## wpiea2021135-print-pdf

## Source details

**Canonical URL:** [wpiea2021135-print-pdf](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021135-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021135-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021135-print-pdf.pdf.json)

---

### Frequency of Unrest Events — Introduction and Research Questions
- The frequency of social unrest has increased recently.
- Only in 2019, social unrest events affected a diverse set of countries, ranging from advanced economies (France and the People's Republic of China Hong Kong Special Administrative Region) to emerging market economies (Chile and Lebanon).
- The latest Global Peace Index (2020) suggests that the number of riots, general strikes and anti-government demonstrations around the world increased by 244 per cent over the last decade (2011 to 2019).
- The COVID-19 pandemic could further accentuate recent trends by disproportionately affecting vulnerable groups and exacerbating inequality and poverty (Furceri, Loungani, Ostry, and Pizzuto, 2020; IMF, 2020), which are important drivers of popular discontent.

Research questions addressed:
- How does social unrest affect economic activity?
- What sectors and components of aggregate demand are more likely to be adversely affected?
- Can the adverse economic implications of social unrest be dampened by strong institutions and available policy space?
- Are emerging markets and advanced economies affected similarly?
- Do different types of unrest have different effects?

### Data: Reported Social Unrest Index (RSUI) and Event Construction
- Data source: Barrett, Appendino, Nguyen and de Leon Miranda (2020).
- RSUI characteristics:
  - Monthly news-based index starting in January 1985.
  - Quantifies extent of social unrest for a large set of countries.
  - Primary source for news articles: Dow Jones’ Factiva news aggregator.
  - Sample restricted to printed articles published in major English-language newspapers and networks in Canada, the UK and the US.
- RSUI formula uses:
  - 푥_{it}: the article count related to unrest in country i in month t.
  - 푧_{t}: the overall article count in period t.
  - 푥̅_{i} and 푧̅: corresponding averages over all time periods.
- Article selection: search criteria include inclusive and exclusive requirements, location criteria, and a minimum word count of 100 words.
- RSUI measures extent of social unrest for each country over time relative to its country mean.

RSUI-implied event algorithm (must satisfy all three criteria):
1. Local peak: 푅푆푈퐼_{it} = max{푅푆푈퐼_{it+1}, 푅푆푈퐼_{it-1}}
2. Index magnitude: one of
   - 푅푆푈퐼_{it} > 푅푆푈퐼̅_{it} + (4 * sd(푅푆푈퐼_{it}))
   - 푅푆푈퐼_{it} is in the top 2%
   - 푅푆푈퐼_{it} exceed the 20-year moving average by 4 times the 20-year st. dev.
3. The unrest article count for the month is at least 10 percent the average monthly count for the country over the past 12 months.

Event labeling and categorization:
- Events satisfying all five criteria used by the authors are labeled as major events.
- Event types (from contemporaneous news keyword classification):
  - a) Political/elections: 490 events
  - b) Socio-economic: 101 events
  - c) Mixed (political and socio-economic): 40 events
  - Events that could not be clearly identified: 205 events
- Quarterly RSUI: aggregated by taking the maximum monthly RSUI over the corresponding quarter.
- Quarterly event dummies: aggregated by taking the maximum over the quarter.
- New events focus: new events must be at least 8 quarters apart.

Sample and macroeconomic data sources:
- Sample: 89 countries with RSUI, quarterly GDP, and commodity terms-of-trade data available over 1990-2019; fragile states excluded.
- Data sources include national accounts, Fiscal Monitor, World Governance Indicators, Ilzetzki-Reinhart-Rogoff exchange rate regimes, Haver, World Uncertainty Index, Gruss and Kebhaj commodity terms-of-trade, CEPII Geodist, ILO EPLEX, WEO, and World Economic Forum Global Competitive Index.
- Sample composition: 32 Advanced Economies and 57 Emerging Markets and Low-Income Countries.

### Methodology — Local Projections, State-Dependence, and IV Strategy
- Method: Local projection method (Jordà, 2005; Teulings and Zubanov, 2014) to estimate impulse responses without constraining shape.
- Benchmark quarterly specification (notation preserved):
  y_{i,t+h} − y_{i,t−1} = α^h_i + γ^h_t + ∑_{j=1}^h μ_{j,h} unrest_{i,t+j} + β^h unrest_{i,t} + θ X_{i,t} + ε_{i,t+h}
  - y: variable of interest (GDP, sectoral value added, demand components, confidence)
  - unrest: RSUI or event dummy
  - α^h_i: country fixed effects
  - γ^h_t: quarter/year fixed effects
  - X: controls including past values of the dependent variable, past terms-of-trade growth, and past RSUI
  - ∑_{j=1}^h μ_{j,h} unrest_{i,t+j}: controls for persistence of unrest events

Addressing endogeneity:
- Distinguish countries with prior negative growth events from those without.
- Distinguish countries undergoing fiscal consolidation episodes from those not.
- IV approach: instrument unrest with regional waves of social unrest using three regional-wave indices:
  - distance-weighted measure across all partners (1/log(dist_i j) weights),
  - distance-weighted measure using only IMF region partners,
  - index using only contiguous countries (sharing a land border).
- Empirical fact: episodes of social unrest in one country are associated with a 1 percent increase in the probability that neighbors experience social unrest in the next six months (Barrett et al. 2020).
- State-dependent local projections:
  - Responses allowed to differ by F(z_{i,t−1}), a smooth function in [0,1], F(0)=1/2.
  - State variables: rule of law, debt level, exchange rate flexibility, labor market flexibility, product market competition.
  - Normalization: state variables standardized with mean zero and standard deviation 1.
  - Exchange rate flexibility indicator: one if Ilzetzki, Reinhart, and Rogoff “fine” classification > 9/2.
  - Functional form: F(z) = exp(−λ_0 z)/(1+exp(−λ_0 z)), with λ_0 = 1.5; results robust to alternative λ_0.

### Results — Impact Based on the Unrest Index (RSUI)
Baseline RSUI shock:
- GDP experiences a steady decline following a shock of one standard deviation in the RSUI.
  - On impact, for the overall sample, GDP declines by over -0.15 percentage points qoq.
  - After 6 quarters, quarterly GDP remains about 0.2 percentage points below its pre-shock level.
- Interpretation of one standard deviation in RSUI:
  - Equivalent to protests following the Peña Nieto election in 2012 or Chile’s presidential election protests in 2013.
  - July 2019 Hong Kong SAR and yellow vest protests of 2018 in France resulted in increases of 4 standard deviations in the RSUI.
  - October/November 2019 in Chile resulted in an increase of 10 standard deviations.

Sectoral heterogeneity:
- Adverse effects on GDP are driven by sharp contractions in services and manufacturing.
- After 6 quarters, effects on services and manufacturing are similar to each other and slightly higher than for GDP.
- Agriculture: impact is not significant at any point over the horizon.

Demand-side components:
- Consumption: negative and significant over most of the six quarters following the unrest shock.
- Investment: appears lower but effect not statistically significant.
- Trade: both exports and imports fall after RSUI spikes; contraction in imports becomes larger after the initial shock relative to exports, leading to an improvement of the trade balance.

Confidence and uncertainty:
- RSUI shock associated with a significant increase in the World Uncertainty Index for a few quarters before dying out within the first year.
- Social unrest associated with a drop in consumer confidence and, to a lesser extent, business confidence; these confidence effects are generally not statistically significant.

Heterogeneity by country group:
- Adverse effects evident across all countries regardless of income level.
- Effect is twice as large on impact in EMDEs relative to AEs and remains somewhat larger throughout the analysis window.

Role of institutions and policy space:
- Strong institutions (rule of law) mitigate economic cost: no decline in activity in aftermath of unrest for countries with strong institutions.
- Ample policy space—measured by public debt level and degree of exchange rate flexibility—helps economies better cope with adverse impacts.
- Countries with weak institutions experience a 0.4 percentage point decline in activity six quarters after episodes of unrest (versus no decline for countries with strong institutions).

### Event-based Results (RSUI-implied Events) and Event Types
- A new RSUI-implied event:
  - Lowers GDP by about 0.6 pp on impact relative to baseline.
  - Effect grows to approximately 1 pp after 1 year.
- Major events (satisfying all five author criteria or large RSUI increases) lead to larger GDP contractions and persistent effects.
- Impact by type:
  - Socio-economic motivated episodes lead to sharper GDP contractions than politics/election-related episodes.
  - Mixed socio-economic and political events associated with the largest GDP contractions.
  - Events that cannot be labeled do not bear a negative effect on economic activity.

### Reverse Causality, Prior Low Growth, and Fiscal Consolidation Controls
Prior low growth:
- Definition: country-specific negative deviations of country growth from the country average growth for at least two out of the last three quarters before social unrest.
- Using this definition, 28 percent of new RSUI-implied events are preceded by low growth events.
- Regression evidence:
  - Social unrest significantly affects growth even if not preceded by adverse growth events, with a GDP contraction of about 1 pp after 6 quarters.
  - Social unrest leads to a larger decline in countries experiencing below-average growth prior to the unrest compared to countries that experience below-average growth but no social unrest event.
- RSUI-based numbers:
  - In low prior growth countries, after a one standard deviation increase in RSUI GDP stands approximately -0.2 percentage points below baseline after six quarters.
  - Countries without prior low growth also suffer a decline, with larger contraction on impact for low-growth countries but similar medium-term effects.

Fiscal consolidation:
- Indicator: cons_i,t = 1 if country i’s primary fiscal deficit over GDP increased by more than 2 percentage points year-on-year.
- Results: social unrest has an adverse effect on economic activity regardless of whether a country is undergoing fiscal consolidation; the adverse impact persists beyond potential role of fiscal consolidations.

### Instrumental Variables (Regional Waves) and Robustness
IV strategy and empirical facts:
- Instruments: regional-wave indices (distance-weighted, IMF-region distance-weighted, contiguous countries).
- Empirical fact: neighbor unrest raises probability of unrest by 1 percent in the next six months (Barrett et al. 2020).

IV estimates:
- A one standard deviation increase in RSUI leads to a 0.2 to 0.3 decline in GDP on impact, depending on instrument; impacts persistent over a 6-quarter window.
- IV estimates are larger than baseline panel regressions.
- For RSUI-implied unrest events, IV estimates imply new episodes lead to a 1 to 2 percent decline in GDP on impact relative to baseline, growing to 2 to 3 percentage points after 6 quarters; these IV-estimated effects are roughly 2 to 3 times as large as baseline estimates.

Robustness checks:
- Results robust to reverse causality and omitted variables concerns.
- Results robust to instrumenting social unrest with regional waves of social unrest.
- Results not driven by fiscal consolidations—GDP declines both in countries with sizeable fiscal improvements and those without.

### Key Quantitative Findings (summary of exact reported numbers)
- On impact, overall-sample GDP declines by over -0.15 percentage points qoq after a one standard deviation RSUI shock.
- After 6 quarters, quarterly GDP remains about 0.2 percentage points below its pre-shock level.
- July 2019 Hong Kong SAR and yellow vest protests in France: RSUI increase of 4 standard deviations.
- October/November 2019 Chile: RSUI increase of 10 standard deviations.
- Countries with weak institutions: 0.4 percentage point decline in activity six quarters after unrest.
- New RSUI-implied event: about 0.6 pp GDP decline on impact; approximately 1 pp after 1 year.
- 28 percent of new RSUI-implied events preceded by low growth.
- Social unrest with no prior low growth: GDP contraction of about 1 pp after 6 quarters.
- RSUI spike in low prior growth countries: GDP about -0.2 percentage points below baseline after six quarters.
- IV (one st.dev. RSUI): 0.2 to 0.3 decline in GDP on impact.
- IV (events): new episodes → 1 to 2 percent GDP decline on impact; 2 to 3 percentage points after 6 quarters.
- Major events (at least 4 standard deviations): on average, GDP remains 1 percentage points below baseline after 6 quarters.

### Annex Findings — Structural Factors and IV with GDP Weights
- Annex A:
  - Low product market competition appears to amplify the negative impact of social unrest on economic activity.
  - Labor market flexibility does not seem to make a significant difference on the negative effect of unrest on growth within a horizon up to 6 quarters.
- Annex B — IV with GDP weights:
  - IV instruments constructed with weights log(GDP_j) / sum_{k∈K−i} log(GDP_k).
  - IV results (GDP-weighted) similar in magnitude: one standard deviation increase in RSUI leads to approximately -0.2 percentage points relative to baseline over 6 quarters; RSUI-implied events show persistent declines of 1 percentage points relative to baseline.

*Source: IMF working paper — Authors’ calculations based on Barrett, Appendino, Nguyen and de Leon Miranda (2020).*

### 1. Frequency of Unrest Events______________________________________________ 3

### 1. Frequency of Unrest Events

### Introduction
- The frequency of social unrest has increased recently.
- Only in 2019, social unrest events affected a diverse set of countries, ranging from advanced economies (France and the People's Republic of China Hong Kong Special Administrative Region) to emerging market economies (Chile and Lebanon).
- The latest Global Peace Index (2020) suggests that the number of riots, general strikes and anti-government demonstrations around the world increased by 244 per cent over the last decade (2011 to 2019).
- The COVID-19 pandemic could further accentuate recent trends: the pandemic disproportionately affects vulnerable groups, and recent studies find it could exacerbate inequality and poverty (Furceri, Loungani, Ostry, and Pizzuto, 2020; IMF, 2020), which are important drivers of popular discontent.

### Research questions addressed
- How does social unrest affect economic activity?
- What sectors and components of aggregate demand are more likely to be adversely affected?
- Can the adverse economic implications of social unrest be dampened by strong institutions and available policy space?
- Are emerging markets and advanced economies affected similarly?
- Do different types of unrest have different effects?

### Key methodological note
- Source: Authors’ calculations based on Barrett, Appendino, Nguyen and de Leon Miranda (2020).

### Summary of high-level findings (from the study)
- Economic activity declines following spikes in the unrest index.
- The adverse effect of unrest on GDP is driven by contractions in manufacturing and services value added (sectoral dimension) and consumption (demand dimension).
- One channel through which social unrest affects activity is by lowering confidence and by increasing uncertainty.
- Strong institutions and ample policy space mitigate the decline in activity following episodes of unrest—countries with strong institutions and ample policy space have more modest declines.
- The economic impact of unrest differs by type of event:
  - Episodes motivated by socio-economic reasons result in sharper GDP contractions compared to those associated mainly with politics/elections.
  - Unrest triggered by a combination of both socio-economic and political issues sees the sharpest contractions.
- Robustness checks and additional results:
  - Results are robust to reverse causality and omitted variables concerns.
  - Episodes of social unrest lead to lower economic activity both in countries that faced low growth prior to social unrest as well as those that experienced above average growth pre-event; however, the impact is larger in countries that experienced adverse growth trajectories prior to social unrest.
  - Results do not appear to be driven by fiscal consolidations—GDP systematically declines both in countries that experience sizeable and swift improvements in the primary deficits and those that do not.
  - Results are robust to instrumenting social unrest with regional waves of social unrest.

---

### 2. Literature and Motivation
- Prior literature focuses more on large, violent conflicts; findings include:
  - Conflicts are a key hurdle to economic growth (Rodrik, 1999).
  - Conflicts produce persistent negative effects on output (Cerra and Saxena, 2008; IMF, 2019; Rother et al. 2016).
  - Conflicts result in large overall macroeconomic costs (Novta and Pugacheva, 2020).
- Political instability literature treats social unrest as detrimental to growth (Alesina et al. 1996; Aisen and Veiga, 2013; Jong-A-Pin, 2009).
- Recent IMF-related work includes:
  - Barrett et al. (2021) on social unrest and stock markets.
  - Hlatshwayo and Redl (forthcoming) on macro-criticality and forecasting with machine learning.
  - Saadi Sedik and Xu (2020) on dynamics among social unrest, pandemic history, economic growth, and inequality.
- Contribution of this paper:
  - Leverages the novel RSUI index (Barrett, Appendino, Nguyen and de Leon Miranda, 2020).
  - Applies local projection method (Jordà, 2005; Teulings and Zubanov, 2014) to:
    - Capture a broad pool of social events, including those less severe than civil conflicts.
    - Analyze higher-frequency economic indicators.
    - Assess impacts on demand-side and supply-side components to identify channels.

---

### 3. Data and Methodology

### Social Unrest Data — Reported Social Unrest Index (RSUI)
- Data source: Barrett, Appendino, Nguyen and de Leon Miranda (2020).
- RSUI characteristics:
  - Monthly news-based index starting in January 1985.
  - Quantifies extent of social unrest for a large set of countries.
  - Primary source for news articles: Dow Jones’ Factiva news aggregator.
  - Sample restricted to printed articles published in major English-language newspapers and networks in Canada, the UK and the US.
- RSUI formula uses:
  - 푥_{it}: the article count related to unrest in country i in month t.
  - 푧_{t}: the overall article count in period t.
  - 푥̅_{i} and 푧̅: corresponding averages over all time periods.
- Article selection: search criteria include inclusive and exclusive requirements, location criteria, and a minimum word count of 100 words.
- RSUI measures extent of social unrest for each country over time relative to its country mean.
- Cross-sectional properties described in Barrett et al. (2020).

### RSUI-implied Events (algorithm criteria)
An RSUI-implied event must satisfy all three criteria:
1. It must be a local peak:
   - 푅푆푈퐼_{it} = max{푅푆푈퐼_{it+1}, 푅푆푈퐼_{it-1}}
2. The index must be large enough to satisfy one of the following conditions:
   - 푅푆푈퐼_{it} > 푅푆푈퐼̅_{it} + (4 * sd(푅푆푈퐼_{it})) or
   - 푅푆푈퐼_{it} is in the top 2% or
   - 푅푆푈퐼_{it} exceed the 20-year moving average by 4 times the 20-year st. dev.

---

### 4. Methods for Impact Estimation
- The paper applies the local projection method proposed by Jordà (2005) and Teulings and Zubanov (2014) to estimate dynamic responses of economic activity to unrest.
- The approach allows analysis of:
  - High-frequency responses to unrest.
  - Sectoral and demand-component breakdowns.
  - Heterogeneity across country groups and event types.
- Instrumental variables strategy and robustness checks used include:
  - Instrumenting social unrest with regional waves of social unrest.
  - Controlling for prior adverse growth events and fiscal consolidations.
  - Annexes provide details on structural factors and IV strategy using GDP weights.

---

*Source: IMF working paper — Authors’ calculations based on Barrett, Appendino, Nguyen and de Leon Miranda (2020).*

### 3. The unrest article count for the month is at least 10 percent the average monthly

### 3. The unrest article count for the month is at least 10 percent the average monthly count for the country over the past 12 months.

### Criteria and event labeling
- An RSUI-implied event is identified when the unrest article count for the month is at least 10 percent the average monthly count for the country over the past 12 months.
- Events satisfying all five criteria used by the authors are labeled as major events.
- Authors construct detailed event timelines from independent narrative sources for more than a dozen country-specific case studies and find these timelines line up closely with RSUI-implied events, supporting that large changes in reported unrest are driven by real-world events rather than media mismatches or bias.

### Unrest data transformations and event categorization
- Quarterly RSUI: for each country, RSUI is aggregated at the quarterly level by taking the maximum monthly RSUI over the corresponding quarter to ensure local spikes are reflected in the quarterly series.
- Quarterly event dummies: the RSUI-implied event dummy and the additional major event dummy are aggregated at the quarterly level by taking the maximum over the quarter.
- New events focus: analysis focuses on new events that are at least 8 quarters apart from each other.
- Event types: events are categorized into three types by consulting contemporaneous news articles and identifying keywords:
  - a) Political/elections
  - b) Socio-economic
  - c) Mixed (intrinsically intertwined political and socio-economic causes)
- Residual category: events for which underlying reasons could not be clearly identified are included in “events that cannot be labeled”.
- Event counts by type:
  - Political/elections: 490 events
  - Socio-economic: 101 events
  - Mixed political and economic reasons: 40 events
  - Events that could not be clearly identified: 205 events

### Sample and macroeconomic data
- Sample: 89 countries with RSUI, quarterly GDP, and commodity terms-of-trade data available over the period 1990-2019; fragile states (as labeled by the World Bank in at least one year since 2006) are excluded.
- Data sources and series:
  - Quarterly GDP and national accounts: national sources (seasonally adjusted by authorities if available; otherwise Haver used for seasonal adjustment).
  - Debt-to-GDP: Fiscal Monitor database.
  - Rule of law: World Governance Indicators (Kaufmann, Kraay, and Mastruzzi, 2010).
  - Exchange rate regimes: Ilzetzki, Reinhart and Rogoff (2019).
  - Confidence indicators: Haver.
  - Uncertainty: World Uncertainty Index (Ahir, Bloom and Furceri, 2018).
  - Commodity terms-of-trade: Gruss and Kebhaj (2019).
  - Bilateral distance and contiguity: CEPII’s Geodist dataset.
  - Labor market legislation: ILO’s EPLEX dataset.
  - Primary balances: IMF’s World Economic Outlook dataset.
  - Product market competition: Global Competitive Index (World Economic Forum).
- Sample listing: 32 Advanced Economies and 57 Emerging Markets and Low-Income Countries with country-specific sample windows (as listed in the source content).

### Econometric approach
- Method: Local projection method (Jordà, 2005; Teulings and Zubanov, 2014) to estimate impulse responses without constraining shape and with less sensitivity to misspecification than VAR models.
- Benchmark quarterly specification (notation preserved in source):  
  y_{i,t+h} − y_{i,t−1} = α^h_i + γ^h_t + ∑_{j=1}^h μ_{j,h} unrest_{i,t+j} + β^h unrest_{i,t} + θ X_{i,t} + ε_{i,t+h}
  - y: variable of interest (GDP, sectoral value added, demand components, confidence)
  - unrest: RSUI or event dummy
  - α^h_i: country fixed effects
  - γ^h_t: quarter/year fixed effects
  - X: controls including past values of the dependent variable, past terms-of-trade growth, and past RSUI
  - ∑_{j=1}^h μ_{j,h} unrest_{i,t+j}: included to control for persistence of unrest events (per Teulings and Zubanov, 2014) to avoid bias in β^h
- Addressing endogeneity concerns:
  - Distinguish countries that faced negative growth events prior to social unrest from those that did not.
  - Distinguish countries that underwent fiscal consolidation episodes from those that did not.
  - Implement an IV approach exploiting regional waves of social unrest.
- State-dependent local projections:
  - Specification allows responses to differ by state via F(z_{i,t−1}), a smooth function taking values in [0,1], with F(0)=1/2.
  - State variables considered:
    a) Rule of law
    b) Debt level
    c) Exchange rate flexibility
    d) Labor market flexibility
    e) Product market competition
  - Normalization: each state variable is standardized with mean zero and standard deviation 1.
  - Exchange rate flexibility: F(z) is an indicator equal to one if the “fine” exchange rate classification from Ilzetzki, Reinhart, and Rogoff (2019) takes a value higher than 9/2, and zero otherwise.
  - Functional form: F(z_{i,t−1}) = exp(−λ_0 z_{i,t−1})/(1+exp(−λ_0 z_{i,t−1})), with λ_0 = 1.5 (as in Auerbach and Gorodnichenko, 2012). Results are robust to alternative choices for λ_0.

### Results — Impact based on the Unrest Index (RSUI)
- Baseline finding: GDP experiences a steady decline following a shock of one standard deviation in the RSUI.
  - On impact, for the overall sample, GDP declines by over -0.15 percentage points qoq.
  - After 6 quarters, quarterly GDP remains about 0.2 percentage points below its pre-shock level.
- Interpretation of one standard deviation in RSUI:
  - A shock of one standard deviation is equivalent to the protests following the Peña Nieto election in 2012 or Chile’s presidential election protests in 2013.
  - The July 2019 protests in Hong Kong SAR and the yellow vest protests of 2018 in France resulted in an increase of 4 standard deviations in the RSUI.
  - The events of October/November 2019 in Chile resulted in an increase of RSUI of 10 standard deviations.
- Sectoral heterogeneity (overview from figures cited):
  - Adverse effects on GDP appear driven by sharp contractions in services and manufacturing.
  - After 6 quarters, effects on services and manufacturing are similar in magnitude to each other and slightly higher than for GDP.

*Source: wpiea2021135-print-pdf — https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021135-print-pdf.pdf*

### 0.4 percentage points relative to baseline, respectively). However, the decline in services

### Impact of Social Unrest on Economic Activity

### Sectoral Impacts
- Services: decline appears more immediate following unrest shocks.
- Manufacturing: decline appears more gradual.
- Agriculture: impact is not significant at any point over the horizon.
- Countries with weak institutions experience a 0.4 percentage point decline in activity six quarters after episodes of unrest (versus no decline for countries with strong institutions).

### Demand-side Components
- Consumption: negative and significant over most of the six quarters following the unrest shock.
- Investment: appears lower but the effect is not statistically significant.
- Trade: both exports and imports fall after spikes in the social unrest index; contraction in imports becomes larger after the initial shock relative to exports, leading to an improvement of the trade balance.

### Confidence and Uncertainty
- A shock to the RSUI is associated with a significant increase in the World Uncertainty Index for the corresponding economy for a few quarters before dying out during the first year.
- Social unrest is associated with a drop in consumer confidence and, to a lesser extent, in business confidence; these confidence effects are generally not statistically significant.

### Heterogeneity by Country Group
- Adverse effects of social unrest are evident across all countries regardless of income level.
- The effect is twice as large on impact in EMDEs relative to AEs and remains somewhat larger throughout the analysis window.

### Role of Institutions and Policy Space
- Strong institutions (measured by rule of law) mitigate the economic cost: no decline in activity in aftermath of unrest for countries with strong institutions.
- Ample policy space—measured by public debt level and degree of exchange rate flexibility—helps economies better cope with adverse impacts.

### Event-based Results (RSUI-implied Events)
- A new RSUI-implied event lowers GDP by about 0.6 pp on impact relative to baseline, and this effect grows to approximately 1pp after 1 year.
- Major events lead to larger GDP contractions and persistent effects on economic activity.

### Impact by Type of Unrest Event
- Socio-economic, political, and mixed events all lead to persistent reductions in economic activity.
- Socio-economic motivated episodes lead to sharper GDP contractions than politics/election-related episodes.
- Events triggered by both socio-economic and political factors are associated with the largest GDP contractions.
- Events that cannot be labeled as socio-economic or political do not bear a negative effect on economic activity.

### Addressing Reverse Causality and Prior Low Growth
- Visual evidence: no clear sign of a decline in growth pre-unrest apart from possibly three quarters prior to unrest (t-3); deviations from average growth are close to or below zero for the four quarters following new episodes of unrest.
- Definition: prior low growth event = country-specific negative deviations of country growth from the country average growth for at least two out of the last three quarters before the social unrest.
- Using this definition, 28 percent of the new RSUI-implied events are preceded by low growth events.
- Regression results:
  - Social unrest significantly affects growth even if not preceded by adverse growth events, with a GDP contraction of about 1pp after 6 quarters.
  - Social unrest leads to a larger decline in economic activity in countries experiencing below-average growth prior to the unrest event compared to countries that experience below-average growth but no social unrest event.
- RSUI-based analysis: in countries with low prior growth, after a one standard deviation increase in the RSUI GDP stands approximately -0.2 percentage points below the baseline after six quarters; countries without prior low growth also suffer a decline, with larger contraction on impact for low-growth countries but similar medium-term effects.

### Controlling for Fiscal Consolidations
- Fiscal consolidation indicator: cons_i,t = 1 if country i’s primary fiscal deficit over GDP increased by more than 2 percentage points year-on-year.
- Results: social unrest has an adverse effect on economic activity regardless of whether a country is undergoing fiscal consolidation; adverse impact of unrest persists beyond potential role of fiscal consolidations.

### Instrumental Variables (Regional Waves) and Robustness
- Instrument strategy: instrument unrest in country i with current and past episodes of social unrest in nearby countries using three regional-wave indices:
  - distance-weighted measure across all partners (using 1/log(dist_i j) weights),
  - distance-weighted measure using only IMF region partners,
  - index using only contiguous countries (sharing a land border).
- Empirical fact used: episodes of social unrest in one country are associated with a 1 percent increase in the probability that neighbors experience social unrest in the next six months (Barrett et al. 2020).
- IV results:
  - A one standard deviation increase in RSUI leads to a 0.2 to 0.3 decline in GDP on impact, depending on the instrument used; impacts are persistent over a 6-quarter window.
  - IV estimates are larger than baseline panel regressions.
  - For RSUI-implied unrest events, new episodes lead to a 1 to 2 percent decline in GDP on impact relative to baseline, growing to 2 to 3 percentage points after 6 quarters; these IV-estimated effects are roughly 2 to 3 times as large as baseline estimates.

### Key Quantitative Findings (summary)
- 0.4 percentage point decline in activity six quarters after unrest in weak-institution countries.
- New RSUI-implied event: about 0.6 pp GDP decline on impact; approximately 1pp after 1 year.
- 28 percent of new RSUI-implied events preceded by low growth (defined above).
- Social unrest with no prior low growth: GDP contraction of about 1pp after 6 quarters.
- RSUI spike: in low prior growth countries GDP about -0.2 percentage points below baseline after six quarters.
- IV: one st.dev. increase in RSUI → 0.2 to 0.3 decline in GDP on impact.
- IV (events): new episodes → 1 to 2 percent GDP decline on impact; 2 to 3 percentage points after 6 quarters.

*Source: wpiea2021135-print-pdf*

### 0.2 percentage points below the pre-shock level after 6 quarters. Second, the adverse effect

### wpiea2021135-print-pdf - 0.2 percentage points below the pre-shock level after 6 quarters. Second, the adverse effect

### Key findings on the macroeconomic impact of social unrest
- A one standard deviation increase in the RSUI leads to a persistent decline in GDP of 0.2 percentage points relative to baseline over a 6-quarter window.  
- The adverse effect on GDP is driven by sharp contractions of services and manufacturing (from a sectoral perspective) and consumption (from the demand side).  
- Social unrest dampens consumer confidence and raises uncertainty.  
- The adverse effect is present across country groups regardless of income level, but the effect upon emerging markets is about two times larger than in advanced economies.  
- Strong institutions and available policy space can dampen the adverse effects of unrest.  
- Major events of social unrest—defined as increases of at least 4 standard deviations in the RSUI—have larger effects: on average, GDP remains 1 percentage points below baseline after 6 quarters of such a social unrest event. The effect of such major events is more than proportional relative to smaller RSUI shocks.  
- The impact of social unrest events is larger than the impact on oil exporters’ GDP of an oil price plunge and is comparable to the impact of an oil price plunge combined with a recession.  
- Unrest episodes motivated by socio-economic considerations result in sharper GDP contractions compared to those associated mainly with politics/elections; episodes triggered by a combination of both socio-economic and political issues see the sharpest contractions.  
- Results are robust to: controlling for previous adverse growth events, controlling for episodes of fiscal consolidation, and implementing an instrumental variables strategy that exploits regional waves of social unrest.

### Annex A — The role of structural factors
- Low product market competition appears to amplify the negative impact of social unrest on economic activity.  
- Labor market flexibility does not seem to make a significant difference on the negative effect of unrest on growth within the horizon of up to 6 quarters.  
- The noted difference between product market competition and labor market flexibility is puzzling and warrants deeper analysis beyond the scope of this paper.

### Annex B — Instrumental variables strategy using GDP weights
- The IV approach instruments social unrest with a GDP-weighted index of global and regional unrest, where for each country i unrest in country j is weighted by log(GDP_j) / sum_{k∈K−i} log(GDP_k), and K−i is either the set of countries other than i in the full sample or in i’s region.  
- Figure B1 (IV approach) shows a one standard deviation increase in the RSUI leads to a persistent decline in GDP of approximately -0.2 percentage points relative to baseline over a 6-quarter window; these results are quantitatively similar to those in Figure 15.  
- Figure B2 (IV approach) shows GDP experiences a persistent decline of 1 percentage points relative to baseline after an RSUI-implied event of social unrest.

*Source: wpiea2021135-print-pdf*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021135-print-pdf.pdf_