusing primarily on macroeconomic growth or unemployment reductions, given stronger micro-level associations with inclusion.

*Source: Appendix A32 and Sections 4 and 6, "Fraying Threads: Exclusion and Conflict in Sub-Saharan Africa" (IMF Working Paper content provided).*

### Appendix A32

### Appendix A32

### Introduction and Research Focus
- Exclusion defined as a condition or process through which certain individuals or groups are systematically denied access to rights, opportunities, or resources available to other segments of the population.
- Perceptions of exclusion amplified by unequal or insufficient provision of basic public services (clean water, food, healthcare, education, economic opportunities).
- Research objective: identify the role of social, political, and economic exclusion in sparking conflicts in Sub-Saharan Africa (SSA) and identify micro-level correlates to guide policy interventions.
- Empirical analysis conducted at two levels:
  - National level: how perceptions of exclusion (economic, social, political) contribute to conflict and instability; claim that intra-state marginalization exerts a more substantial influence on conflict likelihood than poverty or economic underdevelopment alone.
  - Individual level: use of survey data to identify socioeconomic and governance factors that cultivate feelings of exclusion.
- Measurement approach: construct Perception of Exclusion Index (PEI) and sub-components using factor analysis rather than ad hoc weighted averages.
- Endogeneity approach: empirical model employs lagged explanatory variables to address potential reverse causality (conflicts at time t unlikely to have impacted previous levels of exclusion).

### Conflict in Sub-Saharan Africa: Context and Quantitative Evidence
- SSA characterized by persistently high conflict incidence and displacement; conflicts often begin with extremist operations and expand regionally.
- Food insecurity and climate impacts:
  - Forecasts for 2023 indicate nearly 142 million individuals in the region will confront acute food insecurity, marking a 12 percent elevation from the preceding year.
  - Over one-third of these individuals have succumbed to acute food insecurity following the outbreak of the COVID-19 pandemic.
  - Recent research: fragile states may incur cumulative output losses amounting to four percent in the wake of severe climatic events, compared to one percent loss in other nations.
  - Sahel subregion: proportion of people facing acute food insecurity surged threefold in the preceding three years, rising from 3.6 million to 10.5 million.
  - Global Hunger Index classifications highlighted: Chad "alarming"; other Sahelian nations "serious".
- Recent trends in conflict incidents and fatalities (ACLED-based reporting; latest data as of 19-08-2023 unless otherwise noted):
  - Fatalities: 42,653 in 2022, 11,686 in 2012, and 6,755 in 2010; stated increase of 531.8 percent (2010 to 2022).
  - Conflict incidents: 27,330 in 2022, 5,497 in 2012, and 2,483 in 2010; stated increase of 1000.4 percent (2010 to 2022).
  - Sahel subregion 2022: incidents jumped from 4,170 in 2021 to 6,120 in 2022 (nearly 20 percent increase); recorded fatalities increased from 6,610 in 2021 to 10,740 in 2022 (34 percent increase); incidents concentrated in Burkina Faso and Mali.
  - Geographic concentration: conflict largely near national borders in often under-provisioned "ungoverned" spaces; hotspots include Liptako-Gourma tri-border area and parts of northern Mozambique, South Sudan, Central African Republic, Ethiopia.
  - Coup activity: post-2020 the region has witnessed 10 successful or attempted coups, signaling an escalation after two decades of relative stability.
  - Overall: forty percent of Sub-Saharan Africa is categorized as either fragile or conflict-ridden.
- Conflict type dynamics:
  - Violence against civilians has increased particularly strongly; rebel group involvement rose markedly since 2018.
  - Latest spatial/conflict mapping data noted as of 28-04-2023 for location-based figures.

### Literature Review: Exclusion and Conflict
- Prior studies highlight ethnic/political marginalization, horizontal inequalities, polarization, and minority exclusion as conflict drivers:
  - Cederman et al. (2010): politically marginalized ethnic groups more likely to engage in armed rebellion.
  - Stewart (2000): pronounced disparities between culturally-defined groups can fuel social unrest.
  - Østby (2008): polarization and horizontal inequalities associated with increased conflict risk.
  - Gurr (2000): minority groups at risk due to political exclusion and discrimination.
  - Murshed and Gates (2005): economic and social exclusion, particularly in rural areas, contributed to conflict in Nepal.
  - Esteban et al. (2012): link between ethnic distribution measures and social conflict.
- SSA-focused findings and gaps:
  - Literature on exclusion’s impact in SSA is limited.
  - Boutellis and Zahar (2017): political and social exclusion significant in Mali conflict.
  - Nyadera and Massaoud (2019): prolonged marginalization and ineffective governance produce ungoverned territories exploited by armed groups; recommendation for multidimensional reforms to enhance inclusiveness in the Sahel.
  - Many studies have emphasized other determinants (resource capture, climate change), e.g., Mbaye (2020); Vesco et al. (2020) review linking natural resources to conflicts.

### Methodology and Measures
- Construction of Perception of Exclusion Index (PEI) and sub-components using factor analysis for statistically-systematic indices of exclusion.
- Empirical strategy:
  - National-level regressions linking PEI to conflict likelihood, using lagged explanatory variables to mitigate reverse causality concerns.
  - Individual-level survey analysis to identify socioeconomic and governance correlates of exclusion perceptions.
- Rationale: perceptions of exclusion (whether justified or not) increase conflict frequency and intensity independent of whether exclusion arises from intentional discriminatory policies.

### Key Findings and Interpretations (as presented)
- Perceptions of exclusion correlate significantly with classification of some SSA countries as Fragile and Conflict-affected States (FCS).
- Intra-state marginalization (social, economic, political exclusion) shown empirically to exert more substantial influence on conflict likelihood than poverty or economic underdevelopment alone.
- Conflict escalation and diffusion patterns indicate that exclusion dynamics interact with climatic shocks, food insecurity, and resource-related disputes to amplify fragility.

### Policy Implications (framing and priorities)
- Addressing perceptions of exclusion is important for conflict prevention and mitigation alongside efforts to manage climate shocks and natural-resource disputes.
- Suggested policy orientation:
  - Target inclusion-enhancing interventions across social, economic, and political domains to reduce alienation that fuels armed mobilization.
  - Strengthen equitable provision of basic public services (clean water, food security, healthcare, education, economic opportunities), especially in border and peripheral regions prone to being "ungoverned".
  - Adopt multidimensional reforms (social and economic) to build resilience in fragile regions such as the Sahel.
  - Prioritize conflict prevention policies that consider both macro-level exclusion indices (PEI) and micro-level drivers revealed by survey data.

*Source: Appendix A32, "Fraying Threads: Exclusion and Conflict in Sub-Saharan Africa" (IMF Working Paper content provided).*

### 4.  A New Exclusion Index

### 4.  A New Exclusion Index

### Methodology: factor-analytic construction
- Constructed an overall exclusion index and three subindices (social, economic, political) using confirmatory factor analysis (CFA) on a long list of selected variables from the Varieties of Democracy (V-Dem) dataset.
- Notation and single-factor CFA setup (as presented):
  - Let ExclVarij represent the jth V-Dem variable related to exclusion for country i, and ExclIndexi denote the latent overall exclusion measure for country i.
  - Single-factor model (Equation (1)):
    - ExclVari1 − μ1 = λ1 · ExclIndexi + εi1
    - …
    - ExclVarin − μn = λn · ExclIndexi + εin
  - Factor loadings λ1, λ2, ..., λn measure relationships between the latent exclusion variable and observed variables.
- Rationale: CFA streamlines extensive V-Dem information and reveals underlying correlations among variables, in contrast to prior ad hoc weighting approaches.

### Normalization and interpretation
- Applied min-max normalization to scale the exclusion index from zero to one.
  - Index range: 0 to 1.
  - Interpretation: higher values indicate greater exclusion; lower values indicate lesser exclusion.
- The index is intended to be a comprehensive, comparable measure summarizing multidimensional exclusion across countries.

### Subindices: social, political, economic
- Selected variables were classified into categories (social, political, economic) and the CFA procedure repeated separately to generate subindices that measure those specific dimensions.
- Purpose: to obtain nuanced assessments of different facets of exclusion and enable finer cross-country comparisons of which dimension predominates.

### Empirical patterns and illustrative findings (visualized in Figures 5–7)
- Figure 5: Perception of Exclusion by Country, 2022
  - Index scaled from 0 (optimal) to 1 (most severe); bar height shows country-level exclusion in 2022.
  - Bars colored according to the World Bank’s 2023 classification of Fragile and Conflict-Affected States (FCSs).
  - General pattern: states classified as fragile tend to manifest elevated values on the exclusion index, with exceptions.
  - Country-level examples and notable observations:
    - Mauritania and Chad: heightened levels of exclusion relative to West African peers.
    - Niger and Burkina Faso: low values on the exclusion metric despite being classified as fragile states.
    - Equatorial Guinea (GNQ), Madagascar (MDG), Rwanda (RWA), and Angola (AGO): not classified as fragile by the World Bank but score high values (above 0.5) on the exclusion measure, indicating potential future conflict risks.
- Figure 6: Exclusion and State Fragility, 2012-2022
  - Shows correlation between average values of the Perception of Exclusion Index and the Fragile State Index (FSI) over 2012-2022.
  - Color differentiation distinguishes fragile and non-fragile states per World Bank classification.
  - Finding: positive correlation between the exclusion index and the FSI, suggesting exclusion as a potential precursor to state fragility.
  - Examples:
    - South Sudan, Chad, Ethiopia: categorized as fragile with elevated FSI scores and high exclusion index values.
    - Botswana, Seychelles: non-fragile with low exclusion index values.
- Figure 7: Exclusion Multidimensions, 2022 (values averaged between 2012 and 2022)
  - Displays relative prominence (not absolute levels) of social, economic, and political exclusion within each SSA country.
  - Insights:
    - Burkina Faso, Niger, Mali: escalated levels of social and political exclusion (all fragile states).
    - Tanzania and Burkina Faso: exclusion driven substantially by social factors.
    - Seychelles and Mauritius: social exclusion minimal.
    - Nigeria and Mali: economic exclusion consequential.
    - Ghana and Botswana: economic exclusion manifests with lesser intensity.
  - Use: highlights where interventions targeting specific dimensions (social, political, economic) may be most impactful.

### Key methodological links to subsequent analysis
- The exclusion index and subindices constructed here are used as main regressors in national-level and individual-level empirical analyses:
  - National-level analysis uses the Exclusion Index (Section 4) as main control variable of interest for models in Section 5 and results in Section 6.1.
  - Individual-level regressions (Section 6.2) use the Exclusion subindices as explanatory variables to uncover determinants of perceptions of exclusion.
- The index construction enables robust empirical strategies (e.g., local projections, lagged regressors) to address endogeneity concerns when estimating effects of exclusion on conflict.

*Source: IMF Working Paper — Fraying Threads: Exclusion and Conflict in Sub-Saharan Africa — Section 4. A New Exclusion Index*

### 6.  Empirical Results

### 6. Empirical Results

### 6.1 The Effects of Exclusion on Conflict
- Baseline estimations use pooled OLS, fixed effects, and random effects. Dependent variables: Fragility (normalized 0–1), Fatalities (logged), and RSUI (normalized 0–1). Regressors include Exclusion index, Sahel G5 dummy, Uncertainty index, Government Effectiveness, and Exclusion×Sahel G5 interaction.
- Key baseline findings (Table 1, pooled/random effects highlights):
  - Pooled model (Fragility, Model 1): Exclusion = 0.818 ∗∗∗ (0.027).
    - Interpretation in prose: "a 1% increase in exclusion results in 0.8% increase in fragility."
  - Random effects (Fragility, Model 3): Exclusion = 0.088 ∗ (0.052). (Random effects still suggest positive effect; prose states "a 1% increase in exclusion results in 0.1% increase in fragility.")
  - Pooled model (Fatalities, Model 4): Exclusion = 5.636 ∗∗∗ (0.378).
  - Fixed effects (Log fatalities, Model 5) shows a strong positive effect of exclusion (coefficient reported in table: Exclusion = 7.285 ∗∗∗ (1.812)).
  - Sahel G5 dummy (Fragility, pooled): Sahel.G5 = 0.482 ∗∗∗ (0.034).
  - Interaction Exclusion:Sahel.G5 (Fragility, pooled): 0.659 ∗∗∗ (0.059).
    - Interpretation in prose: "a 1% increase in exclusion is associated with an additional 0.7% increase in fragility in Sahel G5 compared to other countries."
  - Uncertainty is consistently positive and significant across models (e.g., Fragility pooled: Uncertainty = 0.234 ∗∗∗ (0.019)).
  - Government Effectiveness is consistently negative and significant for Fragility in pooled and random effects (e.g., Fragility pooled: Gov.Effectiveness = −0.150 ∗∗∗ (0.014)).
  - Observations and R2:
    - Observations: 623 (Fragility models), 722 (Fatalities models), 242 (RSUI models).
    - R2 values shown per model (e.g., Fragility pooled R2 = 0.546).
- Dynamics from Local Projections (LPIRFs, five-year impulse responses; Figure 9):
  - Median response estimates to a 1% rise in the Exclusion index over five years:
    - Fragile State Index (FSI): approximately 0.1 percent.
    - log fatalities: approximately 0.4 percent.
    - Reported Social Unrest Index (RSUI): approximately 0.6 percent.
  - RSUI shows the most significant and immediate response to exclusion, followed by fatalities (longer lag) and then FSI (less pronounced but still significant).
- Sub-index IRF patterns (Figures 10–12):
  - Social exclusion:
    - Strong positive correlation with FSI that intensifies over time.
    - Most persistent and largest impulse response on RSUI.
    - Initial effect on fatalities is insignificant (delayed response).
  - Economic exclusion:
    - Positively correlated with FSI but less pronounced.
    - Effect on fatalities is inconsistent and occasionally positive.
    - Effect on RSUI similar but slightly smaller than social exclusion.
  - Political exclusion:
    - No clear impact on FSI.
    - Muted or insignificant effect on fatalities.
    - Smallest effect overall; some influence on RSUI but weaker.
- Synthesis of section findings:
  - Exclusion has a positive, statistically significant, and persistent effect on conflict measures.
  - Social exclusion is the most critical driver of increased fragility and social unrest.
  - Economic exclusion plays a role but is less pivotal for fatalities.
  - Political exclusion is the least consistent and generally the weakest predictor of conflict outcomes.
- Policy implication highlighted:
  - Policies aiming to reduce conflict should prioritize reducing social exclusion and related economic exclusions, given their stronger and more persistent effects on unrest and fragility.

### 6.2 The Main Drivers of Exclusion
- Micro-level regressions (Tables 2 and 3) model positive measures of inclusion: trust in government (ConfNatGvt), perception that country is good for ethnic/racial minorities (GdPlEthMin), and perception that country is good for religious minorities (GdPlRelMin).
- Key correlates of inclusion (Table 2, full sample; coefficients with standard errors in parentheses):
  - HealthSatisf = 0.03 ∗∗∗ (0.01) (positive for ConfNatGvt; also positive for GdPlEthMin and GdPlRelMin).
  - EduSatisf = 0.11 ∗∗∗ (0.01) (positive for ConfNatGvt).
  - AirSatisf = 0.04 ∗∗∗ (0.01) (positive for ConfNatGvt).
  - WaterSatisf = 0.04 ∗∗∗ (0.01) (positive for ConfNatGvt).
  - Food Insecure = −0.02 ∗∗∗ (0.01) (negative for ConfNatGvt).
  - ShelterInsecure shows negative relations for some inclusion measures (e.g., −0.03 ∗∗ (0.01) for GdPlEthMin).
  - Corruptionindx = −0.25 ∗∗∗ (0.02) (negative for ConfNatGvt).
  - LawandOrderindx = 0.28 ∗∗∗ (0.02) (positive for ConfNatGvt).
  - IncomeQuint generally no notable effect (e.g., IncomeQuint = −0.003 (0.004) for ConfNatGvt).
  - Social, economic, political indexes included (coefficients vary; economicindex = 2.69 ∗ (1.59) in ConfNatGvt; politicalindex = −0.93 (1.29) in ConfNatGvt).
  - Adjusted R2: ConfNatGvt = 0.18; GdPlEthMin = 0.08; GdPlRelMin = 0.09.
  - Observations: ConfNatGvt 49,686; GdPlEthMin 49,713; GdPlRelMin 33,328.
- Sahel G5 subsample (Table 3) shows similar patterns:
  - HealthSatisf = 0.02 ∗∗∗ (0.01) (ConfNatGvt) and = 0.06 ∗∗∗ (0.01) (GdPlEthMin).
  - EduSatisf = 0.15 ∗∗∗ (0.01) (ConfNatGvt).
  - AirSatisf = 0.07 ∗∗∗ (0.01) (ConfNatGvt) and = 0.14 ∗∗∗ (0.01) (GdPlEthMin).
  - WaterSatisf = 0.06 ∗∗∗ (0.01) (ConfNatGvt).
  - Food Insecure = −0.02 ∗∗∗ (0.01) (ConfNatGvt) and = 0.02 ∗ (0.01) (GdPlEthMin) — note direction differs by outcome.
  - ShelterInsecure = −0.02 ∗∗∗ (0.01) (ConfNatGvt) and = −0.04 ∗∗∗ (0.01) (GdPlEthMin).
  - Corruptionindx = −0.24 ∗∗∗ (0.02) (ConfNatGvt); LawandOrderindx = 0.25 ∗∗∗ (0.02) (ConfNatGvt).
  - Social, economic, political index coefficients differ in sign and magnitude for Sahel G5 (e.g., socialindex = −36.00 ∗∗∗ (3.46) for ConfNatGvt; politicalindex = −2.95 ∗∗∗ (1.29) for ConfNatGvt).
  - Observations: 10,104 and 10,139 for the two Sahel G5 regressions shown. Adjusted R2: 0.21 and 0.13.
- Interpretive summary:
  - Higher satisfaction with public goods and services (health, education, air and water quality) contributes positively to inclusion perceptions.
  - Food insecurity and shelter insecurity contribute negatively to inclusion measures.
  - Institutional quality matters: higher law and order scores positively correlate with inclusion; higher corruption correlates negatively.
  - Individual employment status and income quintile generally have no effect on inclusion measures in these regressions.
  - Results hold in the Sahel G5 subsample, suggesting consistent policy relevance across the region.
- Policy implications drawn:
  - Governments should prioritize improving the quality of institutions (reduce corruption, improve law and order) and provision of public goods (healthcare, education, air and water quality, food and shelter sufficiency) rather than focusing primarily on macroeconomic variables such as levels of economic growth and unemployment.
  - Because social and economic factors are the most influential drivers of exclusion, policy efforts to reduce conflict should emphasize mitigating social exclusion and related economic exclusions.
  - Political reforms, while important, may be the most difficult to implement and appear least likely to reduce incidence of unrest and destabilizing violence according to these empirical results.

*Source: 6. Empirical Results, Fraying Threads: Exclusion and Conflict in Sub-Saharan Africa (WP/2024/004).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024004-print-pdf.pdf_