## Adding Fuel to the Fire: How Weather Shocks Intensify Conflict

## Source details

**Canonical URL:** [Adding Fuel to the Fire: How Weather Shocks Intensify Conflict](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024112-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024112-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024112-print-pdf.pdf.json)

---

### Background and motivation
- Climate change is associated with increasing temperatures and erratic precipitation; the relationship between climate change and conflict has gained increasing attention from academics and policy makers.
- The COP28 declaration on climate, relief, recovery, and peace (December 2023) elevated the intersection of climate change, conflict, fragility, and humanitarian crises in UN-level policy discussions.
- The number of conflicts around the world since the early 1990s has been rising, and more sharply over the last decade.
- Possible pathways through which climate shocks can influence conflict include resource scarcity, food insecurity, displacement, and economic shocks.
- This paper analyzes the effect of weather shocks—measured as fluctuations in temperature and precipitation—on conflict to inform understanding of the human toll and policy responses.

### Main findings
- Weather shocks—in particular higher temperatures—significantly worsen conflict.
- Weather shocks may not trigger the onset of new conflict but they exacerbate the intensity of conflict where it already exists.
- Estimation and projection:
  - Under a high emissions scenario (RCP 8.5), and all else equal, by 2060 conflict deaths as a share of the population for a median country facing conflict could increase by 12.3 percent due to rising temperatures.
- Policy implication: integrate climate resilience into peace and security efforts and design climate adaptation policies that support conflict prevention and resolution.

### How this paper extends the literature
- Uses within-country heterogeneity: a high-frequency database with 300,480 observations matched by georeferencing for 2,504 subregions across 168 countries (including advanced, emerging, and low-income countries).
- Captures within-year dynamics: monthly frequency data between 2013 and 2022 to account for shocks of different durations.
- Focuses on conflict intensity at the subregional level (defined as the number of conflict-related deaths as a share of the population), using log form in regressions, rather than a dichotomous onset measure.

### Literature overview and pathways from climate shocks to conflict
- Empirical literature generally finds limited evidence that climate shocks affect conflict onset but suggests they can increase conflict duration, severity, and intensity.
- Cited empirical patterns:
  - Deviations from moderate temperatures and precipitation patterns systematically increase conflict risk.
  - Large floods did not ignite new conflict but fueled existing armed conflicts.
  - Rainfall variability affects instances of conflict in Africa.
  - Warmer-than-normal temperatures raise the risk of violence in East Africa; wetter deviations decrease the risk.
  - Climate-related disasters increase armed conflict risk where conducive socioeconomic and political conditions coexist.
- Identified pathways:
  - Resource scarcity: competition over freshwater, arable land, forests, fisheries; unequal distribution of relief/reconstruction can intensify pre-existing inequalities.
  - Food security: climate-related declines in agricultural productivity can exacerbate food insecurity and social unrest.
  - Displacement: climate-induced migration can increase competition in host communities and exacerbate tensions.
  - Economic shocks: disasters damage infrastructure, reduce productivity, shrink government revenues, and increase poverty and instability.

### Data and empirical approach
- Dataset coverage and key measures:
  - 2,504 subregions across 168 countries, monthly frequency over 2013–2022, totaling 300,480 observations.
  - Conflict: Uppsala Georeferenced Event Dataset (UCDP) fatality data; analysis uses the best estimate of fatalities.
  - Population: Gridded Population of the World (version 4) used to calculate subregional population; dataset covers years 2000, 2005, 2010, 2015, and 2020.
  - Conflict intensity: dependent variable defined as total fatalities per capita at the subregional level (used in log form for regression).
  - Climate variables: temperature and precipitation from the Climatic Research Unit Gridded Time Series (0.5-degree grid), derived by interpolation of monthly climate anomalies from weather station observations.
  - Nighttime lights: Visible Infrared Imaging Radiometer Suite (VIIRS) on Suomi NPP satellite, monthly regional-level data used as a proxy for economic activity and included as a robustness control.
- Visual correlations:
  - Binned scatter plots across the 300,480 observations show conflict intensity tends to be higher when temperatures are high and when precipitation is low.
  - Country-level examples illustrated for Afghanistan, Burkina Faso, Nigeria, Somalia.

### Empirical specification and identification
- Dependent variable: log conflict intensity (number of conflict-related deaths as a share of the population).
- Key regressors: log temperature and log precipitation; nighttime lights (log) included as a control in robustness checks.
- Fixed effects and controls:
  - Subregion fixed effects capture time-invariant subregional characteristics.
  - Time-country fixed effects capture time-varying common shocks at the country level.
- Identification: effects of weather on conflict are identified from within-subregion temporal variation, isolated from country-month common shocks.
- Standard errors are clustered at the subregional level.

### Key empirical findings (4. Empirical Results)
- Where conflict exists, higher temperatures exacerbate conflict intensity.
- Regressions testing whether changes to temperature and precipitation explain the incidence (onset) of conflict were not statistically significant.
- For subregions where conflict intensity is greater than zero:
  - A 1 percent increase in temperature is associated with a 0.2 percent increase in conflict intensity at the subregional level.
- Under RCP 8.5, by 2060 conflict deaths as a share of the population for a country already facing conflict could increase by 12.3 percent due to rising temperatures.
- Precipitation is not found to be significantly associated with conflict intensity once controlling for time-country fixed effects.

### Mechanisms and caveats
- Higher temperatures can influence conflict through resource scarcity, food insecurity, displacement, and economic shocks; identifying specific pathways requires further analysis.
- Results underestimate destructive impacts of conflict that do not result in higher deaths, including destruction of human and physical capital.
- Conflict onset is driven by complex factors (governance, social dynamics, politics, historical conflicts, socioeconomic conditions); weather shocks exacerbate intensity where conflict already exists but may not trigger new conflict.

### Robustness and specification checks
- Results are robust to alternative specifications, including:
  - Including temperature and precipitation variables simultaneously.
  - Adding nighttime lights as a proxy for regional economic activity (though the paper does not use this as baseline because nighttime lights may be affected by climate variation).
  - Alternative subsamples: only low-income countries; only emerging market economies.
- Significance notation used in tables: ***p < 0.01, **p < 0.05, *p < 0.1.

### Selected regression statistics
- Table 1 (selected entries):
  - Temperature coefficients (columns shown): 0.121***, 0.098**, 0.175***, 0.166***
  - Precipitation coefficients (columns shown): -0.026***, -0.015**, 0.006, 0.001
  - Constant terms (columns shown): -12.390***, -11.936***, -12.276***, -12.53***, -12.01***, -12.50***
  - Observations (columns shown): 18,024; 18,010; 17,645; 16,947; 16,942; 16,577
  - R-squared (columns shown): 0.661; 0.661; 0.664; 0.781; 0.781; 0.783
  - Fixed effects: Time FE = YES (in applicable columns); Subregion FE = YES; Time X Country FE = YES (in applicable columns)
- Table 2 (selected entries):
  - Temperature coefficients (columns shown): 0.207***; 0.199***; 0.168***; 0.152***; 0.200**; 0.182*
  - Precipitation coefficients (columns shown): 0.008; 0.003; 0.038; 0.024; -0.023; -0.019
  - Nighttime Lights coefficients (columns shown): -0.067**; -0.070**; -0.066**
  - Constant terms (columns shown): -12.13***; -11.50***; -12.12***; -11.70***; -11.29***; -11.71***; -13.31***; -12.65***; -13.21***
  - Observations (columns shown): 15,435; 15,425; 15,074; 7,935; 8,073; 7,816; 9,012; 8,869; 8,761
  - R-squared (columns shown): 0.791; 0.791; 0.793; 0.662; 0.656; 0.661; 0.806; 0.808; 0.808
  - Subregion FE = YES (in applicable columns); Time X Country FE = YES (in applicable columns)
- Note: Robust t-statistics are reported in parentheses in the source tables.

### Policy implications and directions for further work
- Findings highlight risks climate change poses for peace and security and underscore the importance of:
  - Integrating climate resilience into peace and security efforts.
  - Designing climate adaptation policies that support conflict prevention and resolution.
- Further research priorities:
  - Better understand causal pathways through which climate shocks influence conflict.
  - Identify specific policies and climate-resilience interventions that can help countries mitigate conflict.

*Source: IMF Working Paper — Adding Fuel to the Fire: How Weather Shocks Intensify Conflict (1. Introduction; 4. Empirical Results).*

### 1. Introduction ........................................................................................................

### 1. Introduction

### Background and motivation
- Climate change is associated with increasing temperatures and erratic precipitation; the relationship between climate change and conflict has gained increasing attention from academics and policy makers.
- The COP28 declaration on climate, relief, recovery, and peace (December 2023) elevated the intersection of climate change, conflict, fragility, and humanitarian crises in UN-level policy discussions.
- The number of conflicts around the world since the early 1990s has been rising, and more sharply over the last decade (Figure 1).
- Possible pathways through which climate shocks can influence conflict include resource scarcity, food insecurity, displacement, and economic shocks.
- This paper analyzes the effect of weather shocks—measured as fluctuations in temperature and precipitation—on conflict to inform understanding of the human toll and policy responses.

### Main findings
- Weather shocks—in particular higher temperatures—significantly worsen conflict.
- Weather shocks may not trigger the onset of new conflict (as conflicts derive from a complex range of factors) but they exacerbate the intensity of conflict where it already exists.
- Estimations indicate that in a high emissions scenario (RCP 8.5), and all else equal, by 2060 conflict deaths as a share of the population for a median country facing conflict could increase by 12.3 percent due to rising temperatures.
- Results highlight risks that climate change poses for peace and security and underscore the importance of integrating climate resilience into peace and security efforts and designing climate adaptation policies that support conflict prevention and resolution.

### How this paper extends the literature
- Uses within-country heterogeneity: a unique high-frequency database with 300,480 observations matched by georeferencing for 2,504 subregions across 168 countries (including advanced, emerging, and low-income countries).
- Captures within-year dynamics: monthly frequency data between 2013 and 2022 to account for shocks of different durations.
- Focuses on conflict intensity at the subregional level (defined as the number of conflict-related deaths as a share of the population), instead of a dichotomous onset measure.

### Literature overview and pathways from climate shocks to conflict
- Existing empirical literature generally finds limited evidence that climate shocks affect conflict onset but suggests they can increase conflict duration, severity, and intensity.
- Key cited patterns and findings from the literature:
  - Deviations from moderate temperatures and precipitation patterns systematically increase conflict risk (Burke, Hsiang, and Miguel 2015).
  - Large floods did not ignite new conflict but fueled existing armed conflicts (Ghimire and Ferreira 2016).
  - Rainfall variability affects instances of conflict in Africa (Hendrix and Salehyan 2012; Miguel, Satyanath, and Sergenti 2004).
  - Warmer-than-normal temperatures raise the risk of violence in East Africa; wetter deviations decrease the risk (O’Loughlin and others 2012).
  - Climate-related disasters increase armed conflict risk where conducive socioeconomic and political conditions coexist (IPCC 2022; Koubi 2019; Ide and others 2020; von Uexkull and others 2016; Nel and Righarts 2008).
- Identified pathways:
  - Resource scarcity: competition over freshwater, arable land, forests, fisheries; unequal distribution of relief/reconstruction can intensify pre-existing inequalities.
  - Food security: climate-related declines in agricultural productivity can exacerbate food insecurity and social unrest.
  - Displacement: climate-induced migration can increase competition in host communities and exacerbate tensions.
  - Economic shocks: disasters damage infrastructure, reduce productivity, shrink government revenues, and increase poverty and instability.

### Data and empirical approach
- Dataset coverage and key measures:
  - 2,504 subregions across 168 countries, monthly frequency over 2013–2022, totaling 300,480 observations.
  - Conflict: Uppsala Georeferenced Event Dataset (UCDP) fatality data; analysis uses the best estimate of fatalities.
  - Population: Gridded Population of the World (version 4) used to calculate subregional population; dataset covers years 2000, 2005, 2010, 2015, and 2020.
  - Conflict intensity: dependent variable defined as total fatalities per capita at the subregional level (used in log form for regression).
  - Climate variables: temperature and precipitation from the Climatic Research Unit Gridded Time Series (0.5-degree grid), derived by interpolation of monthly climate anomalies from weather station observations.
  - Nighttime lights: Visible Infrared Imaging Radiometer Suite (VIIRS) on Suomi NPP satellite, monthly regional-level data used as a proxy for economic activity and included as a robustness control.
- Visual correlations:
  - Binned scatter plots (Figure 3) across the 300,480 observations show conflict intensity tends to be higher when temperatures are high and when precipitation is low.
  - Country-level examples illustrated in Figure 4 (Afghanistan, Burkina Faso, Nigeria, Somalia).

### Empirical specification and identification
- Estimation uses a panel regression model where the dependent variable is log conflict intensity (number of conflict-related deaths as a share of the population).
- Key regressors: log temperature and log precipitation; nighttime lights (log) included as a control in robustness checks.
- Fixed effects and controls:
  - Subregion fixed effects capture time-invariant subregional characteristics (e.g., culture, geography).
  - Time-country fixed effects capture time-varying common shocks at the country level (including political, socioeconomic, and governance factors).
- Identification strategy: effects of weather on conflict are identified from within-subregion temporal variation, isolated from country-month common shocks.
- Standard errors are clustered at the subregional level.

*Source: IMF Working Paper — Adding Fuel to the Fire: How Weather Shocks Intensify Conflict (1. Introduction).*

### 4. Empirical Results

### 4. Empirical Results

### Key empirical findings
- Where conflict exists, higher temperatures exacerbate conflict intensity (Table 1).
- Regressions testing if changes to temperature and precipitation explain the incidence of conflict were not statistically significant.
- For the sample of subregions where conflict intensity is greater than zero, panel regressions at monthly frequency show:
  - A 1 percent increase in temperature is associated with a 0.2 percent increase in conflict intensity at the subregional level (Table 1, column 4).
- Annualized projection under a high emissions scenario (RCP 8.5), all else equal:
  - By 2060 conflict deaths as a share of the population for a country already facing conflict could increase by 12.3 percent due to rising temperatures.
  - Estimates are based on the median effect in the subsample of fragile and conflict affected states (FCS). FCS are defined based on the World Bank’s FCS list.
- Precipitation is not found to be significantly associated with conflict intensity once controlling for time-country fixed effects (Table 1, column 5).

### Mechanisms and caveats
- Higher temperatures can influence conflict through resource scarcity, food insecurity, displacement, and economic shocks; a separate analysis would be needed to identify specific pathways.
- Results underestimate destructive impacts of conflict that do not result in higher deaths, including destruction of human and physical capital.
- Conflict onset is driven by a complex range of factors (governance, social dynamics, politics, historical conflicts, socioeconomic conditions); weather shocks exacerbate intensity where conflict already exists but may not trigger new conflict.

### Robustness and specification checks
- Results are robust to alternative specifications (Table 2):
  - Including temperature and precipitation variables simultaneously.
  - Adding regional economic activity proxied by nightlight data (Table 2, columns 1-3).
  - Alternative subsamples:
    - Only low-income countries (Table 2, columns 4-6).
    - Only emerging market economies (Table 2, columns 7-9).
- The regression specification including nighttime lights is not taken as the baseline because nighttime lights may themselves be affected by climate variation.

### Selected regression statistics (from Table 1 and Table 2 in the source)
- Table 1 (selected entries):
  - Temperature coefficients (columns shown): 0.121***, 0.098**, 0.175***, 0.166***
  - Precipitation coefficients (columns shown): -0.026***, -0.015**, 0.006, 0.001
  - Constant terms (columns shown): -12.390***, -11.936***, -12.276***, -12.53***, -12.01***, -12.50***
  - Observations (columns shown): 18,024; 18,010; 17,645; 16,947; 16,942; 16,577
  - R-squared (columns shown): 0.661; 0.661; 0.664; 0.781; 0.781; 0.783
  - Fixed effects: Time FE = YES (in applicable columns); Subregion FE = YES; Time X Country FE = YES (in applicable columns)
- Table 2 (selected entries):
  - Temperature coefficients (columns shown): 0.207***; 0.199***; 0.168***; 0.152***; 0.200**; 0.182*
  - Precipitation coefficients (columns shown): 0.008; 0.003; 0.038; 0.024; -0.023; -0.019
  - Nighttime Lights coefficients (columns shown): -0.067**; -0.070**; -0.066**
  - Constant terms (columns shown): -12.13***; -11.50***; -12.12***; -11.70***; -11.29***; -11.71***; -13.31***; -12.65***; -13.21***
  - Observations (columns shown): 15,435; 15,425; 15,074; 7,935; 8,073; 7,816; 9,012; 8,869; 8,761
  - R-squared (columns shown): 0.791; 0.791; 0.793; 0.662; 0.656; 0.661; 0.806; 0.808; 0.808
  - Subregion FE = YES (in applicable columns); Time X Country FE = YES (in applicable columns)
- Note formatting in tables: Robust t-statistics are reported in parentheses; significance levels: ***p < 0.01, **p < 0.05, *p < 0.1.

### Policy implications and directions for further work
- The findings highlight the risks that climate change poses for peace and security and underscore the importance of:
  - Integrating climate resilience into peace and security efforts.
  - Designing climate adaptation policies that support conflict prevention and resolution.
- Further work is needed to:
  - Better understand the causal pathways through which climate shocks influence conflict.
  - Identify specific policies and climate-resilience interventions that can help countries mitigate conflict.

*Source: 4. Empirical Results, wpiea2024112-print-pdf.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024112-print-pdf.pdf_