## _wp1391

## Source details

**Canonical URL:** [_wp1391](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1391.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1391.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1391.pdf.json)

---

### Introduction and research question
- Focus: effects of political instability on economic performance, especially output and growth.
- Motivating event: the sequence of events dubbed "the Arab spring" and attendant concerns about growth outlook for directly involved countries and the MENA region.
- Central empirical question: How long do the adverse effects of political instability on output last, and what are their dynamics over the medium term?
- Proxy for instability used extensively: number of major cabinet changes (CC).

### Data and descriptive statistics
- Sample: panel of annual data for up to 183 countries spanning from 1980 to 2010.
- Social conflict and political instability sources: Cross-National Time-Series Data Archive (CNTS) and Political Instability Task Force (PITF).
- Ten instability indicators used:
  - (i) cabinet changes (CC);
  - (ii) government crises (GC);
  - (iii)-(v) three regime instability indices (RI1-RI3) from principal component analysis;
  - (vi) general strikes (GS);
  - (vii) antigovernment demonstrations (DE);
  - (viii) composite measure from PCA of CNTS indicators (b)-(i);
  - (ix) composite measure WI (weighted average of CNTS indicators (b)-(i));
  - (x) indicator of adverse regime change (AD).
- Other data sources: IMF WEO, World Bank WDI, Penn World Table (PWT version 7.0), World Governance Indicators (WGI), Fraser Institute’s Economic Freedom of the World (EFW) 2010.
- Sample sizes and selected descriptive figures (preserved exactly):
  - Cabinet changes (CC): total of 4,549 observations; minimum 0; maximum 5.
  - Regime Instability Index 3 (RI3): 4,448 observations.
  - Adverse Regime Change indicator (AD): 5,856 observations.
  - Real GDP per capita (dependent variable): up to 5,081 observations; minimum $117.6; maximum $159,144.50.
    - Note: Real GDP per capita is expressed in constant PPP international dollars of 2005. The maximum corresponds to Qatar in 2009.
- Key summary statistics (Table 1) preserved exactly for selected variables:
  - Real GDP: Obs. 5,081; Mean 9,780.3; Std. Dev. 12,431.7; Min 117.6; Max 159,144.5
  - Price of investment: Obs. 5,081; Mean 4.0; Std. Dev. 0.7; Min -24.1; Max 6.7
  - Openness: Obs. 5,081; Mean 4.2; Std. Dev. 0.6; Min 2.0; Max 6.1
  - Population growth: Obs. 5,445; Mean 1.7; Std. Dev. 1.6; Min -44.4; Max 12.8
  - Cabinet changes (CC): Obs. 4,549; Mean 0.5; Std. Dev. 0.6; Min 0.0; Max 5.0
  - Adverse regime change (AD): Obs. 5,856; Mean 0.0; Std. Dev. 0.1; Min 0.0; Max 1.0
  - Weighted index of political instability (WI): Obs. 5,271; Mean 768.2; Std. Dev. 1,571.5; Min 0.0; Max 2,1250
  - Principal component indicator (PC): Obs. 5,271; Mean 0.0; Std. Dev. 1.5; Min -0.8; Max 18.2

### Co-movement and patterns over time
- CC across regions display high co-movement; pair-wise correlations positive and as high as 0.52 (excluding correlations of individual regions with the whole World).
- Excluding the 1980s: pair-wise correlations as high as 0.70 between cabinet changes in Sub-Saharan Africa and Central and Eastern European countries.
- Pair-wise correlations (full sample): as high as 0.96 between Cabinet Changes and Regime Instability Index 1 (RI1); 0.56 between Cabinet Changes and Government Crises.
- Exception: correlation between the Commonwealth of Independent States and the Advanced Economies is -0.03.
- Time patterns (figure evidence described):
  - Short-term: negative association between social instability and contemporaneous growth (confirmed by Figure 3).
  - Medium-term (7 years after an instability episode): social instability and changes in real GDP are not correlated (Figure 4).
- Implication: important distinction between short-term and medium-term effects.

### Short-term empirical specification and results
- Short-term model features:
  - Dependent variable: log of real GDP for country i at time t.
  - Key regressor: measure of social and political instability (discrete values according to number of episodes; 0 otherwise).
  - Controls: country fixed effects; vector X restricted to investment prices, trade openness, population growth, and (log of) initial GDP; two lags of real GDP growth; time trend.
  - Variance-covariance matrix: clustered sandwich estimator.
  - Robustness checks: exclusion of each region; OLS with country and time fixed effects; two-step system-GMM treating instability as endogenous.
- Conserved short-term effect sizes:
  - Benchmark (CC as instability measure):
    - Coefficient on cabinet changes is negative and statistically significant at 1%.
    - An additional cabinet change is associated with a reduction in real GDP growth of about 1 percentage point.
  - Robustness excluding regions: an additional cabinet change associated with reduction in real GDP growth of about 1 percentage point; point estimates smaller (in absolute value) when MENA or SSA excluded.
  - Across all ten indicators:
    - Strikes and antigovernment demonstrations: lowest effects (0.3-0.6 percentage points) in OLS results; 0-0.2 percentage points in system-GMM results.
    - Adverse regime changes (AD): largest effect — 7 percentage points (both OLS and system-GMM reported).
- OLS contemporaneous coefficients (examples preserved):
  - S_t commonly ≈ -0.011 with t-statistics such as ( -3.87 )*** in baseline specifications.
  - Representative OLS by measure (Table 4):
    - CC: -0.011 ( -3.87 )***
    - GC: -0.021 ( -3.39 )***
    - RI1: -0.006 ( -4.19 )***
    - AD: -0.070 ( -3.37 )***
- GMM-System contemporaneous estimates (Table 5) (examples preserved):
  - CC: -0.017 ( -4.20 )***
  - GC: -0.027 ( -3.37 )***
  - AD: -0.092 ( -3.23 )***
  - Diagnostics: Hansen ^2 reported (e.g., 173.72); AR(1) significant negative z (e.g., -3.81***); AR(2) generally not significant.

### Medium-term methodology (local projections) and main findings
- Method: local projections following Jorda (2005) and Teulings and Zubanov (2010); horizons k = 0,...,8; country fixed effects and country-specific time trends; controls X_it; lags chosen equal to two (robust to other lag choices); White robust standard errors where appropriate.
- Main medium-term findings:
  - The statistical significance of the effect of social instability (proxied by cabinet changes) on output vanishes after only 2 years.
  - The medium-term effect is considerably smaller than the contemporaneous effect and not statistically significant.
  - Across sub-samples excluding each region:
    - In most sub-samples the effect becomes insignificant after only 2 years.
    - Exception: sub-sample excluding Sub-Saharan countries — effect statistically significant up to 3 years after the episode.
  - Sensitivity to lag structure:
    - Medium-term effect ranges from -1.1 percentage points (two lags) to -0.1 percentage points (six lags); point estimates not statistically different from each other.
  - Fixed effects bias check: excluding country fixed effects shows negligible bias; point estimate difference small and not statistically significant.
  - Dynamic panel bias check: two-step system-GMM confirms the statistical significance of the effect of social instability on output vanishes over time.
  - Alternative instability indicators: re-estimating for all indicators confirms medium-term effect is not statistically different from zero.
- Figures described:
  - Figure 5: Dynamic effect of cabinet changes on output (percentage points) — IRF with 95% confidence bands for years 0–7.
  - Figures 6–8: Robustness IRFs across specifications and indicators.

### Role of governance reforms in medium-term dynamics
- Reform measure and specification:
  - Reform indicator (Reforms_it) = change in a composite governance measure in the two (or three) years following the social instability episode.
  - Composite governance indicator constructed via PCA on six WGI indicators: Voice and Accountability; Political Stability; Government Effectiveness; Regulatory Quality; Rule of Law; Control of Corruption.
- Main findings:
  - Medium-term response of output to social instability depends on the country’s ability to implement governance reforms after the episode.
  - Conditional responses (Figure 9):
    - Countries implementing governance reforms in the two (or three) years following instability: output remarkably increases compared to pre-crisis level.
    - Countries where institutional governance deteriorated: output significantly contracts.
    - First and third quartile lines marked when interaction term and overall impact statistically significant at least at a 10 percent confidence level.
  - Heterogeneity:
    - For adverse regime changes, magnifying effect of governance reform is particularly strong for reforms occurring within three years of the adverse regime change.
    - Cabinet changes produce patterns very similar to the baseline.
- Instrumental variables for potential endogeneity of reforms:
  - Instruments used: (i) parliamentary system (presidential vs. parliamentary), (ii) number of veto players who drop from the government in any given year, (iii) initial level of the governance indicator, (iv) lagged changes in governance indicators.
  - Instrument diagnostics:
    - Chi-square test of joint significance = 13.23 (higher than critical value 10 suggested by Staiger and Stock (1997) for strong instruments).
    - Hansen J-statistics and Kleibergen-Paap Wald type F-statistics validate exogeneity and strength.
  - IV results:
    - Instrumenting two- and three-year changes in governance indicators confirms that over the medium-term (five years after the social instability episode) the effect of social conflict on output is positively related with the change in the governance indicator.
- Other reforms:
  - Product market reform and labor market reform included as additional controls.
  - Findings: product and labor market reforms have significant medium-term impact but effects are considerably lower than reforms directly improving governance.
  - Appendix statements:
    - Seven years after social instability, contraction in output is about 2 percent larger in countries where product market flexibility has decreased than in countries where flexibility has increased.
    - Labor market reforms have significant effects only after seven years and need to be larger-scale than governance or product market reforms to restore medium-term output above pre-crisis levels.

### Governance and probability of instability (Probit results)
- Probit regression of probability of instability on initial governance and political controls; marginal coefficients computed at sample mean.
- Cabinet changes:
  - An improvement of one standard deviation of the governance indicator (equal to 2.2) decreases the probability of a cabinet change by about 3 percent (column III).
  - Unconditional probability of the occurrence of a cabinet change ≈ 37 percent.
  - Text preserved exactly: "the results imply a 9 percent increase in the probability of cabinet changes."
  - Control relationships: probability of cabinet changes increases with political fractionalization; decreases with margin of majority and level of autocracy.
  - Correct specification rates across specifications: 67 to 75 percent.
- Adverse regime changes:
  - An improvement of one standard deviation of the governance indicator (equal to 2.2) decreases the probability of a cabinet change by about 0.6 percent (column VI).
  - Unconditional probability of the occurrence of an adverse regime change ≈ 2 percent.
  - Text preserved exactly: "the results imply a 30 percent increase in the probability of an adverse regime change."
  - Control relationships: probability of adverse regime changes decreases with level of autocracy; relation non-linear with partial autocracy showing larger probability.
  - Correct specification rates: 81 to 83 percent.
- Representative probit coefficients (Table 6 examples preserved):
  - Governance_t: -0.028 ( -4.71 )***; -0.022 ( -2.57 )***; -0.005 ( -7.23 )***; -0.003 ( -3.76 )***
  - Fractionalization_t: 0.224 (3.80)***; 0.1883 (4.22)***
  - Margin majority_t: -0.223 ( -3.80 )***; -0.188 ( -4.22 )***
  - Autocracy_t: -0.004 ( -2.95 )***; -0.002 ( -2.05 )**

### Induced political instability and growth
- Two-stage approach:
  - First stage: Probit fitted probability of adverse regime changes from initial governance.
  - Second stage: regress fitted (induced) probability on output growth.
- Findings:
  - Social instability induced by the level of governance has a strong negative effect on growth.
  - Magnitude larger than baseline estimates:
    - Examples (Table 7): Induced regime changes_t = -0.561 ( -2.70 )*** in (II); -0.107 ( -1.91 )* in (III).
    - Non-induced regime changes_t = -0.054 ( -2.00 )** in (III).
    - Adverse regime changes_t = -0.059 ( -2.29 )** in (I).
  - Including initial level of governance in second stage:
    - Initial level of governance has a direct positive effect on growth (e.g., Governance_t = 0.020 (3.59)*** in (IV)).
    - Induced component remains negative and statistically significant.
  - Decomposition: induced effect considerably larger than non-induced component.
- Representative regression statistics:
  - Lagged dependent y_{t-1} strongly negative (e.g., -0.167 ( -7.57 )***).
  - N: 1,917 (columns I, II, IV); 1,689 (column III). R^2 ~0.28–0.30.

### Conclusions and policy implications
- Key empirical conclusions:
  - Negative short-term effect of political instability on contemporaneous output growth confirmed.
  - Detrimental effects of instability on growth are not likely to persist over the medium-term: they vanish after only 2-3 years following an episode of conflict.
  - Medium-term effect is differentiated: countries that improve governance within the first 2-3 years after conflict experience higher medium-term output growth; those that do not reform experience worse medium-term outcomes.
  - Product and labor market reforms produce similar but quantitatively smaller medium-term gains compared with governance improvements.
  - Initial quality of governance significantly and negatively related to the probability of instability episodes, particularly adverse regime changes.
- Policy recommendations (preserved formulation and emphasis):
  - Prioritize timely reforms to improve governance within the first 2-3 years after an episode of political instability or social conflict to achieve better medium-term growth outcomes.
  - Implement structural reforms in product and labor markets as complementary measures, acknowledging they have beneficial but quantitatively smaller medium-term effects compared with governance improvements.
  - Strengthen initial governance institutions to reduce the probability of future instability episodes, including adverse regime changes.

*Italicized source attribution: IMF Working Paper _wp1391 — sections 1, 3.2, 3.3, 3.3.1, 3.3.2, and 4 (excerpts preserved exactly).*

### 1. Introduction

### _wp1391 - 1. Introduction

### Context and research question
- Focus: effects of political instability on economic performance, especially output and growth.
- Motivating event: the sequence of events dubbed "the Arab spring" and attendant concerns about growth outlook for directly involved countries and the MENA region.
- Central empirical question: How long do the adverse effects of political instability on output last, and what are their dynamics over the medium term?

### Key conceptual points
- Political instability and economic growth may be jointly determined: instability can deter investment, cause loss of human capital (e.g., migration), and generate fiscal/monetary volatility leading to high inflation; poor economic performance and high inequality can in turn generate social and political unrest.
- Political instability is recurrent globally; number of major cabinet changes (CC) is used extensively as a proxy for social and political instability.

### Main headline result
- The adverse effects of instability on output start to vanish after only 2-3 years (depending on the instability measure), leaving the medium-term effects as practically null on average.
- Differentiated cross-country effects: countries that reform within 2-3 years following a conflict episode show a higher output growth rate in the medium term, while the opposite occurs for countries that do not implement reforms.

*Source: _wp1391 - 1. Introduction*

---

### Data and descriptive statistics
- Sample: panel of annual data for up to 183 countries spanning from 1980 to 2010.
- Social conflict and political instability sources:
  - Cross-National Time-Series Data Archive (CNTS) — indicators include: (a) cabinet changes; (b) government crises; (c) assassinations; (d) general strikes; (e) guerrilla warfare; (f) purges; (g) riots; (h) revolutions; (i) anti-government demonstrations.
  - Political Instability Task Force (PITF) dataset.
- Ten indicators of social instability used:
  - (i) cabinet changes (CC);
  - (ii) government crises (GC);
  - (iii)-(v) three regime instability indices (RI1-RI3) from principal component analysis;
  - (vi) general strikes (GS);
  - (vii) antigovernment demonstrations (DE);
  - (viii) composite measure from PCA of CNTS indicators (b)-(i);
  - (ix) composite measure WI (weighted average of CNTS indicators (b)-(i));
  - (x) indicator of adverse regime change (AD).
- Other data sources: IMF WEO, World Bank WDI, Penn World Table (PWT version 7.0), World Governance Indicators (WGI), Fraser Institute’s Economic Freedom of the World (EFW) 2010.
- Variable inventory and definitions provided in the Annex.

### Sample sizes and descriptive figures (preserved exactly)
- Cabinet changes (CC): total of 4,549 observations; minimum 0; maximum 5.
- Regime Instability Index 3 (RI3): 4,448 observations.
- Adverse Regime Change indicator (AD): 5,856 observations.
- Real GDP per capita (dependent variable): up to 5,081 observations; minimum $117.6; maximum $159,144.50.
  - Note: Real GDP per capita is expressed in constant PPP international dollars of 2005. The maximum corresponds to Qatar in 2009.

### Co-movement and correlations
- CC across regions display a high degree of co-movement; pair-wise correlations are all positive and as high as 0.52 (excluding correlations of individual regions with the whole World).
- Excluding the 1980s: pair-wise correlations as high as 0.70 between cabinet changes in Sub-Saharan Africa and Central and Eastern European countries.
- Pair-wise correlations between instability indicators for full sample: as high as 0.96 between Cabinet Changes and Regime Instability Index 1 (RI1); 0.56 between Cabinet Changes and Government Crises.
- Exception: correlation between the Commonwealth of Independent States and the Advanced Economies is -0.03.
- High CC–RI1 correlation noted as CC is used in constructing RI1.

### Patterns over time
- Figure evidence (described):
  - Short-term: negative association between social instability and contemporaneous growth (confirmed by Figure 3).
  - Medium-term (7 years after an instability episode): social instability and changes in real GDP are not correlated (Figure 4).
- Implication: distinction between short-term and medium-term effects is important.

*Source: _wp1391 - 1. Introduction*

---

### Empirical analysis framework (overview)
- Goals:
  - Assess effect of social instability on contemporaneous (short-term) growth.
  - Analyze dynamic response of output up to 8 years following instability (medium-term).
  - Assess role of reforms implemented after instability episodes in shaping output effects.
- Dynamic horizon explicitly studied: up to 8 years following occurrence of instability.

### Short-term empirical specification (equation (1))
- Dependent variable: log of real GDP for country i at time t.
- Key regressor: measure of social and political instability (discrete values according to number of episodes; 0 otherwise).
- Controls:
  - Country-specific effects to account for different growth trends and time-invariant influences.
  - Vector X restricted to: investment prices, trade openness (share of total exports and imports in GDP), population growth, and (log of) initial GDP.
  - Two lags of real GDP growth and a time trend included.
- Estimation notes:
  - Variance-covariance matrix estimated using the clustered sandwich estimator to account for within-country serial autocorrelation.
  - Robustness checks: exclusion of each region in turn; OLS with country and time fixed effects; two-step system-GMM treating instability as endogenous.
  - Concern: potential bias from lagged dependent variable and fixed effects (Nickell, 1981) mitigated by the large number of time periods. Endogeneity of instability addressed with system-GMM.

### Short-term empirical findings (preserved effect sizes)
- Benchmark (CC as instability measure):
  - Coefficient on cabinet changes is negative and statistically significant at 1%.
  - An additional cabinet change is associated with a reduction in real GDP growth of about 1 percentage point.
- Robustness:
  - Excluding each region at a time yields similar results: an additional cabinet change associated with reduction in real GDP growth of about 1 percentage point.
  - Point estimates are smaller (in absolute value) when MENA or Sub-Saharan (SSA) countries are excluded, suggesting the effect may be more detrimental for these regions.
- Results across all ten indicators:
  - Negative and significant effect of social instability on contemporaneous growth is robust across indicators.
  - Ranking by magnitude (examples preserved):
    - Strikes and antigovernment demonstrations: lowest effects (0.3-0.6 percentage points) in OLS results; 0-0.2 percentage points in system-GMM results.
    - Adverse regime changes (AD): largest effect — 7 percentage points (both OLS and system-GMM reported).
- Statistically significant control variables commonly: time trend, initial level of GDP, first lag of real GDP growth.

### Inference on methodology
- System-GMM two-step estimation treating social instability as endogenous confirms significance of instability on contemporaneous growth and reinforces the ranking of indicator magnitudes.

*Source: _wp1391 - 1. Introduction*

---

### Medium-term dynamics and reforms (summary)
- Medium-term dynamics:
  - Adverse effects of instability on output start to vanish after 2-3 years on average (measure-dependent).
  - By 7 years after a social instability episode, social instability and changes in real GDP are not correlated for the full sample.
  - Empirical analysis extends dynamic response of output up to 8 years following occurrence.
- Role of reforms:
  - Countries that implement reforms that improve governance within 2-3 years following a conflict episode show higher output growth in the medium term.
  - Countries that do not implement reforms show the opposite pattern (worse medium-term growth outcomes).

*Source: _wp1391 - 1. Introduction*

### 3.2 Medium-term

### 3.2 Medium-term

### Methodology: local projections and impulse response estimation
- IRFs estimated directly from local projections following Jorda (2005) and Teulings and Zubanov (2010).
- Estimated equation (annual data) for each horizon k = 1,...,8 includes:
  - country fixed effects,
  - country-specific time trends (Time_it),
  - controls X_it as described in the previous section,
  - lags of variables with the number of lags (l) chosen equal to two (results robust to other lag choices).
- Corrections for heteroskedasticity applied using White robust standard errors when appropriate.
- IRFs plotted using estimated β_k for k = 0,1,...,8 with 95% confidence bands from associated standard deviations.
- Joint estimation of the k equations also performed; coefficients identical to baseline, t-statistics moderately smaller but remain statistically insignificant.

### Medium-term empirical findings (output response to social instability)
- Main result: the statistical significance of the effect of social instability (proxied by cabinet changes) on output vanishes after only 2 years.
- The medium-term effect is considerably smaller than the contemporaneous effect and not statistically significant.
- Across sub-samples excluding each region in turn:
  - In most sub-samples the effect becomes insignificant after only 2 years.
  - Exception: sub-sample excluding Sub-Saharan countries — effect statistically significant up to 3 years after the episode.
- Sensitivity to lag structure:
  - Medium-term effect ranges from -1.1 percentage points (two lags) to -0.1 percentage points (six lags); point estimates not statistically different from each other.
  - Similar robustness holds for different horizons k and different social instability indicators.
- Fixed effects bias check (Teulings and Zubanov, 2010 concern):
  - Re-estimating equation (2) excluding country fixed effects (Panel E, Figure 6) shows negligible bias; point estimate difference small and not statistically significant.
- Dynamic panel bias check:
  - Two-step system-GMM estimator (controls for fixed effects and lagged dependent variables) yields results where the statistical significance of the effect of social instability on output vanishes over time (Panel F, Figure 6).
- Alternative social instability indicators:
  - Re-estimating equation (2) for all indicators confirms medium-term effect of social instability on output is not statistically different from zero (Figure 8 and other indicators).

### Robustness summary
- Results robust to:
  - including a common time trend, time fixed effects, country-specific time trends (Panels B–D, Figure 6),
  - excluding regions one at a time (Panels B–H, Figure 7),
  - different lag parameterizations,
  - joint estimation of k equations,
  - exclusion of country fixed effects,
  - two-step system-GMM estimation,
  - alternative social instability indicators.

*Italicized source attribution: IMF Working Paper section 3.2, _wp1391 - 3.2 Medium-term*

### 3.3 The Effect of Reforms

### Empirical specification and governance measure
- Estimated model (equation (3)) introduces interaction between social instability and reforms in governance:
  - Reform indicator (Reforms_it) = change in a composite governance measure in the two (or three) years following the social instability episode.
- Composite governance indicator constructed via principal component analysis on six governance indicators (Kaufmann et al., 2010):
  - i) Voice and Accountability,
  - ii) Political Stability,
  - iii) Government Effectiveness,
  - iv) Regulatory Quality,
  - v) Rule of Law,
  - vi) Control of Corruption.
- Reforms measured over two- or three-year changes (to capture lagged and gradual implementation effects).

### Main findings: governance reforms shape medium-term output response
- The medium-term response of output to social instability depends on the country’s ability to implement governance reforms after the episode.
- Conditional responses (Figure 9):
  - Countries implementing governance reforms in the two (or three) years following instability: output remarkably increases compared to pre-crisis level.
  - Countries where institutional governance deteriorated: output significantly contracts.
  - First and third quartile lines marked when interaction term and overall impact statistically significant at least at a 10 percent confidence level.
- Robustness: results hold across all specifications and checks from the previous section.

### Heterogeneity by type of instability and timing
- For adverse regime changes, the magnifying effect of governance reform is particularly strong for reforms occurring within three years of the adverse regime change (Figure 10, panel B).
- Re-estimations for all social instability indicators largely confirm the governance-reform gradient; cabinet changes produce very similar patterns to the baseline.

### Endogeneity and instrumental variable strategy
- Concern: reforms may be endogenous (unobserved factors affect GDP dynamics and probability of reform).
- Instruments for change in governance indicator:
  - (i) parliamentary system (presidential vs. parliamentary),
  - (ii) number of veto players who drop from the government in any given year,
  - (iii) initial level of the governance indicator,
  - (iv) lagged changes in the governance indicators.
- Theoretical justification: political variables (political system and number of veto players) considered strictly exogenous.
- Joint significance and instrument strength:
  - Chi-square test of joint significance = 13.23 (higher than critical value 10 suggested by Staiger and Stock (1997) for strong instruments).
  - Hansen J-statistics and Kleibergen-Paap Wald type F-statistics validate exogeneity and strength.
- IV results:
  - Instrumenting two- and three-year changes in the governance indicators confirms that over the medium-term (five years after the social instability episode) the effect of social conflict on output is positively related with the change in the governance indicator (Figure 11).

### Other reforms and relative magnitudes
- Additional reform controls considered:
  - (i) product market reform,
  - (ii) labor market reform.
- Findings:
  - Medium-term response of output to social instability remains a function of change in governance.
  - Product and labor market reforms also have a significant medium-term impact, but their effects are considerably lower than reforms directly improving governance (Appendix).

*Italicized source attribution: IMF Working Paper section 3.3, _wp1391 - 3.2 Medium-term*

### 3.3.1 Governance and probability of political instability

### Probit model and controls
- Probit regression of the probability of social and political instability episodes on:
  - initial level of governance,
  - controls: political polarization, government fractionalization, share of seats of opposition parties, degree of centralization, margin of majority (fraction of seats held by the government), autocracy indicator, dummy for partial autocracy (center of autocracy distribution).
- Outcomes analyzed: probability of cabinet changes and probability of adverse regime changes.
- Model reports marginal coefficient estimates computed at the sample mean.

### Key quantitative results
- Cabinet changes:
  - An improvement of one standard deviation of the governance indicator (equal to 2.2) decreases the probability of a cabinet change by about 3 percent (column III).
  - Unconditional probability of the occurrence of a cabinet change ≈ 37 percent.
  - Text interpretation: "the results imply a 9 percent increase in the probability of cabinet changes." (text preserved exactly).
  - Control variable relationships:
    - Probability of cabinet changes increases with political fractionalization.
    - Probability of cabinet changes decreases with margin of majority and level of autocracy.
  - Correct specification rates across specifications: 67 to 75 percent.

- Adverse regime changes:
  - An improvement of one standard deviation of the governance indicator (equal to 2.2) decreases the probability of a cabinet change by about 0.6 percent (column VI).
  - Unconditional probability of the occurrence of an adverse regime change ≈ 2 percent.
  - Text interpretation: "the results imply a 30 percent increase in the probability of an adverse regime change." (text preserved exactly).
  - Control variable relationships:
    - Probability of adverse regime changes decreases with level of autocracy, but relation is non-linear: regimes at the center of the autocracy distribution (partial autocracy) are characterized by a larger probability.
  - Correct specification rates across specifications: 81 to 83 percent.

### Interpretation
- Quality of governance plays a key role in reducing the probability of instability episodes, particularly adverse regime changes, and in mitigating their medium-term effects on output.

*Italicized source attribution: IMF Working Paper section 3.3.1, _wp1391 - 3.2 Medium-term*

### 3.3.2 Governance, induced political instability and growth

### Induced-probability construction and second-stage regression
- Construct fitted value of the probability of adverse regime changes from Probit regression using initial governance (this fitted value = probability induced by level of governance).
- Regress this fitted (induced) probability on output growth.

### Findings on induced instability effects
- Social instability induced by the level of governance has a strong negative effect on growth.
- Magnitude of the induced effect is significantly larger than baseline estimates (Table 7, columns II and III).
- Robustness:
  - Result holds across different specifications of equation (1).
  - Robust to inclusion of autocracy and partial autocracy in first-stage regression.
  - Including initial level of governance as a control in the second-stage regression:
    - Initial level of governance has a direct effect on growth not channeled through the probability of adverse regime changes.
    - The induced component of the probability of adverse regime changes (driven by level of governance) remains negative and statistically significant (column III).
- Decomposition:
  - The induced effect of adverse regime changes is considerably larger than the non-induced component (column IV).

*Italicized source attribution: IMF Working Paper section 3.3.2, _wp1391 - 3.2 Medium-term*

### 4. Conclusions and Policy Implications

### 4. Conclusions and Policy Implications

### Key empirical findings
- Confirms the negative short-term effect of political instability on contemporaneous output growth documented by previous studies.
- Shows that the detrimental effects of instability on growth are not likely to persist over the medium-term: they vanish after only 2-3 years following an episode of conflict.
- Provides evidence that the medium-term effect of political instability and social conflict on output growth is differentiated across countries depending on their ability to implement governance-improving reforms within the first 2-3 years after an episode of conflict.
- Finds that countries that improve their levels of governance after periods of conflict experience, over the medium-term, output growth that is significantly higher than in countries that do not improve governance.
- Finds similar, though weaker, conclusions for structural reforms aimed at improving product and labor markets.
- Presents evidence that the probability of instability episodes, particularly adverse regime changes, is significantly and negatively related with the initial level of governance.

### Policy implications and recommendations
- Prioritize timely reforms to improve governance within the first 2-3 years after an episode of political instability or social conflict to achieve better medium-term growth outcomes.
- Implement structural reforms in product and labor markets as complementary measures, acknowledging they have beneficial but quantitatively smaller medium-term effects compared with governance improvements.
- Strengthen initial governance institutions to reduce the probability of future instability episodes, including adverse regime changes.

*Source: 4. Conclusions and Policy Implications*

### References

### _wp1391 - References

### Bibliographic references
- Lists cited works relevant to political instability, economic growth, governance, and econometric methods including authors and exact article/book/journal details as provided (Aisen & Veiga; Alesina et al.; Barro; Bates et al.; Beck et al.; Cukierman et al.; Durlauf et al.; Furceri & Zdzienicka; Heston et al.; Jong-A-Pin; Jorda; Kaufmann et al.; Keefer & Stasavage; Levine & Renelt; Nickell; Rodrik; Roodman; Sala-i-Martin; Staiger & Stock; Teulings & Zubanov).

### Annex — Data description (dependent and control variables)
- Output
  - GDP per capita (rgdpl in the WEO dataset, y in this paper): PPP converted GDP per capita (Laspeyres methodology), derived from growth rates of private consumption, government expenditures, and investment at 2005 constant prices.
- Social conflict indicators (CNTS, PITF, Jong-A-Pin)
  - Major cabinet changes (polit11; CC): number of times in a year a new premier is named and/or 50% of cabinet posts assumed by new ministers.
  - Changes in Effective Executive (polit12): number of times control of executive power changes hands (new executive independent of predecessor).
  - Coups d'État (polit03): number of extra-constitutional or forced changes in the top government elite (successful coups counted; unsuccessful coups not counted).
  - Major Constitutional Changes (polit04): number of basic alterations in constitutional structure (significant alterations only).
  - Legislative Election (polit15): number of elections for lower house in a given year (limited by-elections included).
  - Party fractionalization index (polit01): F = 1 - sum_{i=1}^{m} t(i)^2 where t(i) is proportion of members of ith party (independents disregarded).
  - Assassinations (domestic1), General Strikes (domestic2), Guerrilla Warfare (domestic3), Major government crises (domestic4; GC), Purges (domestic5), Riots (domestic6), Revolutions (domestic7), Anti-government Demonstrations (domestic8).
  - Weighted index of political instability (domestic9; WI): weights — Assassinations (25), Strikes (20), Guerrilla Warfare (100), Government Crises (20), Purges (20), Riots (25), Revolutions (150), Anti-Government Demonstrations (10).
  - Adverse regime change (AD): dummy = 1 for adverse regime change episode (PITF).
- Composite indicators (PCA first component per Cattell’s scree plot)
  - Principal component indicator of political instability (PC): includes domestic1 through domestic8.
  - Composite measure JAP (Jong-A-Pin 2009) when ICRG not used.
  - Regime instability indices: RI1 (polit11, polit12); RI2 (polit11, polit04, polit03, polit12, domestic4); RI3 (polit11, polit12, polit04, polit03, polit12, domestic4, polit15, polit01).
  - Violence index (VI): includes domestic1, polit03, domestic7.
- Macroeconomic & demographic controls
  - Price of investment (PI, from PWT): price level of investment (in logs).
  - Openness (Open, from PWT): (log) Openness at 2005 constant prices, in percent.
  - Population growth (Popg, SPPOPGROW from WDI): annual percent rate.
- Governance and regulation indicators
  - Governance (WGI, Kaufmann et al. 2010): composite of Voice and Accountability; Political Stability and Absence of Violence/Terrorism; Government Effectiveness; Regulatory Quality; Rule of Law; Control of Corruption — aggregated via PCA (first component).
- Political indicators (DPI, Polity IV)
  - System (DPI): parliamentary (2), assembly-elected president (1), president (0).
  - Polarization (Keefer & Stasavage 2003): max difference between chief executive’s party’s value and three largest opposition parties.
  - Government fractionalization, Margin majority, Opposition’s share (DPI).
  - Autocracy (Polity IV Project).
- Regulation quality indices (Fraser Institute EFW 2010, range 0–10)
  - Business/product markets (area5c): composite of Price controls; Administrative requirements; Bureaucracy costs; Cost of starting a business; Extra payments / bribes; Licensing restrictions; Cost of tax compliance.
  - Labor markets (area5b): composite of Hiring regulations and minimum wage; Hiring and firing regulations; Centralized collective bargaining; Hours regulations; Mandated cost of worker dismissal; Conscription.

### Key summary statistics (Table 1)
- Output and macroeconomic/demographic controls
  - Real GDP: Obs. 5,081; Mean 9,780.3; Std. Dev. 12,431.7; Min 117.6; Max 159,144.5
  - Price of investment: Obs. 5,081; Mean 4.0; Std. Dev. 0.7; Min -24.1; Max 6.7
  - Openness: Obs. 5,081; Mean 4.2; Std. Dev. 0.6; Min 2.0; Max 6.1
  - Population growth: Obs. 5,445; Mean 1.7; Std. Dev. 1.6; Min -44.4; Max 12.8
- Social conflict indicators
  - Cabinet changes (CC): Obs. 4,549; Mean 0.5; Std. Dev. 0.6; Min 0.0; Max 5.0
  - Government crises (GC): Obs. 5,276; Mean 0.1; Std. Dev. 0.4; Min 0.0; Max 5.0
  - Weighted index of political instability (WI): Obs. 5,271; Mean 768.2; Std. Dev. 1,571.5; Min 0.0; Max 2,1250
  - Principal component indicator (PC): Obs. 5,271; Mean 0.0; Std. Dev. 1.5; Min -0.8; Max 18.2
  - Regime instability index 1 (RI1): Obs. 4,549; Mean 0.0; Std. Dev. 1.2; Min -0.8; Max 10.6
  - Regime instability index 2 (RI2): Obs. 4,522; Mean 0.0; Std. Dev. 1.4; Min -0.9; Max 14.6
  - Regime instability index 3 (RI3): Obs. 4,448; Mean 0.0; Std. Dev. 1.5; Min -1.1; Max 14.1
  - General strikes (GS): Obs. 5,276; Mean 0.1; Std. Dev. 0.5; Min 0; Max 7
  - Antigovernment demonstrations (DE): Obs. 5,275; Mean 0.5; Std. Dev. 1.5; Min 0; Max 26
  - Adverse regime change (AD): Obs. 5,856; Mean 0.0; Std. Dev. 0.1; Min 0.0; Max 1.0

### Empirical results — core estimates (selected coefficients and statistics)
- OLS contemporaneous effect of social conflicts on output (Table 2, baseline specifications)
  - S_t coefficient (effect of social conflict): commonly reported as -0.011 with t-statistics (e.g., -3.87)*** across many specifications. Example entries:
    - (I) S_t = -0.011 ( -3.87 )***
    - (VI) S_t = -0.013 ( -4.48 )***
  - Lagged dependent y_{t-1} often negative and significant, e.g., -0.077 ( -4.35 )*** in (I).
  - y_{t-1} positive and occasionally significant (e.g., 0.094 (1.89)*).
  - Time_t trend positive and significant, e.g., 0.002 (4.52)***.
  - N ranges: 3,904 up to 4,079 across specifications; R^2 range ~0.06 to 0.18.
- Subsample OLS results (Table 3)
  - S_t remains negative and significant across subsamples. Examples:
    - Full sample (I): -0.011 ( -3.87 )***
    - Excluding AE (II): -0.013 ( -3.83 )***
    - Excluding SSA (VIII): -0.009 ( -2.69 )***
  - N varies by exclusion: e.g., Full sample N = 3,904; Excluding AE N = 3,200; Excluding SSA N = 2,860.
- OLS by different instability measures (Table 4)
  - S_t coefficients by measure:
    - CC: -0.011 ( -3.87 )***
    - GC: -0.021 ( -3.39 )***
    - RI1: -0.006 ( -4.19 )***
    - RI2: -0.008 ( -4.46 )***
    - RI3: -0.006 ( -4.19 )***
    - GS: -0.006 ( -2.45 )**
    - DE: -0.003 ( -3.37 )***
    - PC: -0.005 ( -3.92 )***
    - WI: -0.005 ( -3.53 )***
    - AD: -0.070 ( -3.37 )***
- GMM-System estimates (Table 5)
  - S_t coefficients generally larger in magnitude:
    - CC: -0.017 ( -4.20 )***
    - GC: -0.027 ( -3.37 )***
    - RI1: -0.009 ( -4.23 )***
    - RI2: -0.011 ( -4.63 )***
    - RI3: -0.009 ( -4.38 )***
    - DE: -0.002 ( -1.60 )*
    - PC: -0.006 ( -2.93 )***
    - WI: -0.005 ( -2.83 )***
    - AD: -0.092 ( -3.23 )***
  - Diagnostic statistics:
    - Hansen ^2 reported (examples): 173.72, 176.12, 178.37 across columns.
    - AR(1) significant negative z (e.g., -3.81***), AR(2) generally not significant.
  - Estimator notes: GMM-System Estimator, two-step using Windmeijer standard errors; Population growth and Time trend as predetermined; other controls instrumented using up to 2 lags.
- Governance and probability of instability episodes (Table 6)
  - Governance_t coefficients (examples):
    - Column (I): -0.028 ( -4.71 )***
    - Column (II): -0.022 ( -2.57 )***
    - Column (IV): -0.005 ( -7.23 )***
    - Column (X): -0.003 ( -3.76 )***
  - Other notable coefficients:
    - Fractionalization_t: 0.224 (3.80)*** in one specification; 0.1883 (4.22)*** in another.
    - Margin majority_t: -0.223 ( -3.80 )***; -0.188 ( -4.22 )***.
    - Autocracy_t: -0.004 ( -2.95 )***; -0.002 ( -2.05 )** in different columns.
    - Middle-type regime_t: 0.063 (1.45); 0.008 (2.59)***; 0.011 (3.22)*** in other columns.
  - Sample sizes and classification performance:
    - N-observations: examples 1,381; 1,032; 1,194; 1,952; 1,417; 1,717.
    - %-instability episodes: reported 37, 37, 37, 2, 2, 2 (across columns).
    - % instability episodes (% non-instability episodes) correctly classified: e.g., 75 (34), 67 (47), 74 (40), 83 (73), 81 (84), 83 (83).
    - Pseudo R^2 examples: 0.01, 0.03, 0.02, 0.21, 0.39, 0.37.
    - ^2 statistics (examples): 22.17***, 43.95***, 35.47***, 52.32***, 104.28***, 140.55.
- Induced political instability and output (Table 7)
  - Adverse regime changes_t: -0.059 ( -2.29 )**
  - Induced regime changes_t: -0.561 ( -2.70 )*** in (II); -0.107 ( -1.91 )* in (III).
  - Non-induced regime changes_t: -0.054 ( -2.00 )** in (III).
  - Governance_t positive: 0.020 (3.59)*** in (IV).
  - Lagged dependent y_{t-1} strongly negative (e.g., -0.167 ( -7.57 )***).
  - N: 1,917 (columns I, II, IV); 1,689 (column III). R^2 ~0.28–0.30.

### Figures and dynamic effects (selected descriptions)
- Figure 3 & Figure 4: Simple regressions reported
  - growth(t) = 2.794 -1.491*CC  with reported t-statistics (44.58) and (-12.15).
  - lny(t+7)-lny(t) = 0.235 + 0.050*CC with t-statistics (6.38) and (0.72).
- Figure 5–8: Impulse response functions (IRFs)
  - Figure 5: Dynamic effect of cabinet changes on output (percentage points) — estimated IRF with 95% confidence bands shown for years 0–7.
  - Figures 6–7: Robustness checks of IRFs across specifications (time fixed effects, no time trends, country-specific time trends, exclusion of regions, GMM, etc.).
  - Figure 8: Dynamic effect of other social instability indicators (e.g., RI1 and AD) on output — IRFs with confidence bands.
- Figures 9–11 and Appendix Figures A1–A2: Interaction effects with governance and reform indicators
  - Conditioning on reforms in governance, product market flexibility, and labor market flexibility:
    - Panel A/B comparisons for reform within 2 years and within 3 years display differing IRFs for "deterioration in governance" vs "improvement in governance" (improvement = third quartile change; deterioration = first quartile change).
    - Appendix summary: Seven years after social instability, contraction in output is about 2 percent larger in countries where product market flexibility has decreased than in countries where flexibility has increased (text statement in Appendix).
    - Labor market reforms have significant effects only after seven years and need to be larger-scale than governance or product market reforms to restore medium-term output above pre-crisis levels (text statement in Appendix).

### Empirical synthesis (explicit numeric highlights from tables/figures)
- Baseline contemporaneous OLS estimate of social conflict (S_t) on output: approximately -0.011 (t-statistics ~ -3.87)***.
- GMM-System contemporaneous estimates of S_t generally larger (e.g., CC: -0.017 ( -4.20 )***; AD: -0.092 ( -3.23 )***).
- Induced regime changes have large negative point estimates on output: induced regime changes_t = -0.561 ( -2.70 )*** in one specification.
- Governance improvements associated with lower probability of cabinet changes and adverse regime changes: Governance_t coefficients negative and significant (examples include -0.028 ( -4.71 )***, -0.005 ( -7.23 )***).
- Time trend coefficients in growth regressions commonly positive and significant (e.g., Time_t = 0.002 (4.52)***).

*Source: _wp1391 - References (PDF). *

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1391.pdf_