## 2–7: The Urgency of Conflict Prevention — Macroeconomic Perspective (selected sections)

## Source details

**Canonical URL:** [2–7: The Urgency of Conflict Prevention — Macroeconomic Perspective (selected sections)](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024256-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024256-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024256-print-pdf.pdf.json)

---

### H3: Catastrophic economic consequences of the armed conflict trap — overview and objectives
- Macroeconomic policy can play an important role in armed conflict prevention.
- Prevention defined as policies, programs, and projects that strengthen institutions, promote inclusive economic growth, and support local community development.
- Two main challenges:
  - Assessing likelihood of conflict in peaceful countries is hard.
  - Uncertainty about which macroeconomic policy levers reduce conflict risk and how they affect violence and the economy.
- Analytical approach: integrate forecasting and policy decision-making via a quantitative model that models dynamics of conflict risk through distinct conflict risk stages.

### H3: Conflict risk stages model — structure and empirical facts
- Stages distinguished and summarized into 4 groups for presentation:
  - Very low risk: stages 1 to 4.
  - High risk with little or no violence: stages 5 and 6.
  - High risk with recent violence: stages 7 to 10.
  - High violence: stages 11 and 12.
- Transition dynamics and persistence:
  - From Very low risk there is a 97% likelihood to remain next quarter; a 2% likelihood of transitioning to High risk with little or no recent violence.
  - Strong diagonal dominance: example likelihood of starting in stage 1 and ending in stage 1 is over 93%.
  - For a country in stage 4, likelihood of moving to any of stages 5-10 is 9%, while likelihood of moving to any of stages 11-12 is 0.2%.
  - The conflict trap: once in High violence, countries cannot directly return to Very low risk; escape typically requires extended periods of elevated risk.
- Mortality intensity (stage averages):
  - Stage 11 average: 6 deaths per 1 million inhabitants.
  - Stage 12 average: 71 deaths per 1 million inhabitants.
- Data and modeling inputs:
  - UCDP violence data from 1989 to today; close to 6 million newspaper articles used.
  - Forecasts: 12-month-ahead outbreak risk via Random Forest trained on January 1989–December 1999 with rolling out-of-sample predictions from January 2000.
  - Aggregation: Hidden Markov Model using PyHHMM; HMM hyperparameters chosen by experimentation and judgment.

### H3: Macroeconomic correlates and quantified damages
- Cross-sectional correlation: Pearson correlation between GDP per capita and the 12 risk stages: -0.40.
- Stylized group comparisons (sample period 2000–2023; about 35 countries in the conflict trap versus 130 countries that avoided it):
  - GDP per capita: countries that avoided the conflict trap rose, on average, from around US$9,000 to US$13,000 in the last two decades.
  - Gross fixed capital formation to GDP: about 20% in conflict trap countries versus 30% in countries that avoided the conflict trap.
  - Government revenues: below 20% of GDP on average in conflict trap countries versus close to 25% of GDP in countries that avoided the conflict trap.
  - Current accounts: more volatile in conflict trap countries.
- Two-way fixed effect OLS regression GDP growth impacts (one year in stage → expected GDP change):
  - Spending 1 year in any of stages 1–4: would not experience GDP losses in excess of 1% on average.
  - Spending 1 year in stage 11: expected GDP to contract by 2.7% on average.
  - Spending 1 year in stage 12: expected GDP to contract by 9.2% on average.
- Additional empirical findings cited (cross-country literature):
  - Collier (1999): civil conflict correlated with contemporaneous reduction of GDP per capita growth of 2.2%.
  - Abadie and Gardeazabal (2003): Basque terror campaign reduced per capita GDP by approximately 10%.
  - Blomberg and Hess (2006): presence of violence equivalent to a 30% tariff on international trade.
  - Martin, Mayer and Thoenig (2008): conflict reduces trade by 20 –25%.
  - Couttenier, Monnet and Piemontese (2022): spillovers into production networks generate over 70% of total economic damage from conflict.
  - Collier et al. (2003): during civil wars countries increase military expenditure from 2.8 to 5% of GDP.
  - Margalef and Mueller (2023): trap dynamics can explain a relative GDP per capita decline of around 50% for the worst affected countries.

### H3: DEWAM (Dynamic Early-Warning Action Model) — mechanics and valuations
- DEWAM integrates monthly forecasts with a dynamic optimal policy simulation and a model of the conflict trap using Dynamic Programming.
- Decision framework:
  - Policymaker faces known transition matrix, damages per stage, and intervention costs.
  - Interventions incur immediate costs and shift transition probability masses to reduce escalation risk.
  - Trade-off: act early on weak signals (false positives accepted) versus react post-outbreak at higher cost.
- Damage valuation inputs:
  - Regression outcomes of GDP loss, fatalities, ODA flows (emergency response and peace/security), and refugee flows.
  - Value of Statistical Life (VSL) assumed at $900,000.
  - Refugee per capita costs assumed to be $100.
- Intervention cost calibrations:
  - Fixed cost proportional to population at $3 per person per year ($0.25 per person per month).
  - Variable cost proportional to expected fatalities with three per-fatality cost values tested: $10,000; $40,000; $200,000 per fatality.
  - Example country-level costs for a Pakistan-sized population:
    - Stage 4 policies might cost $164mn every quarter.
    - Stage 12 interventions would cost $9.45bn every quarter (variable cost assumption of $200,000 per fatality).
- Baseline policy effectiveness calibration: 10% (noted as conservative; other work suggests up to close to 50% reductions may be possible).

### H3: Returns to prevention — simulation outcomes and sensitivity
- Sample considered: countries in the bottom 75% ranked by GDP per capita.
- Neutral/central findings (policy instruments targeted monthly, initiated within a quarter):
  - Returns range from $26 to $75 per $1 spent on prevention for countries that have not recently suffered from violence.
  - For countries at High risk that have suffered from violence recently, returns could be as high as $103 per $1 spent on prevention.
  - Prevention policies in High violence countries: return of $27 per $1 spent on prevention (reported elsewhere in the paper).
- Comparison with prior work:
  - World Bank (2018) found $1 of preventative activities generates on average $16 of gains; DEWAM results higher due to persistence, high damages, and higher post-crisis intervention costs.
- Appendix B robustness and scenario structure:
  - Policy effectiveness levels tested: 2%, 5%, and 10%.
  - Discount rate for BCRs: 4%.
  - Three scenarios:
    - Pessimistic: 2% effectiveness; post-conflict variable cost $200,000 per fatality.
    - Neutral: 5% effectiveness; post-conflict variable cost $40,000 per fatality.
    - Optimistic: 10% effectiveness; post-conflict variable cost $10,000 per fatality.
  - Neutral scenario result highlighted: in stage 8 each $1 spent on prevention saves a monetary equivalent of $201.
  - Reported BCRs are averages across bottom 75% GDP per capita countries; BCR < 1 implies not cost-effective.
  - Including GDP growth effects markedly increases BCRs; excluding growth reduces the case for true prevention.

### H3: Empirical identification of macroeconomic policy effects — methodology and key results
- Policy shock construction:
  - WEO forecasts published in April and October from 1990 to 2022; forecast error = difference between current WEO edition forecast and previous edition’s forecast, normalized by GDP.
  - Shock binary cutoffs:
    - Change in public expenditure (percent of GDP): ≤ −1.1 (Cutoff percentile 10).
    - Government balance (percent of GDP): ≥ 1.5 (Cutoff percentile 90).
    - Unemployment rate: ≤ −.81 (Cutoff percentile 10).
- Estimation approach: matched difference-in-difference following Mueller and Rauh (2024) and Callaway and Sant’Anna (2021) with matching on predicted intensity pre-treatment and estimation window [−4, 4] semesters.
- Key empirical findings:
  - Positive shocks to the overall fiscal balance:
    - Improvement in fiscal balance by 1.5% of GDP (top 10 percentile relative to forecast) leads to reduction in conflict deaths by 9% (post-treatment).
    - Empirical-calibration-consistent policy effectiveness used in simulations implies violence reduction between 7 and 10%.
  - Labor market shocks:
    - Likelihood of violence falls by 1.1 percentage points following a drop in the unemployment rate of about 0.81 percentage points (bottom 10th percentile).
    - Severity of violence falls by 5.2% following such unemployment declines.
  - Public expenditure cuts:
    - Cuts of at least 1.1% of GDP (lowest 10th percentile) increase the share of months in conflict by 2 percentage points and increase fatalities by 8.5%.
  - IMF program approvals:
    - Approvals result in slight reduction in likelihood of any battle death by about 1.5 percentage points on average; effect amplifies after around a year to about 4 percentage points.
  - Coverage gap in FCS list:
    - High violence stages: FCS 0.86, Non FCS 0.14.
    - High risk with little or no recent violence: FCS 0.23, Non FCS 0.77.
    - High risk with recent violence: FCS 0.23, Non FCS 0.77.
    - Very low risk: FCS 0.03, Non FCS 0.97.
    - Implication: 77% of High risk cases are not on the World Bank FCS list.

### H3: Policy implications and recommendations (summary)
- Fiscal policy design:
  - Improving fiscal balance matters for prevention, but improvements should not come at the expense of productive expenditures.
  - Increasing revenues (revenue mobilization and capacity development) is particularly valuable.
  - Fiscal consolidation via sharp public expenditure cuts can worsen incidence and severity of conflict.
- Labor market and inclusion:
  - Policies that foster a resilient labor market (employment guarantees, employment programs) reduce conflict risk and severity.
  - Addressing exclusion and increasing employment opportunities directly reduces fragility.
- State capacity and governance:
  - Building state capacity and strengthening the social contract has large direct and indirect economic benefits through violence reductions.
  - Program design should include institution-building goals (modernize revenue administrations, tax auditing) while avoiding increased exclusion.
  - Anti-corruption measures and accountability (e.g., independent watchdogs, audits) can bolster legitimacy and revenue mobilization.
- International engagement:
  - IMF-supported programs and international assistance can reduce conflict risk by boosting growth, catalyzing donor support, and raising revenues.
  - Targeting prevention efforts toward High risk but non-violent settings (many not on FCS lists) could yield large returns.
- Operational guidance for IFIs and partners:
  - Adopt early warning systems and internal processes to identify high-risk non-violent settings.
  - Internalize returns to prevention policies across all high-risk situations, not only countries already classified as fragile.
  - Resist “blood leads” phenomena; prioritize prevention even where outbreaks have not materialized.

*Italic: Source — Excerpted content from wpiea2024256-print-pdf (IMF Working Paper — The Urgency of Conflict Prevention – A Macroeconomic Perspective; sections 2, 3.1, 4, 5.3, 7, and Appendix B).*

### 2. T

### 2. THE CATASTROPHIC ECONOMIC CONSEQUENCES OF THE ARMED CONFLICT TRAP

### Overview and objectives
- The paper argues that macroeconomic policy can play an important role in armed conflict prevention.
- Prevention is defined as policies, programs, and projects that strengthen institutions, promote inclusive economic growth, and support local community development (United Nations and World Bank, 2018).
- Two challenges for making prevention an objective of macroeconomic policy:
  - Assessing the likelihood of conflict in peaceful countries is hard (Mueller and Rauh 2022a).
  - Uncertainty about which macroeconomic policy levers reduce conflict risk and how they affect both violence and the economy.
- The analysis integrates forecasting and policy decision-making via a quantitative model (Mueller, Rauh, Ruggieri and Seimon (2024a)) that models dynamics of conflict risk through distinct conflict risk stages.

### The conflict risk stages model (conceptual framework)
- Stages distinguished: Very low risk, High risk with little/no recent violence, High risk with recent violence, and High violence.
- Transition dynamics example: from Very low risk there is a 97% likelihood to remain next quarter, a 2% likelihood of transitioning to High risk with little or no recent violence; likelihoods <1% are not shown.
- Once in High violence, countries cannot directly return to Very low risk but must transition through High risk stages — the essence of the "conflict trap".
- The complete framework links conflict risk stages with data on economic damages (GDP), fatalities, displacement, and humanitarian aid.
- Policies are modeled as costly manipulations of transition probabilities between stages, enabling trade-offs between prevention costs and avoided macroeconomic damages over dynamic horizons.

### Macroeconomic costs of conflict (empirical background highlighted)
- Cross-country and regional estimates cited:
  - Collier (1999): civil conflict correlated with a contemporaneous reduction of GDP per capita growth of 2.2%.
  - Abadie and Gardeazabal (2003): Basque terror campaign reduced per capita GDP by approximately 10%.
  - Mueller (2016): supports per-capita measurement of violence intensity for macroeconomic damage studies.
- Mechanisms and magnitudes:
  - Violence destroys resources and human capital, causes long-term impacts (lost schooling, mental scars), severs transport links, and generates strong spillovers across production networks.
  - Blomberg and Hess (2006): presence of violence is equivalent to a 30% tariff on international trade.
  - Martin, Mayer and Thoenig (2008): conflict reduces trade by 20 –25%.
  - Couttenier, Monnet and Piemontese (2022): spillovers into production networks generate over 70% of total economic damage from conflict.
  - Collier et al. (2003): during civil wars countries increase military expenditure from 2.8 to 5% of GDP.
  - Margalef and Mueller (2023): trap dynamics can explain a relative GDP per capita decline of around 50% for the worst affected countries.
- Anticipation effects:
  - Anticipation or risk of violence raises uncertainty, lowers investment, and depresses asset prices (Bloom 2009; Baker, Bloom and Davis 2016; Besley and Mueller 2018; Besley and Mueller 2012; Tapsoba 2023; Elster, Zussman and Zussman 2017).
  - The conflict trap reinforces poverty traps via higher risk premia, lower savings and investment, increased fixed costs, and worsening resource traps (Chami et al. 2021a).

### Dynamic returns to prevention and policy-relevant quantitative findings
- Returns to prevention (including dynamic impacts) are large:
  - For countries that have not suffered recently from violence: returns range from $26 to $75 per $1 spent on prevention.
  - For countries with recent violence: returns could be as high as $103 per $1 spent on prevention.
  - Prevention policies in High violence countries: return of $27 per $1 spent.
- Empirical links between macroeconomic variables and conflict risk:
  - Higher fiscal balance associated with lower conflict risk; however, conflict risk increases following expenditure cuts.
  - A resilient labor market reduces conflict risk; higher unemployment increases likelihood and intensity of violence.
- International engagement (IMF programs):
  - IMF programs tend to occur in stages of high conflict risk.
  - After controlling for endogeneity, IMF programs lead to reductions in the likelihood of violence by 1.5–4 percentage points on average.
- Coverage and risk assessment implication:
  - 77% of high-risk countries with recent violence were not on the World Bank-IMF FCS list, indicating potential undercoverage by existing fragility classifications and a need for more nuanced risk assessment.

### Policy implications emphasized
- Fiscal policy design:
  - Improving fiscal balance matters for prevention, but efforts to improve fiscal balance should not come at expense of productive expenditures.
  - Increasing revenues (revenue mobilization and capacity development) is particularly valuable.
- Labor market and growth policies:
  - Policies that foster a resilient labor market increase the opportunity cost of taking up arms and reduce conflict risk.
- International support:
  - International engagement, including IMF programs and capacity development, can help build state capacity and reduce conflict risk.
  - More attention to prevention in non-violent but high-risk settings could yield large benefits given the high returns to prevention.
- Broader coordination:
  - Macroeconomic policy must be planned in the broader conflict context and aligned with peace maintenance; economic gains are unsustainable if they jeopardize peaceful equilibria between opposition and state entities.

*Source: IMF Working Paper — The Urgency of Conflict Prevention – A Macroeconomic Perspective (sections 1–3 and Figures/tables referenced).*

### 3.1 The stages of conflict

### 3.1 The stages of conflict

### Data and definition of conflict
- Conflict measure: Uppsala Conflict Data Program (UCDP) definition — contested incompatibility concerning a government or territory where the use of armed force between two parties (one of them the government) results in at least 25 battle related deaths in a calendar year.
- Time coverage and inputs:
  - UCDP violence data available from 1989 to today.
  - UCDP provides candidate data for recent fatalities updated monthly; treated as true UCDP fatality observations.
  - All kinds of conflict fatalities coded by UCDP are incorporated.
  - Model inputs include UCDP data and information from close to 6 million newspaper articles.

### Modeling approach: forecasts and aggregation
- Forecasting sub-models:
  - ConflictForecast sub-models (Mueller and Rauh (2022a)) produce pseudo-out-of-sample forecasts for 12-month-ahead conflict outbreak risk using information at time T.
  - Forecasts generated by training a Random Forest using data from January 1989 to December 1999.
  - Rolling out-of-sample predictions from January 2000 onwards.
- Aggregation into stages:
  - Forecasts are aggregated using a Hidden Markov Model (HMM) to estimate conflict risk stages.
  - HMM setup: observable variables O (the forecasts) generated by hidden states Z; transitions follow a first-order Markov chain with start probability vector π and transition matrix A.
  - Key computational tasks: Likelihood (P(O|λ)), Decoding (best hidden state sequence Z), Learning (estimate λ) — solved via Viterbi, Forward-Backward, and Baum-Welch algorithms respectively.
  - HMM state-space is discrete; number of states is a hyperparameter chosen through experimentation and ad-hoc judgment to distill the cycle of peace and conflict.
- Specific model inputs to HMM (O):
  - 12 month ahead outbreak risk forecasts from the best (historical violence and text features) and text-only models,
  - 12 month ahead intensity forecast,
  - a polynomial of degree three describing predicted risk of conflict re-emergence using months since last armed conflict (captures high risk immediately after conflict that decays exponentially but remains relatively high for several years).
- Implementation detail: Python package PyHHMM is used for HMM implementation.

### Definition and interpretation of stages
- Countries assigned to a stage monthly for January 2000 to April 2023.
- 12-stage model summarized into 4 groups for presentation:
  - Very low risk: stages 1 to 4
  - High risk with little or no violence: stages 5 and 6
  - High risk with recent violence: stages 7 to 10
  - High violence: stages 11 and 12
- Notes on interpretation:
  - Stages summarize likelihood of conflict and expected fatalities; they are not directly observed and do not claim causality.
  - Ranking of stages is derived from dynamic summary of likelihood of conflict and number of fatalities within the HMM; stages rank higher if more likely to escalate into stages 11 and 12.
  - In stages 1 to 4, predicted fatalities and escalation risk are very low; deaths per capita are 0 on average in these stages.
  - Stages 5 and 6: peaceful for years but elevated outbreak risk.
  - Stages 7 to 10: elevated risk driven by recent violence and susceptibility to recurrence.
  - Stages 11 and 12: ongoing violence differentiated by intensity:
    - Stage 11 average: 6 deaths per 1 million inhabitants.
    - Stage 12 average: 71 deaths per 1 million inhabitants.
  - Example country patterns:
    - Mexico: extended periods in stage 11 (homicides associated with organized crime).
    - Afghanistan, Syria, Central African Republic: frequently in stage 12.
- Exceptions and data notes:
  - 98% of observations in stages 7-10 experienced armed conflict within the last 3 years; exceptions exist (example: Lebanon in stage 8 from March 2005 to March 2006).
  - Table summaries report mean values across observations conditional on stage; distributions may be driven by outliers.

### Transition dynamics and the conflict trap
- Transition matrix constructed from transitions observed January 2000 to April 2023; visualized in Figure 4a (12-stage) and summarized in Figure 4b (4-group).
- Key empirical observations from transition dynamics:
  1. Stages are persistent:
     - Strong diagonal dominance in transition matrix.
     - Example: likelihood of starting in stage 1 and ending in stage 1 is over 93%.
  2. Peace is stable:
     - Rare for countries to move directly from stable peace to intense conflict.
     - Example: for a country in stage 4, likelihood of moving to any of stages 5-10 is 9%, while likelihood of moving to any of stages 11-12 is only 0.2%.
  3. The conflict trap:
     - Rare transitions from intense conflict to stable peace; escape typically requires extended periods of elevated risk before settling into peaceful stages.
     - This persistence and sequencing define the “conflict trap.”
- Modeling implications:
  - Stages in the low-risk range (1 to 6) are largely clusters of risk data coming from machine learning; forecasts are necessary to distinguish these stages.
  - Policies can causally affect transition likelihoods (example: power-sharing agreements produce systematic forecast errors for 12-month-ahead forecasts, indicating causal impact on transitions).
  - Persistent stage dynamics and the conflict trap imply dynamic benefits to prevention (keeping countries in low-risk stages generates significant downstream benefits).

### Macroeconomic correlates and costs of conflict
- Cross-sectional relationships:
  - Pearson correlation between GDP per capita and the 12 risk stages: -0.40.
- Group definitions for macro comparisons:
  - Countries in the “conflict trap” defined as those in groups 3 and 4 in the 4-stage summary (countries that started most semesters in stages 7-12).
  - Comparison group: countries that avoided the conflict trap (started most semesters in stages 1-6).
- Stylized facts (sample period 2000-2023; about 35 countries in the conflict trap versus 130 countries that avoided it):
  - Growth and levels:
    - Countries that avoided the conflict trap: GDP per capita rose, on average, from around US$9,000 to US$13,000 in the last two decades.
    - Countries in the conflict trap grew more slowly over the same period.
  - Investment:
    - Gross fixed capital formation to GDP: about 20% in conflict trap countries versus 30% in countries that avoided the conflict trap.
  - Current account:
    - Countries in the conflict trap have more volatile current accounts (Figure 5c).
  - Government revenue:
    - Government revenues: below 20% of GDP on average in conflict trap countries versus close to 25% of GDP in countries that avoided the conflict trap.
    - Low revenue supports view of weaker state capacity in conflict trap countries.
- Quantified macroeconomic damages (from two-way fixed effect OLS regressions associating risk stages with outcomes):
  - GDP growth impacts:
    - Spending 1 year in any of stages 1-4: would not experience GDP losses in excess of 1% on average.
    - Spending 1 year in stage 11: expected GDP to contract by 2.7% on average.
    - Spending 1 year in stage 12: expected GDP to contract by 9.2% on average.
  - Other outcome relationships shown in Figure 6 include effects on government revenue, debt levels, and share of informality (coefficients and 95% confidence intervals derived from two-way fixed effect regressions).
- Methodological note:
  - OLS specification: y_{i,t} = outcome; α_i country fixed effect; λ_t year fixed effect; d_{s,i,t} dummy for being in risk stage s at time t; ε_{i,t} error term.
  - Identification strategy follows two-way fixed effects used in Cerra and Saxena (2008) and Mueller (2013); reverse causality and omitted variable bias are acknowledged concerns.

*Source: IMF Working Paper — section 3.1 "The stages of conflict" (content as provided).*

### 4.   The Dynamic Gains of Conflict Prevention

### 4.   The Dynamic Gains of Conflict Prevention

### Overview of DEWAM and motivation
- DEWAM = Dynamic Early-Warning Action Model; integrates monthly forecasts from ConflictForecast with a dynamic optimal policy simulation and a model of the conflict trap.
- ConflictForecast: machine learning forecasting module that leverages close to 6 million news articles to forecast armed conflict outbreaks and the intensity of violence twice per month.
- Key trade-off: prevention requires acting on uncertain signals and accepting false positives; prevention allocates resources to crises that may never materialize because not all development-friendly policies are stabilizing.
- DEWAM uses Dynamic Programming to account for future optimal policy reactions to changing conflict dynamics and to evaluate the trade-off between early, lower-cost interventions and later, higher-cost crisis responses.

### Model structure and mechanics
- Decision framework:
  - Policymaker faces a known transition matrix (Markov chain), damages associated with each conflict risk stage, and intervention costs.
  - Interventions can be implemented at any time; costs are incurred immediately, but interventions affect outcomes by shifting probability masses in the Markov chain.
  - The model allows simulation of acting on weak signals early (lower intervention cost) versus reacting post-outbreak (higher intervention cost plus conflict damages).
- Integration elements:
  - Conflict risk stage model (Section 3) links forecasts with damage and cost estimates.
  - Damage vector, intervention cost vector, and policy effect on transition probabilities are inputs to the dynamic programming optimization (robustness checks in Appendix B).

### Damage valuation and components
- Simulation of damages uses:
  - Regression outcomes of GDP loss (Figure 6).
  - Data on fatalities, ODA flows (categories for rebuilding and humanitarian aid), and refugee flows.
- ODA classification: spend only relating to emergency response and peace/security.
- Valuation assumptions:
  - Value of Statistical Life (VSL) assumed at $900,000 (Leon and Miguel 2017).
  - Refugee per capita costs assumed to be $100 (total UNHCR budget in 2021 divided by total number of refugees).

### Intervention cost assumptions
- Purpose: illustrate dynamic-framework benefits under reasonable cost assumptions rather than derive precise costs of specific interventions.
- Two calibrated cost types (calibrated with policymakers; Mueller et al. 2022):
  - Fixed cost of intervention:
    - Cost proportional to population at $3 per person per year ($0.25 per person per month).
  - Variable cost of intervention:
    - Costs rise in line with the expected number of fatalities in each stage.
    - Three per-fatality cost values tested: $10,000 per fatality, $40,000 per fatality, and $200,000 per fatality.
    - Interventions become extremely costly in stages 11 and 12; little distinction across stages 5 to 10 (typically non-violent).
- Example country-level cost context:
  - Stage 4 policies might cost $164mn every quarter for a country with the population size of Pakistan.
  - Stage 12 interventions would cost $9.45bn every quarter (e.g., mediation for a ceasefire, large peacekeeping deployment, and/or DDR initiatives).
  - Note: Pakistan stated costs assume a variable cost assumption of $200,000 per fatality.

### How interventions are modeled to affect outcomes
- Policy effect is abstracted: interventions decrease likelihood of transitioning to higher (more violent) stages and increase likelihood of moving to lower (more peaceful) stages.
- Baseline calibration assumes policy effectiveness of 10%.
- Conservative note: Mueller and Rauh (2024) suggest violence reduction of close to 50% with interventions are possible, so the 10% calibration is conservative.

### Returns to prevention — simulation results
- Sample: countries in the bottom 75% ranked by GDP per capita.
- Under a neutral scenario on prevention effectiveness, rates of return:
  - Range from $26 to $75 per $1 spent on prevention for countries that have not recently suffered from violence.
  - For countries at High risk that have suffered from violence recently, rates could be as high as $103 per $1 spent on prevention.
- These findings are higher than previous estimates (World Bank 2018 found $1 of preventative activities generates on average $16 worth of gains) because DEWAM accounts for persistence of High violence, substantial economic damages, and higher costs of interventions once in High violence.
- Simulations assume policy instruments can be targeted using monthly forecasts and initiated within a quarter.

### Implications for policy evaluation and design
- Effective prevention requires:
  - Integrating forecasts with estimates of conflict damages and dynamics in a unified framework (DEWAM).
  - Acceptance of false positives as a feature of early-action approaches.
  - Accounting for dynamic policy reactions when evaluating returns to prevention.
- DEWAM supports a preventative approach in contexts with no open or recent violence, assuming targeted, timely interventions informed by monthly forecasts.
- Macroeconomic policy role: subsequent sections investigate whether macroeconomic policies and international engagement affect conflict risk (evidence on fiscal balance improvements reducing violence by 7-10 percent is previewed).

### Empirical identification of macroeconomic policy effects (methodology notes)
- Focus: causal connection between government balance/public expenditure shocks and conflict (incidence = months with violence; intensity = log number of fatalities).
- Policy shocks constructed from forecast errors in IMF World Economic Outlook (WEO) projections:
  - WEO forecasts published in April and October each year from 1990 to 2022; semi-annual frequency used.
  - Forecast error = difference between current WEO edition forecast and previous edition’s forecast (normalized by GDP for comparability across countries).
- Shock binary definitions and cutoffs (Table 4):
  - Change in public expenditure (percent of GDP): ≤ −1.1 (Cutoff percentile 10).
  - Government balance (percent of GDP): ≥ 1.5 (Cutoff percentile 90).
  - Unemployment rate: ≤ −.81 (Cutoff percentile 10).
  - Note: Cutoff values are in percent. Public expenditure shocks = reductions in expectations below 1.1 percent contraction of public consumption expenditure. Government balance shocks = increases in expectations above 1.5 percentage points. Unemployment decreases > 0.8 percentage points treated as shocks.
- Identification strategy:
  - Matched difference-in-difference methodology following Mueller and Rauh (2024) and Callaway and Sant’Anna (2021).
  - Control group constructed by matching countries with similar predicted intensity of conflict prior to policy implementation to ensure comparable pre-treatment conflict trajectories.
  - Estimation window: [−4, 4] semester window around treatment; equation (1) estimates ρj coefficients capturing difference-in-difference effects at j periods relative to treatment.
  - Desirable pattern for causal inference: ρ coefficients near zero up to j = 0 (pre-treatment) and diverging thereafter attributable to treatment.

*Source: IMF Working Paper — chapter 4, “The Dynamic Gains of Conflict Prevention.”*

### 5.3 Results

### 5.3 Results

### Impact of positive shocks to the fiscal balance
- On average, positive shocks to the overall fiscal balance:
  - Decrease the incidence of violence (LHS chart in Figure 9).
  - More significantly, reduce the severity of violence (RHS chart in Figure 9).
- Pre-treatment estimates are not significantly different from 0, supporting the matched difference-in-difference approach.
- Post-treatment estimate: an improvement in the fiscal balance by 1.5% of GDP (top 10 percentile relative to forecast) leads to a reduction in conflict deaths by 9%.
- The policy effectiveness calibration used in the paper’s simulations implies a reduction in violence between 7 and 10%, consistent with the empirical result above.

### Relation to simulations and cost comparisons
- Intervention costs to improve the government balance by 1.5% of GDP, via official grants:
  - Around US$100 per capita for a country like Libya.
  - Around US$10 per capita for a country like Chad.
- Chad has received on average grants worth US$22–28 per capita (about US$350 million) for economic and other development goals.
- Assumed intervention costs for direct violence reduction interventions in the paper: US$3 per capita (under fixed-costs assumption in Annex B.2).
- Benefit-cost perspective:
  - The model’s infinite-horizon calculations in Table 3 show preventative intervention is justified even if it cost over US$200 per capita per year.
  - The future discounted benefits are more than 70 times higher than the assumed costs of US$3 per capita.

### Potential causal channels linking fiscal balances and conflict
- Fiscal space allows governments to react to negative shocks and prevent economic contractions—important to prevent a fall into fragility.
- Stronger fiscal capacity/ state capacity has deterrent effects on challenges to the ruling government and can discourage attempts to subvert institutions.
- Complementing security provision with increased public service provision tends to be most effective in reducing violence (evidence cited from Berman, Shapiro and Felter 2011; Berman et al. 2013; Sánchez de la Sierra 2020).

### Labor market (unemployment) shocks and violence
- Large shocks to unemployment might affect intensity and severity of violence (Figure 10), though parallel pre-trends do not hold for conflict intensity.
- Empirical estimates:
  - Likelihood of violence falls by 1.1 percentage points following a drop in the unemployment rate relative to the forecast (bottom 10th percentile, i.e. a fall by about 0.81 percentage points).
  - Severity of violence (conflict deaths) falls by 5.2%.
- Policy relevance:
  - Supports results where employment guarantees or employment programs can reduce conflict levels (Fetzer 2020; Blattman and Annan 2016).

### Public expenditure cuts and conflict
- How fiscal improvements are achieved matters: sharp public expenditure cuts relative to forecast (lowest 10th percentile, cuts of at least 1.1% of GDP) worsen incidence and severity of conflict.
- Empirical estimates for such public expenditure cuts:
  - Increase the share of months in conflict by 2 percentage points.
  - Increase the number of fatalities by 8.5%.
- Interpretation: maintaining essential government functions and pre-existing social contracts is critical; fiscal consolidation that undermines expenditure delivery can escalate conflict.

### Broader implications for state capacity and policy design
- Cautious support for the view that fiscal consolidation and increased state capacity via higher government revenues reduces conflict intensity and incidence (Besley and Persson 2011a).
- State capacity is most useful for violence reduction when it advances economic inclusion and mitigates effects of external shocks.
- Macroeconomic stability and reduction of fragility can be complementary objectives:
  - Revenue mobilization raises fiscal capacity and protects expenditure.
  - Policies that increase accountability can both save resources and strengthen the social compact.
  - Examples cited: randomized municipal audits (Controladoria-Geral da Uniao (CGU) in Brazil) that increase political accountability and reduce misuse of public resources; tax audits that can increase revenues and shift norms (Weigel 2020).
- Caution on resource windfalls: commodity windfalls that do not generate large employment gains can trigger violence (Dube and Vargas 2013); mining activity can trigger social unrest and conflict (Berman et al. 2017; Sonno 2024).

### Alignment with climate and social protection policies
- Policies recommended for climate adaptation in FCS countries (Jaramillo et al. 2023), such as building fiscal buffers and strengthening institutional capacity for emergency responses, align with conflict prevention goals.
- Strengthening social safety nets aligned with conflict prevention when implemented through employment programs.
- Transfer programs have mixed effects (Rohner 2024 review).

### Engagement by the international community and IMF programs
- Official Development Assistance (ODA) to countries in situations of fragility and conflict was around US$130 billion in 2022 (OECD 2024).
- IMF-supported programs:
  - Provide frameworks for fiscal and monetary policies, deliver technical assistance, and catalyze ODA and other financing.
  - Are proxies for international engagement on economic stabilization despite lower rates of successful completion in fragile settings.
  - Have been found to contribute to higher growth rates and state building, with tax revenues increasing in fragile states after program starts (Collyns, Kuruc and Takagi 2021).
- MONA database usage:
  - A binary indicator is constructed if any IMF program or program review was approved in a given month (variable takes value one on the month the program or program review is approved at the IMF Executive Board, and zero otherwise).

### Empirical impact of IMF programs on violence
- Matched difference-in-difference analysis around program/program review approvals (18 months before and after IMF Board approval) shows:
  - On average, program and program review approvals result in a slight reduction in the likelihood of any battle death in a given month by about 1.5 percentage points.
  - Effect amplifies after around a year to a reduction of around 4 percentage points.
- Interpretation: IMF programs’ roles in boosting growth, catalyzing donor support, raising revenues, and restoring confidence are visible in lowering conflict risk, though other concurrent interventions by MDBs, donors, and UN agencies are not accounted for.

### Coverage of high-risk countries in the FCS list and implications for prevention
- Table 5 results (groups calculated using data from 2022 and 2023 compared to the latest FCS list):
  - Very low risk: FCS 0.03, Non FCS 0.97
  - High risk with little or no recent violence: FCS 0.23, Non FCS 0.77
  - High risk with recent violence: FCS 0.23, Non FCS 0.77
  - High violence: FCS 0.86, Non FCS 0.14
- Many countries categorized as High risk by the paper are not on the World Bank’s FCS list:
  - High violence stages have an 86% likelihood to be on the FCS list.
  - High risk only have a 23% likelihood to be on the FCS list.
- Policy implication: 77% of cases categorized as High risk are not in the FCS list, yet they are where the largest intervention gains for prevention may be achieved. Greater awareness and prevention-focused engagement in these countries could yield additional gains.

*Source: wpiea2024256-print-pdf - 5.3 Results*

### 7.   Policy Recommendations

### 7.   Policy Recommendations

### For development partners and International Financial Institutions (IFIs)
- 1. Adopt systems to identify risks, such as early warning.
  - While countries with High violence make the headlines, there are currently 34 countries at High risk.
  - Conflict prevention has high returns, but it has the highest returns in situations where outbreaks have not yet materialized.
  - The focus should be on building systems to identify risks and possibly internal early warning systems to inform policy decisions in high-risk environments.
  - This may require specific adaptions to internal processes to allow early warning systems to better guide policy adjustments and project design.
- 2. Improve attention to prevention in policy design.
  - The IMF is relatively active in the kind of situations that our model identifies as ripe for high returns from a preventative approach.
  - Presently, the IMF’s FCS Strategy focuses primarily on countries that are already classified as fragile and conflict-affected.
  - Given that there are many situations of high risks across all economies (fragile or not), consideration should be given to internalizing the return to prevention policies in all situations of high risks.
- 3. Resist the “blood leads” phenomena.
  - While man-made disasters require humanitarian support, our framework clearly shows that prevention is most effective in countries at risk that may not be on everybody’s radar.
  - Policy design needs to take the devastating effects of armed conflict into account.
  - In countries with high risk, policies need to be systematically evaluated for their repercussions on fragility.
  - For example, government expenditures are often part of stabilizing political bargains and expenditure reductions can negatively affect risks.
- 4. Help build state capacity to strengthen the social contract.
  - Developing fiscal capacity increases resilience and thereby protects other economic and social development goals, such as inclusion (Besley and Mueller 2021).
  - Program design needs to include specific institution-building goals, like modernization of revenue administrations, tax auditing systems, with a short-term objective to increase revenues and a medium-term objective to make tax collection more efficient, equitable, and transparent.
  - These objectives are deeply political but can play a key role in the escape of the conflict trap if they do not increase exclusion.
  - One way to achieve this is building institutions that help establish checks and balances and steer policy towards policy instruments that work for common purposes (LSE-Oxford Commission, 2018 and J-Pal, 2020).

### For governments of countries in a situation of high risk of conflict
- 1. Recognize the tremendous economic costs of conflict.
  - The high economic toll of internal armed conflict should be understood and policy options designed accordingly.
  - Conflict has unintended and cascading dynamics which can be impossible to predict or undo.
- 2. Address exclusion.
  - When designing economic institutions and policies, actions should be taken to reduce economic and social exclusion, as these are major drivers of violence.
  - Access to justice and to other public services directly reduces grievances.
  - Economic exclusion plays an important role in fragility. Increasing employment opportunities directly reduces fragility.
- 3. Support tax compliance.
  - Taxation increases expectations towards the state and these norms of reciprocity rely on citizens seeing that the state is using revenues for common purposes.
  - To strengthen compliance, increases in taxation should go hand-in-hand with the effective and equitable provision of public services.
  - A corrupt administration feeds into fragility by affecting social norms (Collier 2021).
  - Independent watchdogs have been found to be an effective way to fight corruption.
- 4. Build capacity.
  - Building state capacity has tremendous economic benefits because a more capable state can provide public services efficiently, address market failures, and promote the development of markets.
  - But it also has large indirect economic effects through the violence reductions it may be accompanied by if enhanced capacity is not used to accelerate economic or political exclusion.
  - The magnitude of these indirect economic effects can be so large that it gives a completely new motivation for the development of state capacity in fragile contexts—purely on economic grounds.

### Overall framing and implementation considerations
- These recommendations aim to promote peace, stability, and development in fragile situations through a macroeconomic lens.
- Low state and fiscal capacity are serious constraints and building capacity takes a generation.
- While the average long-term economic gains of capacity building are substantial, they require patience and persistence, and the acceptance that gains may not materialize in every case.
- International actors can provide support in the interim.
- Coordinated efforts by IFIs, national governments, and partner countries can help prevent the devastating impacts of internal conflicts.

*wpiea2024256-print-pdf - 7.   Policy Recommendations*

### References

### wpiea2024256-print-pdf - References

### Key findings from Appendix B: Dynamic Early Warning and Action Model (DEWAM)
- Combining static losses with dynamics shows stage 9 is associated with more damaging expected outcomes than stage 3 despite similar contemporaneous GDP effects because stage 9 has elevated risk of rapid escalation to stages 11 or 12 and potential persistence in high-intensity conflict.
- Average GDP losses from a year in stage 3 are -0.7% and -1.5% in stage 9.
- Dynamic losses are essential to communicate the purpose of the DEWAM: acting now to prevent future damages requires understanding situations with high escalation risk even if current violence is low.
- The DEWAM calculates optimal decisions in all possible future worlds, accounting for imperfect forecasts and the possibility that failed prevention may be partially offset by later optimal interventions (dynamic programming is used to conduct simulations).

### Policy simulation scenarios and assumptions
- Policy effectiveness levels tested: 2%, 5%, and 10%.
- Transition probability changes depend on net likelihoods of moving to higher versus lower stages; probability of staying in the same stage can increase if upward movement likelihood exceeds downward movement likelihood.
- Reported BCRs represent the average return across all countries that fall into the bottom 75% according to GDP per capita.
- Benefit-cost ratio (BCR) is computed as gross gains divided by intervention costs; future gains/losses are discounted using a rate of 4%. A BCR less than 1 implies the intervention is not cost-effective.
- Three scenarios to gauge robustness:
  - Pessimistic: policies effective in 2% of applications; intervening in periods of ongoing conflict is extremely costly (variable cost of $200,000 per fatality).
  - Neutral: policies effective in 5% of cases; intervention costs increase by $40,000 per fatality in each stage.
  - Optimistic: policies effective in 10% of cases; late interventions are more financially feasible (variable cost of $10,000 per fatality).
- Lower costs of intervening in conflict-affected stages make prevention more attractive.
- The BCR calculations differentiate between including and excluding effects of economic growth because economic benefits of preventing conflict are large and can make policies always appear cost-effective when included.

### Results: Benefit-Cost Ratios and policy timing
- When including GDP growth effects, economic and social policy benefits are greatest in situations of elevated risk (stages 5-10).
- According to the model, the optimal time for tailoring policy to fragility is in stage 8 or even 9.
- In the neutral scenario, returns to interventions are maximized in stage 8: each $1 spent on prevention saves a monetary equivalent of $201 in prevented economic losses, fatalities, displacement, and humanitarian aid.
- Even without explicitly modeling higher costs of post-conflict reconstruction, BCR figures in stages 5 and 6 are comparable to those in stages 7 to 10, particularly in the pessimistic scenario, supporting preventive macro policies in circumstances with no open or recent violence.
- When excluding the effect of risk stages on GDP growth, returns are markedly lower. In the pessimistic scenario, only stages 6, 8, 9 and 10 yield cost-effective interventions (BCR ≥ 1).
- The case for true prevention (policies in stages 5 and 6 where conflict has yet to break out) is generally weaker when GDP growth effects are excluded; however, the analysis supports post-conflict reconstruction efforts (interventions in stages 7-10) to help countries escape the conflict trap.
- The main text presents results aggregated into four groups: Very low risk; High risk with little or no violence; High risk with recent violence; and High violence. Core paper results are a weighted average of the neutral scenario that includes GDP loss; BCRs are weighted according to the frequency of stages in a given group.

*Italic: Source — Excerpt from the References and Appendix B (Dynamic Early Warning and Action Model) of wpiea2024256-print-pdf.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024256-print-pdf.pdf_