## Explaining Income Divergence in the Western Indian Ocean

## Source details

**Canonical URL:** [Explaining Income Divergence in the Western Indian Ocean](https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026165-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2026/english/wpiea2026165-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2026/english/wpiea2026165-source-pdf.pdf.json)

---

### Overview and scope
- Focus: explaining income divergence in four Western Indian Ocean island nations: Mauritius, Seychelles, Madagascar, and Comoros.
- Approach: two-pronged—qualitative assessment of growth divergence and quantitative evaluation using Fully Modified Ordinary Least Squares (FMOLS) and Dynamic Ordinary Least Squares (DOLS).
- Time coverage:
  - Qualitative assessment: post-independence period.
  - Quantitative analysis: begins in 1980 due to data limitations (particularly for Comoros).

### Key empirical facts and per-capita trajectories (1980–2022)
- Real GDP per capita changes since 1980:
  - Mauritius: increase of 354 percent.
  - Seychelles: increase of 138 percent.
  - Madagascar: contraction of 35 percent.
  - Comoros: increase of 17 percent.
- Growth classification:
  - Mauritius and Seychelles: among top performers in Sub-Saharan Africa (SSA) with average annual real GDP growth rates of approximately 3 percent.
  - Comoros and Madagascar: more limited growth; divergence widened despite modest initial income gaps at independence.

### Qualitative analysis highlights (Section II)
- Geography and disaster risk:
  - Comoros: area approximately 2,000 square kilometers; population nearly one million; tropical marine climate; tropical cyclones 64 percent and floods 35 percent of average natural-disaster losses (World Bank).
  - Madagascar: nearly 600,000 square kilometers; population approximately 30 million; highly vulnerable to cyclones, floods, droughts.
  - Mauritius: 2,040 km²; coastline 177 km; population over one million; cyclones peak December–March.
  - Seychelles: about 500 km²; population about 120,000; Mahé accounts for 86 percent of population; islands vulnerable to flooding and landslides.
- Political background and governance contrasts:
  - Comoros: independence 1975; recurrent instability, coups, secession efforts; institutional development constrained.
  - Madagascar: independence June 26, 1960; several political crises notably 2009–2012; centralized governance and weak accountability.
  - Mauritius: independence March 12, 1968; stable democracy and relatively strong institutions.
  - Seychelles: independence June 29, 1976; prolonged single-party rule to 2020, state-led development, later multiparty institutions.
- Economic structure and sectoral shifts:
  - Mauritius: transformed from agriculture (sugar) to manufacturing (textiles) and services (tourism, finance, ICT).
  - Seychelles: shifted toward tourism and fish processing; by 2022 tertiary sector contributed 10.8 percentage points to growth.
  - Madagascar and Comoros: remained more agriculture-dependent with limited structural transformation.
- Social policies:
  - Mauritius and Seychelles prioritized education and health (free or broadly accessible education; universal health coverage in Seychelles; strong literacy and life-expectancy outcomes).
  - Comoros and Madagascar: more limited public spending in social sectors.

### Empirical strategy and data
- Outcome variable:
  - Real GDP per capita in constant 2015 USD, logged (log GDPPC_it). Source: World Bank World Development Indicators.
- Main regressors and measures:
  - Education Equality Indicator (V-Dem), levels.
  - Political Corruption Index (V-Dem), levels.
  - Public debt: general government public debt as percent of GDP, logged; squared log included in nonlinear specifications.
  - Trade openness: exports plus imports of goods and services as percent of GDP, logged; treated as a deterministic regressor.
  - Additional controls: initial GDP per capita, investment/capital stock, credit-to-investment ratio, real exchange rate, reserves, FDI, health equality, governance indicators.
- Sample and limitations:
  - Empirical analysis restricted to 1980–2022; four cross-sections (Mauritius, Seychelles, Madagascar, Comoros).
  - Data constraints, particularly for Comoros, required proxies (V-Dem) and limited generalizability.

### Estimators and baseline specification
- Estimators used: FMOLS and DOLS, pooled and weighted pooled variants.
- Main estimator chosen: weighted pooled FMOLS (weighted FMOLS) for efficiency and stability in small-N, large-T panel and to account for heterogeneity in error variance.
- Baseline weighted pooled FMOLS specification (common long-run coefficients):
  - log GDPPC_it = α + β1.(EducEqual_it) + β2.log(Public Debt/GDP)_it + β3.(Corruption_it) + β4.log(Trade/GDP)_it + ε_it
- Estimation adjustments:
  - Bartlett kernel and Newey-West fixed bandwidth for serial correlation and heteroskedasticity corrections.
  - DOLS specifications include leads and lags of first differences of regressors (automatic selection by SIC).

### Baseline empirical results (weighted pooled FMOLS core estimates — Table 2a)
- Education Equality coefficients (weighted FMOLS, four progressive specifications I–IV):
  - I: 0.574*** (0.017)
  - II: 0.550*** (0.030)
  - III: 0.492*** (0.031)
  - IV: 0.582*** (0.048)
- Log Public Debt (% GDP) (selected columns):
  - II: 0.053** (0.022)
  - III: 0.061*** (0.022)
  - IV: 0.166*** (0.041)
- Political Corruption Index:
  - III: -1.933*** (0.026)
  - IV: -2.032*** (0.039)
- Log Trade Openness:
  - IV: 0.201*** (0.054)
- Sample and fit:
  - Observations: I 229; II 189; III 189; IV 179
  - Cross-sections: 4
  - R²: I 0.62; II 0.54; III 0.67; IV 0.69
- Interpretation:
  - Education equality: consistently positive and highly significant across specifications.
  - Political corruption: large negative association with log GDP per capita.
  - Public debt: positive association at moderate levels; relationship weakens at higher debt levels (nonlinear evidence below).
  - Trade openness: positive and significant in the preferred specification.

### Robustness checks and alternative estimators (Table 2b, 2c, 2d)
- Alternative estimators produce broadly consistent directions and significance.
- Selected estimates across FMOLS pooled, DOLS weighted pooled, and DOLS pooled (Table 2b):
  - Education Equality:
    - FMOLS Pooled: 0.639*** (0.088)
    - DOLS W. Pooled: 0.376*** (0.096)
    - DOLS Pooled: 0.536*** (0.089)
  - Log Public Debt (% GDP):
    - FMOLS Pooled: 0.122*** (0.042)
    - DOLS W. Pooled: 0.068** (0.027)
    - DOLS Pooled: 0.086* (0.047)
  - Political Corruption Index:
    - FMOLS Pooled: -2.387*** (0.425)
    - DOLS W. Pooled: -1.617*** (0.288)
    - DOLS Pooled: -1.988*** (0.450)
  - Log Trade Openness:
    - FMOLS Pooled: 0.201*** (0.055)
    - DOLS W. Pooled: 0.189*** (0.051)
    - DOLS Pooled: 0.210*** (0.050)
  - Observations: FMOLS Pooled 179; DOLS W. Pooled 181; DOLS Pooled 181
  - R²: 0.68, 0.72, 0.74 respectively
- Additional controls (selected coefficients from Tables 2c and 2d):
  - Log Investment/Capital Stock: 0.331*** (0.063)
  - Log Credit/Investment: 0.280*** (0.036)
  - Log Real Exchange Rate: -0.063** (0.027)
  - Log International Reserves: 0.180*** (0.056)
  - Log FDI % GDP (inflows): 0.113*** (0.015)
  - Health Equality: 0.261** (0.103)
  - Law Equality: 0.701*** (0.048)
  - Performance Legitimation: 0.410*** (0.043)
  - Initial GDP per capita (log): 0.263*** (0.099) in one specification (indicative of divergence among the four economies)

### Extensions focused on public debt (Table 2e)
- Nonlinearity and interaction with education:
  - Education Equality:
    - i: 0.481*** (0.049)
    - ii: 0.533*** (0.048)
    - iii: 0.388*** (0.051)
  - Log Public Debt (% GDP):
    - i: 0.300*** (0.046)
    - ii: 0.102** (0.041)
    - iii: 0.165*** (0.043)
  - Education × Log Debt (column ii): 0.233*** (0.053)
  - Log Public Debt (% GDP)² (column iii): -0.158*** (0.049)
  - Observations: 179; Cross-sections: 4
- Interpretation:
  - Public debt exhibits a nonlinear association with GDP per capita: positive at moderate levels, diminishing (negative squared term) at higher debt.
  - Positive and significant interaction (Education × Log Debt) suggests the positive association of debt with income is stronger when education equality is higher.
  - All estimates reported as associations, not causal effects.

### Synthesized key quantitative findings
- Institutional quality:
  - Political Corruption Index: consistently large negative association with log GDP per capita (e.g., ranges from about -1.6 to -2.4 across estimators).
  - Law equality, executive respect for constitution, and performance legitimation positively associated with higher income.
- Human capital:
  - Education equality: stable, strong positive association (coefficients typically between 0.38 and 0.64 across specifications and estimators).
  - Health equality: positive in robustness checks (e.g., 0.261** (0.103)).
- Trade openness and external integration:
  - Trade openness: generally positive and significant (e.g., 0.201*** (0.054) in baseline IV); FDI inflows and international reserves also positive and significant.
- Public debt:
  - Positive association at moderate levels; diminishing marginal association at higher debt levels (squared term negative and significant).
- Initial GDP per capita:
  - Positive association in some specifications, indicating divergence rather than conditional convergence among the four islands.

### Policy implications emphasized in the text
- Strengthen institutional quality and governance (reduce corruption; enhance rule of law; improve executive respect for constitutional rules).
- Invest in human capital development:
  - Promote education equality and expand access to high-quality basic education.
  - Improve health equity and public health spending.
- Promote trade openness and trade facilitation to support diversification and market access.
- Adopt sound macroeconomic policies that support productive investment and stability:
  - Use public borrowing to finance productive investment, mindful of diminishing returns at high debt levels.
  - Leverage reserves, exchange-rate flexibility, and policies that attract sustained FDI.
- Sectoral strategy for small islands:
  - Use institutional strength to translate sectoral advantages (e.g., tourism, fisheries) into sustained, diversified growth while managing exposure to external shocks.

### Methodological limitations, diagnostics, and suggestions for future research
- Limitations:
  - Small-N (four countries) limits external validity and power of unit-level inference.
  - Data constraints: limited coverage for V-WGI and other institutional indicators over full post-independence period; national accounts measurement differences may affect output measures, especially in Comoros and Madagascar.
  - Results interpreted as associations; causality not established.
- Diagnostics (selected test statistics and unit-root findings):
  - Cross-section dependence: most variables show significant Breusch-Pagan LM and Pesaran scaled LM statistics; probabilities often 0.000.
  - Unit root tests:
    - Trade (percentGDP): CIPS truncated t-stat -2.85; p-value <0.01 (but alternative t-stat -1.30; p-value >=0.10).
    - Log GDP per capita, Education Equality, Political Corruption Index, Log Public Debt: CIPS and other tests report p-value >=0.10 (unit-root cannot be rejected).
  - Post-estimation residual stationarity:
    - FMOLS Pool.W: Levin, Lin & Chu t* -3.09; CIPS Truncated CIPS <0.01 (stationary residuals).
    - DOLS Pool.W: Levin, Lin & Chu t* -2.91; CIPS Truncated CIPS <0.01.
- Suggestions for future research:
  - Expand sample to a broader set of small island developing states.
  - Examine the role of external shocks and climate vulnerability explicitly.
  - Use more granular measures of institutional quality and policy implementation.
  - Move beyond real GDP per capita to analyze poverty, inequality, resilience, and broader human development outcomes.

*IMF Working Paper — Introduction and Sections II–Conclusion excerpts from "Explaining Income Divergence in the Western Indian Ocean" (wpiea2026165-source-pdf).*

### Introduction ...........................................................................................................

### wpiea2026165-source-pdf - Introduction

### Overview and scope
- Focus: Explaining income divergence in four Western Indian Ocean island nations: Mauritius, Seychelles, Madagascar, and Comoros.
- Approach: Two-pronged—qualitative assessment of growth divergence and quantitative evaluation using Fully Modified Ordinary Least Squares (FMOLS) and Dynamic Ordinary Least Squares (DOLS).
- Time coverage: Qualitative assessment spans the post-independence period; quantitative analysis begins in 1980 due to data limitations (particularly for Comoros).

### Key empirical facts and trajectories
- Real GDP per capita changes since 1980 (exact figures preserved):
  - Mauritius: increase of 354 percent.
  - Seychelles: increase of 138 percent.
  - Madagascar: contraction of 35 percent.
  - Comoros: increase of 17 percent.
- Growth classification:
  - Mauritius and Seychelles: among top performers in Sub-Saharan Africa (SSA) with average annual real GDP growth rates of approximately 3 percent.
  - Comoros and Madagascar: experienced more limited growth.

### Main research question and methods
- Objective: Investigate key correlates of differences in growth trajectories of real GDP per capita across the four island nations.
- Qualitative component: national accounts data and literature-based review of structural, macroeconomic, and institutional factors.
- Quantitative component: FMOLS and DOLS estimations to assess long-run associations with growth using small, partially heterogeneous panel data.

### Principal findings (as stated in the source)
- Divergent post-independence growth trajectories are associated with differences in:
  - Institutional quality.
  - Macroeconomic and structural policies, especially those related to education and trade.
- Mauritius and Seychelles benefited from:
  - Stronger governance.
  - Policies that fostered productive investment.
  - Expanded educational access.
  - Promotion of trade openness.
- Initial conditions at independence varied, but policy frameworks and institutional evolution appear to have been associated with divergent development outcomes.
- Early advantages are captured in the core model through levels and trends in independent variables such as education equality, health outcomes, and corruption indices.
- Initial conditions are also tested explicitly as a separate variable in robustness checks.

### Theoretical and literature context (condensed)
- Theoretical frameworks referenced:
  - Solow-Swan model: emphasizes physical and human capital accumulation, exogenous technological progress, convergence prediction.
  - Endogenous growth theory: emphasizes policy, incentives, endogenous technological progress, role of savings, investment, education, and government policy in determining growth rates.
- Four broad categories of growth drivers discussed: macroeconomic policies, structural factors, institutional quality, and the external environment.
- Empirical literature highlights (selected findings as presented):
  - De Long and Summers (1991): equipment investment correlated with growth—an additional percentage point of GDP invested in equipment could lead to a GDP increase of approximately one-third of a percentage point.
  - Barro (1996): higher initial schooling and life expectancy, lower fertility, reduced government consumption, stronger rule of law, lower inflation, and improved terms of trade correlated with higher growth conditional on initial per capita GDP.
  - Sala-i-Martin et al. (2004): of 67 variables, 18 significantly correlated with growth; most influential included relative price of investment, primary school enrollment rates, and initial real GDP per capita.
  - Dao (2011): growth dependent on urban sanitation and water access, rural water access, gross capital formation as percent of GDP, adult literacy rates, bank liquid reserves relative to assets, globalization indices, and social/political globalization measures.
  - Chirwa and Odhiambo (2016): determinants differ by development level; in developing countries, key determinants include foreign aid, foreign direct investment, fiscal policy, investment levels, trade openness, human capital, demographics, monetary policy, natural resources, and geographic/political/financial influences.
  - African-focused studies: identify roles for investment, education, government effectiveness, trade openness, external debt, and vulnerability to shocks; effects can vary by specification and country set.

### Small states and sectoral considerations
- Mixed evidence on whether small size is a disadvantage:
  - Easterly and Kraay (2000): argue against treating small states as distinct for policy advice.
  - Aiyar (2008): no consistent evidence smallness harms growth.
  - Briguglio (2022): small states face constraints like limited economies of scale and high vulnerability to external shocks.
- Sectoral specialization (tourism) is prominent for small island economies: supports diversification and employment but increases exposure to external shocks and demand volatility.
- Institutional quality is critical in translating sectoral advantages into sustained growth.

### Region- and country-specific evidence
- Limited empirical work specific to the four islands due to data constraints.
- Hamori and Razafimahefa (2003): trade effects on growth depend on economy size and trade importance relative to GDP.
- Mauritius literature highlights drivers of the “growth miracle”: heterodox export strategy, strong institutions, favorable trading environment, well-managed Export Processing Zone, spending on education, avoidance of currency overvaluation, and business facilitation.

### Role of legal origins
- Legal origin literature (La Porta et al. 1998, 2008) discussed as historical context shaping institutional development; not explicitly operationalized in empirical analysis in this paper.
- Main propositions preserved from source: legal traditions influence legal rules and regulations, which correlate with economic and social outcomes; divergence between civil law (policy-implementing focus) and common law (market-supporting focus) affects institutional design and economic performance.

### Policy implications highlighted in the text
- Emphasized policy levers for similarly situated economies:
  - Strengthen institutional quality and governance.
  - Invest in human capital development (education and health).
  - Promote trade openness.
  - Adopt sound macroeconomic policies that support productive investment and stability.

*Source: IMF Working Paper — Introduction section of "Explaining Income Divergence in the Western Indian Ocean" (content unit: wpiea2026165-source-pdf - Introduction).*

### Section II: Qualitative Analysis

### Section II: Qualitative Analysis

### A. Geography, Political and Economic Background

- This section examines structural factors highlighted by the growth literature as likely to influence long-term growth performance: geography, political background, and economic structure. It shows that the four island nations share similar geographic features and broadly similar colonial legacies, although their political histories and post-independence trajectories differ.

a. Geographical settings and risks of natural disasters
- Comoros
  - Location: Mozambique Channel between the east coast of Africa and Madagascar.
  - Area/population: approximately 2,000 square kilometers; population nearly one million.
  - Climate: tropical marine; November to April hot and humid with northeastern monsoons and potential cyclones; rest of year cooler and drier.
  - Natural disaster losses (World Bank): tropical cyclones 64 percent of average losses; floods 35 percent.
- Madagascar
  - Location: Indian Ocean off the southeast coast of Africa.
  - Area/population: nearly 600,000 square kilometers; population approximately 30 million.
  - Geography: lowlands and steep bluffs on east coast; volcanic mountains in northern Tsaratanana Massif; protected harbors and broad plains on west coast; plateau and semi-arid region in southwest.
  - Climate: two seasons — hot and rainy from November to April; cooler and drier from May to October.
  - Vulnerabilities: highly vulnerable to climate change; significant risks from cyclones, floods, and droughts across different regions.
- Mauritius
  - Location: archipelago east of Madagascar.
  - Area/coastline/population: covering 2,040 km² with a coastline of 177 km; population over one million.
  - Composition: main island of Mauritius, Rodrigues, and atolls Agalega and Saint Brandon; part of volcanic Mascarene Islands.
  - Climate: humid and subtropical; hot and rainy November to April; cyclones peak December to March; cooler and drier May to October.
  - Disaster exposure: direct cyclone hits rare but effects experienced almost annually; wind, rain, and storm surges contribute to significant losses (World Bank).
- Seychelles
  - Location: western Indian Ocean.
  - Area/population/islands: about 500 km2; population about 120,000; comprises 115 islands including around 40 granitic islands and coralline islands stretching over 1,200 kilometers.
  - Population distribution: main inhabited islands Mahé, Praslin, La Digue; Mahé accounts for 86 percent of population.
  - Topography: granitic islands feature hills up to 940 meters; coralline islands are flat and lack fresh water.
  - Climate/risks: tropical and humid moderated by monsoons (cooler southeast winds May to September; warmer northwest winds March to May); flooding from intense rainfall leads to landslides, fatalities, and infrastructure damage.

b. Political background and governance
- Comoros
  - Independence: from France in 1975.
  - Political history: recurrent episodes of political instability including attempted or successful coups and repeated secession efforts by individual islands; 1989 assassination of President Abdallah noted.
  - Institutional consequences: instability and destructive cyclones in the 1980s hindered institutional development; formal governance structures remain relatively constrained; communities often rely on traditional governance and diaspora-funded initiatives.
  - Languages: Arabic and French official.
- Madagascar
  - Independence: June 26, 1960.
  - Political history: marked by several political crises, notably between 2009 and 2012.
  - Institutional constraints: centralized governance, weak accountability, limited institutional capacity, vulnerability to vested interests; reform implementation has sometimes been uneven and public engagement limited.
  - Languages/membership: Malagasy and French official; member of the ‘’Organisation Internationale de la Francophonie’’.
- Mauritius
  - Independence: March 12, 1968 (Mauritius Independence Act).
  - Political history: colonized by Dutch, French, British; path to independence marked by intercommunal tensions between Creole and Indian communities.
  - Institutions: stable democracy with relatively strong institutions; successive governments broadly committed to economic development.
  - Languages: English official in administration, courts, and business; English and French dominate education and professional settings.
- Seychelles
  - Independence: June 29, 1976.
  - Political history: same ruling party (under successive names) in power from 1977 to 2020; state-led development model; new constitution in 1993 reintroduced multiparty democracy; 2016 opposition coalition won control of legislature.
  - State role: state provided free public services, maintained a large public enterprise sector, and funded social programs through revenues from strategic sectors like tourism and fishing.
  - Languages: Seychelles Creole, English, French; English primary language of government and business.

c. Economic performance
- Comoros
  - Independence-era structure: predominantly agricultural with focus on vanilla, cloves, and ylang-ylang; subsistence farming widespread; fishing supported coastal livelihoods; minimal industrial sector and heavy reliance on imports.
  - Sectoral contributions to growth: in 1997 primary sector contributed 1.4 percentage points to growth; secondary sector added 0.4. By 2022 contributions declined to 0.7 (primary) and 0.1 (secondary), with tertiary contributing 0.5—reflecting a gradual but modest shift in economic structure.
- Madagascar
  - Independence-era structure (1960): agriculture accounted for 34 percent of GDP and employed most of the population; key crops rice, cassava, sweet potatoes, maize, coffee; export crops vanilla, cloves, sugar; livestock and coastal fishing important.
  - Growth performance: persistent institutional challenges and recurring political instability associated with weaker economic performance; long-term growth relatively weak compared with peers, prolonged downturns and inconsistent recoveries have kept poverty levels high; stop-go growth pattern and limited institutional effectiveness major barriers to inclusive development.
- Mauritius
  - Independence-era structure (1968): remote monocrop producer, agriculture-based, heavily reliant on sugarcane with tea, vegetables, cut flowers also contributing; agriculture accounted for over 22 percent of GDP; per capita income US$260 at independence.
  - Transformation: since independence, diversified into manufacturing, tourism, and financial services, overcoming early challenges of high unemployment, low wages, and economic volatility.
- Seychelles
  - Independence-era structure (1976): primarily agricultural centered on copra, cinnamon, vanilla, other spices; agriculture constrained by limited arable land and poor soil quality; fishing important; manufacturing minimal; tourism in infancy.
  - Sectoral contributions to growth: by 1999 primary sector contributed 0.4 percentage points to growth, secondary sector detracted -1.4. By 2022 primary, secondary, and tertiary sectors contributed 0.5, 1.1, and 10.8 percentage points respectively, reflecting a shift toward a service-based economy.

### B. Divergent Growth Trajectories: Exploring Associated Factors

- The analysis approaches drivers of economic activity in two ways: (i) a national accounts perspective showing how investment and trade shaped different trajectories; (ii) examination using literature-based determinants: macroeconomic policies, structural factors, institutional quality, and the external environment.
- Expectations guiding the analysis:
  - (i) improvements in institutional quality should be associated with higher income levels,
  - (ii) expansions in education and health should support income convergence,
  - (iii) trade openness should coincide with structural transformation and export growth.

- Core observation
  - Mauritius and Seychelles have significantly outperformed Madagascar and Comoros in per capita income growth over the last four decades. Initial income gaps at independence were relatively modest, but Mauritius and especially Seychelles experienced sustained and robust growth in real GDP per capita, while Madagascar and Comoros experienced relatively flat trajectories and widening development gaps.

a. National accounts perspective
- Key factors from GDP composition (nominal):
  - Trade openness (exports plus imports of goods and services as a percentage of GDP) and investment (gross fixed capital formation) were key engines of growth in Mauritius and Seychelles; these countries maintained high trade openness and capital formation.
  - Investment in Madagascar has been relatively strong but more volatile; investment in Comoros has been lower and more volatile.
  - Sectoral diversification and structural transformation: Mauritius transitioned from agriculture (sugar) to manufacturing (textiles) and later services (tourism, finance, ICT) from the late 1970s onward. Seychelles leveraged tourism and expanded fish processing during the 1990s and 2000s. These shifts were supported by investment and sustained trade integration.
  - Government consumption: spending in social sectors such as education contributed to long-term growth in Mauritius and Seychelles; public spending was limited in Madagascar and Comoros, reflecting fiscal constraints and weaker state capacity.
  - Private consumption: as a share of GDP, private consumption has been comparatively higher in Madagascar and Comoros, reflecting lower investment and government spending.

b. Macroeconomic policies, structural factors, institutional quality, and external environment
- Macroeconomic stability
  - Over the post-1980 period, macroeconomic stability was generally stronger in Mauritius and Seychelles, while Madagascar and Comoros experienced greater volatility—associated with weaker investment and growth outcomes.
  - This suggests a positive association between macroeconomic stability and income growth (quantitatively tested in Section III).
- Public debt
  - While public debt has increased in both Mauritius and Seychelles, this rise has been accompanied by sustained gains in income per capita (illustrated in referenced figures). 

*Italic source-attribution: Section II: Qualitative Analysis, wpiea2026165-source-pdf*

### 1. This suggests an association

### Explaining Income Divergence in the Western Indian Ocean

### Association between public borrowing and productive investment
- The text suggests an association between public borrowing and productive investments, including infrastructure, rather than a causal interpretation.
- Examples:
  - Mauritius: public funds supported a diversification strategy shifting the economy from a sugar-dependent base toward textiles, tourism, and financial services.
  - Seychelles: significant investments in education and healthcare, resulting in universal health coverage and high literacy rates; attracted substantial foreign direct investment to develop a strong tourism sector, supported by improvements in infrastructure and service quality.
- Contrast:
  - Comoros and Madagascar: public debt increased rapidly but without a corresponding improvement in income per capita; possible reasons include less effective use of borrowing for productive investment and external factors such as political instability or constrained governance.

### Infrastructure- and external-integration indicators
- Panel 3 highlights stronger infrastructure-related performance in Mauritius and Seychelles (e.g., mobile and internet access increased strongly from the 1990s onward).
- Exchange rate and reserves:
  - Comoros: modest demand growth and a relatively stable exchange rate maintained through a peg to the euro; peg supported by relatively comfortable foreign exchange reserves and remittance inflows from France.
  - Mauritius, Seychelles, and Madagascar: moved from fixed exchange rate arrangements to more flexible ones, reducing pressures to accumulate reserves.
- Foreign direct investment:
  - FDI inflows have been significantly higher and more sustained in Mauritius and Seychelles compared with Madagascar and Comoros (figure referenced).
- Tariff trends:
  - Figure 5: Tariff rate, applied, simple mean, all products (%) shown across periods 1980-89, 1990-99, 2000-2009, 2010-2019 for Comoros, Madagascar, Mauritius, Seychelles.

### Social policies: education and health
- Education (Panel 4):
  - Both Seychelles and Mauritius prioritized education through comparatively higher government spending (see Panel 4, first chart).
  - Mauritius: free primary, secondary, and tertiary education (historical note footnoted: primary free prior to independence; secondary free in 1977; tertiary free 1965–2000, fee paying 2000–2019, free from 2019).
  - Seychelles: one of the highest literacy rates in Africa due to comprehensive inclusive educational policies.
  - Alternative indicators, such as the education equality index (V-Dem), show Seychelles and Mauritius performing strongly.
  - Panel 4 charts include:
    - Government spending in education (% GDP) over 1980–2020 for the four countries.
    - School enrollment, secondary (% gross) with missing data interpolated.
    - Literacy rate, adult total (% of people ages 15 and above) (Most recent available data).
    - Educational equality index (V-Dem, average 2020-22).
- Health (Panel 5):
  - Mauritius and Seychelles: stronger health outcomes due to proactive social policies and higher government spending in healthcare (see panel 5).
  - Mauritius: free public healthcare services, high life expectancy, low infant mortality rates.
  - Seychelles: universal healthcare coverage and significant improvements in public health indicators.
  - Panel 5 charts include:
    - Domestic general government health expenditure (% of GDP) over 2000–2018 for the four countries.
    - Life expectancy at birth, total (years) across periods 1980-89, 1990-99, 2000-2009, 2010-2019.
    - Mortality rate across the same periods.
    - Health equality index (V-Dem, average 2020-22).
  - V-Dem health equality: captures extent to which basic health services are accessible across social groups; higher scores indicate more equitable access.

### Political environment and institutional quality
- Political stability:
  - Mauritius and Seychelles: relatively stable political environments facilitating consistent policy implementation and investor confidence.
  - Madagascar and Comoros: more volatile political contexts.
- Institutional strengthening:
  - Mauritius and Seychelles: sustained efforts to improve transparency, accountability, and public administration efficiency; Mauritius implemented anti-corruption measures and regulatory improvements; Seychelles undertook public sector reforms.
- Governance challenges in Madagascar and Comoros have constrained their ability to leverage public borrowing and attract investment.

### Trade policy, export diversification, and FDI
- Trade agreements and market access:
  - Mauritius: long-standing WTO membership and preferential access to major markets, particularly the European Union and the United States (including through AGOA).
  - Seychelles: trade linkages with the European Union and past access to the U.S. market under AGOA until its graduation in 2017.
- Export promotion and diversification:
  - Mauritius: export diversification in sugar, textiles, and clothing; emphasis on tourism, textiles, and financial services; investment incentives reinforced export promotion.
  - Seychelles: focus on tourism, fisheries, and offshore financial services; policies created an attractive environment for FDI via improved infrastructure, open foreign exchange policy, and tax incentives.
- Trade facilitation:
  - Mauritius and Seychelles improved trade facilitation by reducing tariffs (see figure 5), improving customs procedures and port infrastructure, and enhancing logistics services.
- Contrasts:
  - Madagascar and Comoros: limited expansion of investment in social sectors; weaker institutional effectiveness; less conducive trade environment; constraints in infrastructure and customs procedures.

### Historical episodes, reforms, and shocks
- Mauritius: establishment of the Export Processing Zone during the 1970s and 1980s marked a pivotal trade reform associated with sustained growth and export diversification.
- Seychelles: expansion of the tourism sector during the 1990s facilitated a structural transformation toward a service-based economy and stronger income performance.
- Madagascar: major political crisis during 2009–12 disrupted economic activity and associated with weaker growth outcomes.
- Comoros: recurrent episodes of instability from independence through the 2000s constrained policy continuity and dampened economic performance.
- These episodes illustrate how timing and nature of reforms and shocks are closely associated with divergence in growth trajectories.

### Section II — Quantitative analysis: scope, data, and main variables
- Purpose:
  - Tests a small number of key correlates identified in the qualitative analysis; does not seek an exhaustive explanation of income divergence.
- Data constraints:
  - Data availability is a major constraint, particularly for Comoros.
  - Standard indicators on education, health, and institutional quality (e.g., World Governance Indicators) are unavailable for much of 1970–2022.
  - Alternative proxies used from V-Dem for longer historical coverage and comparability.
  - Empirical analysis restricted to period starting in 1980—the earliest year for which a consistent comparable dataset across all four countries could be constructed.
  - National accounts revisions and differences in treatment of informal and subsistence activities may affect measured output levels, especially in Comoros and Madagascar.
- Main variables (as summarized in the variable dictionary):
  - Real GDP per capita: outcome variable; real GDP per capita in constant 2015 USD expressed in natural logarithms. Source: World Bank World Development Indicators.
  - Public debt: General government public debt as a percent of GDP; used in natural logarithms. The squared logarithm of public debt is included in nonlinear specifications. Source: IMF World Economic Outlook and related IMF databases, subject to country coverage.
  - Political Corruption Index: institutional-quality indicator measuring extent of political corruption, used in levels. Source: Varieties of Democracy database.
  - Education Equality Indicator: measures extent to which access to high-quality basic education is distributed equally across the population, used in levels. Source: Varieties of Democracy database.
  - Trade openness: exports plus imports of goods and services as a percent of GDP; used in natural logarithms. Source: World Bank World Development Indicators.
  - Additional controls: Initial GDP per capita, investment or capital-stock measures, credit-to-investment ratio, real exchange rate, reserves, FDI, health equality, and governance indicators. Sources: World Bank, IMF databases, and V-Dem, depending on variable coverage.
- Macroeconomic policy interpretation:
  - Public Debt (percent of GDP): captures fiscal stance; moderate debt may support growth by financing productive investment, while excessive debt can hinder growth due to crowding out and debt overhang.
- Institutional quality:
  - Political Corruption Index (V-Dem): measures extent of political corruption across branches of government; high corruption expected to negatively impact growth by deterring investment and weakening institutional efficiency.
- Structural policies:
  - Social Policies: Education Equality (V-Dem) reflects equitable availability of high-quality basic education; greater educational equality expected to enhance human capital and support long-term growth.
  - Trade Policies: Trade Openness defined as sum of exports and imports as a percentage of GDP; openness may improve efficiency and market access but can expose small undiversified economies to external shocks.
- Other external factors considered in robustness checks include foreign direct investment, reserve accumulation, and exchange rate dynamics; data limitations prevent inclusion of more granular external indicators in the baseline specification.

### Empirical strategy and methodology
- Estimation methods:
  - Panel Fully Modified Ordinary Least Squares (FMOLS) and Dynamic Ordinary Least Squares (DOLS) used to address endogeneity and serial correlation in non-stationary panel data.
  - DOLS incorporates leads and lags of first differences of regressors to mitigate endogeneity and omitted dynamics; often achieves higher efficiency in finite samples compared with FMOLS.
  - FMOLS corrects for long-run covariance to ensure unbiased estimates; DOLS dynamically adjusts for endogenous feedback effects.
- Rationale:
  - Methods are appropriate for small samples and partially heterogeneous panels and robust in handling non-stationary panel data with cointegration, small sample sizes, and endogeneity.
- Next steps (empirical strategy will be as follows): the source continues beyond this excerpt to describe the implementation of FMOLS and DOLS estimations and robustness checks.

*Source: IMF Working Paper — Explaining Income Divergence in the Western Indian Ocean*

### 1. We start by providing background on the different estimators that will be used: FMOLS and DOLS,

### 1. We start by providing background on the different estimators that will be used: FMOLS and DOLS,

### Background on estimators and choice of main estimator
- Estimators used: FMOLS and DOLS, both pooled and weighted pooled variants.
- Main estimator selected: weighted pooled FMOLS (referred to as weighted FMOLS) because:
  - It is efficient and stable in a small-N, large-T panel setting.
  - It handles cross-sectional dependence without overfitting.
  - It acknowledges heterogeneity in the error term while pooling slope coefficients.
- Sample: four partially heterogeneous island economies (differences in openness, political and institutional conditions; shared traits: small island status, import dependence, vulnerability to external environment).
- FMOLS-MG was considered theoretically valid but proved unreliable with only four countries: unit-level estimates unstable, mean-group average sensitive to outliers, weak inference with so small an N.

### Estimator specifications (explicit forms and features)
- Common setup: dependent variable is log GDP per capita (log GDPPC_it), annual data from 1980 to 2022, coefficients α and β constant across cross-sections in pooled specifications, ε_it idiosyncratic error.
- FMOLS Pooled Specification:
  - log GDPPC_it = α + β1.(EducEqual_it) + β2.log(Public Debt/GDP)_it + β3.(Corruption_it) + β4.log(Trade/GDP)_it + ε_it
  - Coefficients β1–β4 represent common long-run coefficients.
- FMOLS Weighted Pooled Specification (Baseline Estimator):
  - Same specification as pooled FMOLS but incorporates weights to account for differences in cross-sectional sample sizes or variance contributions; weights give greater importance to more reliable cross-sections.
- DOLS Pooled Specification:
  - Same long-run homogeneity assumption as pooled FMOLS but adds dynamic leads and lags of independent variables to address endogeneity:
    - log GDPPC_it = α + β1.(EducEqual_it) + β2.log(Public Debt/GDP)_it + β3.(Corruption_it) + β4.log(Trade/GDP)_it + Σ_k δ_k ΔX_it−k + ε_it
  - δ_k are coefficients on leads and lags of differenced regressors.
- DOLS Weighted Pooled Specification:
  - Pooled DOLS specification with a weighting mechanism analogous to weighted FMOLS.
- Estimation adjustments:
  - All weighted pooled estimates use Bartlett kernel and Newey-West fixed bandwidth to correct for serial correlation and heteroskedasticity.
  - Trade enters as a deterministic regressor.
  - DOLS estimates apply automatic leads and lags specification based on the SIC criterion.

### Baseline results (weighted pooled FMOLS core estimates — Table 2a)
- Four progressive model columns (I, II, III, IV); robust standard errors in parentheses; significance codes: *** p<0.01, ** p<0.05, * p<0.1.
- Education Equality coefficients:
  - I: 0.574*** (0.017)
  - II: 0.550*** (0.030)
  - III: 0.492*** (0.031)
  - IV: 0.582*** (0.048)
- Log Public Debt (% GDP):
  - II: 0.053** (0.022)
  - III: 0.061*** (0.022)
  - IV: 0.166*** (0.041)
- Political Corruption Index:
  - III: -1.933*** (0.026)
  - IV: -2.032*** (0.039)
- Log Trade Openness:
  - IV: 0.201*** (0.054)
- Sample and fit:
  - Observations: I 229, II 189, III 189, IV 179
  - Cross-sections: 4 (all columns)
  - R²: I 0.62, II 0.54, III 0.67, IV 0.69
- Interpretation notes:
  - Education equality is consistently positive and highly significant (coefficients range 0.49 to 0.58 across specifications).
  - Public debt positive and significant at moderate levels; relationship described as associative rather than causal and weakens at higher debt levels (nonlinear results reported below).
  - Political corruption index negative and highly significant (around -1.9).
  - Trade openness positive and significant (0.20 in IV).
  - Results underscore complementary roles of education, institutional quality, fiscal capacity, and trade integration.

### Robustness checks: estimator sensitivity (Table 2b)
- Alternative estimators used: pooled FMOLS, weighted pooled DOLS, pooled DOLS.
- Table 2b coefficients and standard errors:
  - Education Equality:
    - FMOLS Pooled: 0.639*** (0.088)
    - DOLS W. Pooled: 0.376*** (0.096)
    - DOLS Pooled: 0.536*** (0.089)
  - Log Public Debt (% GDP):
    - FMOLS Pooled: 0.122*** (0.042)
    - DOLS W. Pooled: 0.068** (0.027)
    - DOLS Pooled: 0.086* (0.047)
  - Political Corruption Index:
    - FMOLS Pooled: -2.387*** (0.425)
    - DOLS W. Pooled: -1.617*** (0.288)
    - DOLS Pooled: -1.988*** (0.450)
  - Log Trade Openness (% GDP):
    - FMOLS Pooled: 0.201*** (0.055)
    - DOLS W. Pooled: 0.189*** (0.051)
    - DOLS Pooled: 0.210*** (0.050)
  - Observations: FMOLS Pooled 179; DOLS W. Pooled 181; DOLS Pooled 181
  - Cross-sections: 4; R²: 0.68, 0.72, 0.74 respectively
- Robustness conclusion:
  - Direction, magnitude, and statistical significance of coefficients broadly consistent across estimators.
  - Education equality coefficient ranges from 0.38 (weighted pooled DOLS reported elsewhere) to 0.64 (pooled FMOLS).
  - Corruption coefficient consistently negative across estimators (range –1.6 to –2.4).
  - Trade openness positive and significant across specifications.
- Nonlinear debt check:
  - Introduction of a squared public debt term (Table 2e, discussed below) confirms a non-linear relationship: marginal association of debt with growth declines at higher levels.

### Robustness checks: inclusion of additional variables (Tables 2c and 2d)
- Weighted pooled FMOLS used to test sensitivity to additional macroeconomic, institutional, and social variables.
- Table 2c selected coefficients (columns i–iv):
  - Education Equality: i 0.636*** (0.049); ii 0.536*** (0.051); iii 0.466*** (0.053); iv 0.558*** (0.054)
  - Log Public Debt (% GDP): i 0.314*** (0.063); ii 0.188*** (0.048); iii 0.139*** (0.042); iv 0.154*** (0.044)
  - Political Corruption Index: i -1.212*** (0.059); ii -2.422*** (0.046); iii -1.190*** (0.041); iv -2.098*** (0.042)
  - Log Trade Openness: i 0.327*** (0.082); ii 0.102* (0.053); iii 0.105** (0.046); iv 0.093* (0.051)
  - Log Initial GDPPC: 0.263*** (0.099) (appears in one specification)
  - Log Investment/Capital Stock: 0.331*** (0.063) (appears in one specification)
  - Log Credit/Investment: 0.280*** (0.036) (appears in one specification)
  - Log Real Exchange Rate: -0.063** (0.027) (appears in one specification)
  - Cross-sections: 4; Observations vary: 164, 167, 179, 179
- Table 2d selected coefficients (columns v–xi):
  - Education Equality coefficients vary: 0.549*** (0.050), 0.342*** (0.069), 0.451*** (0.049), 1.025*** (0.048), 0.942*** (0.046), 0.968*** (0.049), 0.445*** (0.048)
  - Log Public Debt (% GDP): 0.148*** (0.044), 0.130** (0.052), 0.105*** (0.039), 0.087** (0.041), 0.211*** (0.032), 0.096** (0.041), 0.147*** (0.044)
  - Political Corruption Index: -1.753*** (0.041), -2.046*** (0.048), -2.331*** (0.039), -2.232*** (0.039), -2.262*** (0.029), -2.643*** (0.039), -1.803*** (0.039)
  - Log Trade Openness: varying signs and significance across models (e.g., -0.041 (0.047), -0.077 (0.054), 0.155*** (0.050), 0.280*** (0.076), 0.278*** (0.101), 0.300*** (0.073), 0.171** (0.048))
  - Selected additional positive and significant variables:
    - Log International Reserves: 0.180*** (0.056)
    - Log FDI % GDP (inflows): 0.113*** (0.015)
    - Health Equality: 0.261** (0.103)
    - Law Equality: 0.701*** (0.048)
    - Performance Legitimation: 0.410*** (0.043)
    - Executive Respect for Constitution: 0.087*** (0.033)
    - Head of State Appointment: 0.052*** (0.010)
  - Cross-sections: 4; Observations vary by column (179, 163, 179, etc.)
- Interpretations from additional controls:
  - Public debt positive at moderate levels but diminishing returns (negative squared term reported in extensions).
  - Real exchange rate and private credit-to-investment ratio significant: external competitiveness and financial intermediation linked to higher income.
  - Investment-to-capital stock ratio significant: importance of physical capital accumulation.
  - Positive coefficient on initial GDP per capita indicates divergence among the four island economies, reflecting structural asymmetries (institutional capacity and governance).

### Extensions: focus on public debt (Table 2e and interaction with education)
- Two policy extensions examined:
  1. Nonlinearity: introduce squared public debt term.
  2. Interaction: public debt × education equality.
- Table 2e selected coefficients (FMOLS Weighted; columns i–iii):
  - Education Equality: i 0.481*** (0.049); ii 0.533*** (0.048); iii 0.388*** (0.051)
  - Log Public Debt (% GDP): i 0.300*** (0.046); ii 0.102** (0.041); iii 0.165*** (0.043)
  - Political Corruption Index: i -1.936*** (0.039); ii -2.109*** (0.040); iii -2.047*** (0.040)
  - Log Trade Openness: i 0.406** (0.160); ii 0.147*** (0.054); iii 0.106* (0.054)
  - Education × Log Debt: column ii 0.233*** (0.053)
  - Log Public Debt (% GDP)²: column iii -0.158*** (0.049)
  - Cross-sections: 4; Observations: 179 for all columns
- Key extension findings:
  - Public debt shows a nonlinear association with GDP per capita: positive at moderate levels, diminishing marginal association as debt rises (negative and significant squared term).
  - Interaction between public debt and education equality is positive and significant (0.233***), implying the positive association between debt and income is stronger when education equality is higher.
  - Results interpreted as associations, not causal estimates.

### Quantitative analysis: synthesized key findings
- Institutional quality:
  - Central correlate of growth: perceived corruption negative and significant; law and liberty protection, executive respect for constitutional rules, performance-based legitimation positively associated with GDP per capita.
- Human capital:
  - Education equality strong and stable positive association with GDP per capita; health equality adds explanatory power in robustness checks.
- Trade openness:
  - Generally positive and significant; effect varies when accounting for exchange rate, reserves, or FDI; international reserves and FDI inflows themselves are consistently positive and significant.
- Public debt:
  - Conditional role in growth: positive association at moderate debt levels, diminishing returns at higher debt (consistent with debt-overhang interpretation).
- Initial GDP per capita:
  - Positive association with subsequent income outcomes (contrary to conditional convergence), indicating divergence among the four economies; highlights the role of initial capital base (Seychelles and Mauritius examples mentioned in text).
- Broader implication:
  - Core findings robust across FMOLS and DOLS variants and under pooled and weighted pooled structures, supporting the association of education equality, trade openness, institutional quality, and fiscal structure with GDP per capita variation in small island developing states.

*IMF Working Paper — Explaining Income Divergence in the Western Indian Ocean (excerpts from supplied content).*

### Conclusion

### Conclusion

### Main findings on divergence in the Western Indian Ocean
- The study analyzes divergence in economic growth trajectories among Mauritius, Seychelles, Madagascar, and Comoros since 1980 using qualitative evidence and FMOLS/DOLS estimations.
- Institutional quality, human capital, trade openness, and fiscal conditions are identified as important correlates of long-run growth.
- Mauritius and Seychelles generally benefited from stronger institutions and more stable policy frameworks; Madagascar and Comoros experienced more persistent political and institutional challenges.

### Key empirical findings
- Geography and historical legacies influenced early development paths, but differences in policy frameworks and institutional evolution are associated with the divergence in income outcomes among the four island economies.
- Stronger governance, greater respect for the rule of law, and higher levels of performance-based legitimacy are positively associated with higher levels of real GDP per capita.
- More equitable access to education (and, to a lesser extent, better health outcomes) is associated with higher income levels, underscoring the importance of investment in human capital.
- Greater trade openness is positively associated with income levels.
- Moderate public debt is associated with higher GDP per capita, although the relationship weakens at higher debt levels.

### Methodological limitations and data constraints
- Data constraints required use of proxy variables for several institutional and structural factors.
- The small sample of countries limits the generalizability of the findings.
- Trade was treated as a deterministic regressor in FMOLS and DOLS estimations because:
  - trade reflects underlying structural characteristics (heavy reliance on imports of essential goods, narrow export base, reliance on foreign direct investment and external financing),
  - and when tested, trade was I(0) with a deterministic constant but I(1) when the constant was removed.

### Suggestions for future research
- Extend the analysis by incorporating a broader sample of small island developing states.
- Examine the role of external shocks and climate vulnerability.
- Use more granular measures of institutional quality and policy implementation.
- Move beyond real GDP per capita to examine how institutional quality and structural policies influence broader development outcomes, including poverty reduction, inequality, resilience to external shocks, human development, and overall economic well-being.

### Diagnostic and estimation notes (selected results)
- All variables tested are cross-sectionally dependent.
- Data were cross-sectionally demeaned to remove fixed effects prior to FMOLS/DOLS estimation.
- Unit root tests: all variables are I(1) except trade which tested I(0) with a deterministic constant (and I(1) when constant removed).
- Cross-sectional dependence test statistics (selected):
  - Log (Public Debt percentGDP): Breusch-Pagan LM 71.01; Pesaran scaled LM 18.77; Bias-corrected scaled LM 18.73; Pesaran CD 3.77; Probabilities 0.000.
  - Education Equality: Breusch-Pagan LM 170.96; Pesaran scaled LM 47.62; Bias-corrected scaled LM 47.59; Pesaran CD 6.64; Probabilities 0.000.
  - Political Corruption index: Breusch-Pagan LM 82.5; Pesaran scaled LM 22.1; Bias-corrected scaled LM 22.1; Pesaran CD 8.3; Probabilities 0.000.
  - GDP per capita (constant): Breusch-Pagan LM 199.05; Pesaran scaled LM 55.73; Bias-corrected scaled LM 55.70; Pesaran CD -0.30; Probabilities 0.000, 0.765 for Pesaran CD.
  - Trade (percentGDP): Breusch-Pagan LM 55.79; Pesaran scaled LM 14.37; Bias-corrected scaled LM 14.34; Pesaran CD 3.75; Probabilities 0.000.
- CIPS unit root test results (selected):
  - Log of GDP per Capita: t-stat -1.81; p-value >=0.10.
  - Education Equality: t-stat -2.18; p-value >=0.10.
  - Political Corruption Index: t-stat -0.98; p-value >=0.10.
  - Log (Public Debt percentGDP): t-stat -1.69; p-value >=0.10.
  - Trade (percentGDP): t-stat -2.85; p-value <0.01 (Truncated CIPS t-stat -2.85; p-value <0.01); alternative t-stat -1.30; p-value >=0.10.
- Post-estimation cross-section dependence on residuals (selected):
  - FMOLS Pool.W: Breusch-Pagan LM 19.97; Pesaran scaled LM 4.03; Bias-corrected scaled LM 3.99; Pesaran CD 4.14; Probabilities 0.00.
  - DOLS Pool.W: Breusch-Pagan LM 19.14; Pesaran scaled LM 3.79; Bias-corrected scaled LM 3.76; Pesaran CD 3.58; Probabilities 0.00.
  - FMOLS Pooled: Breusch-Pagan LM 11.00; Pesaran scaled LM 1.44; Bias-corrected scaled LM 1.40; Pesaran CD 2.80; Probabilities 0.09, 0.15, 0.16, 0.01 respectively.
  - DOLS Pooled: Breusch-Pagan LM 17.33; Pesaran scaled LM 3.27; Bias-corrected scaled LM 3.23; Pesaran CD 3.65; Probabilities 0.01, 0.00, 0.00, 0.00 respectively.
- Post-estimation unit root tests on residuals (selected):
  - FMOLS Pool.W: Levin, Lin & Chu t* -3.09; CIPS Truncated CIPS <0.01; Probabilities indicate stationarity.
  - DOLS Pool.W: Levin, Lin & Chu t* -2.91; CIPS Truncated CIPS <0.01; Probabilities indicate stationarity.
  - FMOLS Pooled: Im, Pesaran and Shin W-stat -2.40; ADF - Fisher X2 22.28; PP - Fisher X2 22.41; Levin, Lin & Chu t* -0.26; Probabilities 0.008, 0.004, 0.004, 0.399.
  - DOLS Pooled: Levin, Lin & Chu t* -2.73; CIPS Truncated CIPS <0.01.
- Note: Null hypotheses reported for tests:
  - Cross-section dependence tests: Null hypothesis is No cross-section dependence (correlation).
  - Unit root tests: Null hypothesis is Unit root.

*IMF Working Paper No. WP/2026/165*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026165-source-pdf.pdf_