## 1. Full Sample QGI-based Ranking

## Source details

**Canonical URL:** [1. Full Sample QGI-based Ranking](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14172.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14172.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14172.pdf.json)

---

### I. Introduction: purpose and conceptualization
- Objective: introduce a Quality of Growth Index (QGI) that encompasses both the intrinsic nature of growth and its social dimensions to assess the “quality of growth” across developing and emerging countries over 1990–2011.
- Conceptual definition: “good quality growth” = high, durable, and socially-friendly growth; emphasizes that how growth is generated (path from level L1 to L2 of income) matters for social outcomes.
- Positioning vs. existing indices:
  - QGI goes beyond the Human Development Index (HDI) by concentrating on the nature of growth rather than just income levels accumulated over long periods.
  - QGI differs from the Social Progress Index by incorporating growth fundamentals (pace, stability, sectoral breadth, outward-orientation) alongside social outcomes.
- Key research questions:
  - How has the quality of growth evolved over time and across regions?
  - Is there convergence in QGI dynamics or a “growth quality trap”?
  - How is QGI related to other development indicators?
- Main findings:
  - The quality of growth has been improving in the majority of countries over the past two decades.
  - The rate of convergence is relatively slow.
  - Considerable cross-country variations across income levels and regions.
  - Political stability, public pro-poor spending, macroeconomic stability, financial development, institutional quality and external factors such as FDI are associated with higher QGI.

### II. Methodology and index structure
- QGI composition:
  - QGI = α·(growth fundamentals) + β·(social dimension)
  - Two building blocks: intrinsic nature of growth sub-index (“growth fundamentals”) and social dimension sub-index (“social outcomes”).
- Growth fundamentals sub-index (four dimensions):
  - Strength (γ1): annual change in real GDP per capita.
  - Stability (γ2): inverse of the coefficient of variation (CV) of growth using a five-year rolling window (CV = standard deviation / average).
  - Diversification of sources (γ3): proxied by 1 − Herfindahl-Hirschman index (HHI) using exports data.
  - Outward-orientation (γ4): proxied by share of net external demand (exports − imports), both as percent of GDP.
- Social dimension sub-index (two components):
  - Health (δ1): aggregation of the reverse of infant mortality rate and life expectancy at birth.
  - Education (δ2): primary school completion rate.
- Interpretation notes:
  - Higher CV implies lower inverse-CV and lower stability; instability is harmful to poverty and equity via hysteresis.
  - Outward orientation can raise productivity but may increase vulnerability to external shocks; volatility accounted for separately.

### III. Standardization, weighting, and aggregation
- Standardization:
  - Z-score: Z = (X − μ) / σ (discussed; sensitive to outliers and unbounded).
  - Min-Max: Z' = (X − Xmin) / (Xmax − Xmin); benchmark uses Xmin and Xmax observed in the sample.
- Benchmark weighting:
  - α = 50 percent; β = 50 percent.
  - Within growth fundamentals: γ1 = γ2 = γ3 = γ4 = 25 percent.
  - Within social dimension: δ1 = δ2 = 50 percent.
  - Health sub-index sub-components: 50 percent each.
- Aggregation:
  - Benchmark: arithmetic mean (simple averaging) for sub-indices and components.
  - Robustness: geometric averaging considered to account for complementarities.
  - Formal note: arithmetic and geometric formulations presented; equal weighting implies substitutability among components, geometric averaging addresses complementarities.

### IV. Data and sample
- Panel coverage: 93 developing countries over 1990–2011.
  - 57 middle-income countries; 36 low-income countries.
- Temporal aggregation: five-year averages to smooth short-term fluctuations:
  - Periods: 1990–94, 1995–99, 2000–04, 2005–11.
- Data sources: IMF World Economic Outlook, World Bank WDI, COMTRADE, International Country Risk Guide, Barro and Lee (2010), Xala-i-Martin (2006).
- Missing data handling:
  - Primary school completion rate: missing observations imputed per Appendix 2 (method uses Barro and Lee average years of primary schooling divided by average duration and neighborhood adjustment over 5 years before/after).
  - Export diversification proxied using exports data due to missing sectoral value added.

### V. Robustness and sensitivity
- Sensitivity tests:
  - Alternative weights for α, β, γi, δi.
  - Alternative aggregation: geometric mean vs. arithmetic mean.
  - Alternative standardization choices; benchmark uses sample observed min/max for Min-Max scaling.
- Justification:
  - Min-Max with sample observed extrema reduces subjectivity compared with hypothetical ideal extrema.
  - Arithmetic mean preferred for simplicity and comparability; geometric mean used to test sensitivity to substitutability/complementarity.

### VI. Key empirical takeaways and specifications (exact figures)
- Sample size: 93 developing countries.
- Income group composition: 57 middle-income countries; 36 low-income countries.
- Time coverage: 1990–2011, averaged over five-year windows (1990–94, 1995–99, 2000–04, 2005–11).
- Benchmark weighting scheme:
  - α = 50 percent; β = 50 percent.
  - γ1 = γ2 = γ3 = γ4 = 25 percent.
  - δ1 = δ2 = 50 percent.
  - Health sub-index: two sub-components weighted at 50 percent each.
- Standardization: Min-Max using sample observed Xmin and Xmax.
- Aggregation: arithmetic mean (benchmark); geometric mean used for robustness.

### III. RESULTS — Overview: stylized facts, ranking, convergence, and comparisons
- Stylized facts (period 2005–2011 rankings and distribution):
  - Top performers: Bulgaria 0.843, China 0.842, Argentina 0.830.
  - Poorest performers: Chad 0.334, Central African Republic 0.402, Niger 0.415.
  - Aggregate QGI statistics:
    - Average QGI: 0.604.
    - Minimum QGI: 0.258 for Niger over 1990–94.
    - Maximum QGI: 0.849 for China over 2000–04.
    - QGI increases from 0.556 in 1990–94 to 0.656 in 2005–2011.
  - Distribution dynamics: density plots indicate rightward shift and narrowing over time, signaling some convergence.
- Regional and income-level patterns:
  - LA countries exhibit the highest QGI scores.
  - Sub-Saharan Africa (SSA) exhibits the lowest QGI and the flattest/leftmost density with thick tails; roughly more than 60 percent of SSA observations lie below the full sample average score (0.604).
  - Income-level ordering: QGI positively correlated with income level: UMIC > LMIC > LIC.
- Fragility and resource endowment:
  - Fragile countries underperformed the sample average by almost 16 percentage point.
  - Resource-rich countries have QGI scores slightly lower than non-resource rich peers; natural resource endowment tended to be more of a curse than a blessing for achieving higher QGI.

### B. Convergence hypothesis (empirical evidence)
- Pooled OLS evidence (selected exact figures):
  - Lagged QGI (one period) coefficient: -0.066*** (standard error (0.0156)).
  - Initial QGI (1990-94) coefficient: -0.068*** (standard error (0.016)).
  - Observations: 279.
  - R-squared: 0.072 (column 1) and 0.074 (column 2).
  - Note: Robust standard errors in brackets; *, **, and *** indicate significance at 10 percent, 5 percent, and 1 percent.
- Interpretation:
  - Past QGI performance is negatively associated with subsequent QGI growth rates, suggesting convergence: lower-performing countries tend to catch up over time, but convergence is relatively slow.
- Country categorization (1990–94 vs 2005–2011):
  - Region 2 (“hopefuls”): below sample average in both periods but improved — largely low-income and/or fragile countries, mostly from SSA and some from MENA.
  - Region 3 (“contenders”): improved from below to above the sample average — examples include Bangladesh, Laos, Nepal, Ghana, Tanzania, Zambia, Algeria, Iran, Morocco, Azerbaijan, Tajikistan, Uzbekistan, Guatemala, Nicaragua.
  - Region 4 (“club of best performers”): above sample average in both periods and improved — chiefly upper-middle and lower-middle income countries; includes Kenya, Namibia, South Africa.
  - Region 5 (“club of superior performers”): above sample average in both periods but experienced a mild drop — comprises Botswana and Malaysia.
  - Virtually all countries improved QGI over the two decades.

### C. QGI versus existing development indicators (pairwise correlations and fits)
- UN HDI:
  - Strong positive relationship: y = 0.7754x + 0.1993, R² = 0.7843.
- Real GDP per capita:
  - Positive but non-linear association: y = 0.0019x + 0.5323, R² = 0.2003.
- Poverty rate:
  - Negative association: y = -0.7372x + 0.6614, R² = 0.4271.
- Income inequality (Gini):
  - Weak/near-zero correlation: y = -0.001x + 0.6762, R² = 0.0046.
- Implication:
  - QGI complements HDI, GDP per capita, poverty and inequality measures as a gauge of inclusive growth.

### D. Drivers of the QGI: pairwise and conditional evidence
- Pairwise associations:
  - Institutional quality measures (quality of bureaucracy, rule of law, control of corruption): positive with QGI; strongest for bureaucracy.
  - Government stability: positive association with QGI.
  - Volatility of inflation: negative correlation with QGI.
  - Credit to the private sector: positive correlation with QGI.
  - Public spending in education and health: positive link with QGI.
  - FDI and remittances: weak association in pairwise plots.
  - Foreign aid: markedly negative correlation (interpreted with caveat that aid is mostly allocated to low-income countries).
- Conditional (OLS with lagged explanatory variables) — selected exact coefficients from Table 3 (1990–2011):
  - Social spending-to-GDP ratio:
    - Column [1]: 1.340*** (0.341)
    - Column [2]: 1.292*** (0.339)
    - Column [3]: 1.318*** (0.348)
    - Column [4]: 1.363*** (0.464)
    - Column [5]: 1.451*** (0.333)
    - Column [6]: 1.296*** (0.338)
    - Column [7]: 1.573*** (0.336)
  - Government stability:
    - Column [1]: 0.017*** (0.005)
    - Column [2]: 0.014*** (0.005)
    - Column [3]: 0.014*** (0.005)
    - Column [4]: 0.007 (0.006)
    - Column [5]: 0.015*** (0.005)
    - Column [6]: 0.017*** (0.005)
    - Column [7]: 0.017*** (0.005)
  - Log of inflation volatility:
    - Column [1]: -0.017*** (0.005)
    - Column [2]: -0.016*** (0.005)
    - Column [3]: -0.017*** (0.005)
    - Column [4]: -0.020*** (0.006)
    - Column [5]: -0.018*** (0.005)
    - Column [6]: -0.016*** (0.005)
    - Column [7]: -0.018*** (0.005)
  - Credit to private sector-to-GDP ratio:
    - Column [1]: 0.106*** (0.027)
    - Column [2]: 0.108*** (0.026)
    - Column [3]: 0.107*** (0.028)
    - Column [4]: 0.061 (0.043)
    - Column [5]: 0.114*** (0.026)
    - Column [6]: 0.092*** (0.027)
    - Column [7]: 0.118*** (0.026)
  - Aggregated institutional quality (columns [1]–[4]):
    - Column [1]: 0.029** (0.014)
    - Column [2]: 0.029** (0.014)
    - Column [3]: 0.029** (0.014)
    - Column [4]: 0.036** (0.014)
  - FDI-to-GDP ratio (column [2]):
    - 0.628** (0.260)
  - Remittances-to-GDP ratio (column [3]):
    - 0.196 (0.160)
  - Foreign Aid-to-GDP ratio (column [3]):
    - -1.184*** (0.209)
  - Rule of law (column [4]):
    - 0.013 (0.008)
  - Quality of bureaucracy (column [4]):
    - 0.038*** (0.012)
  - Control of corruption (column [4]):
    - -0.005 (0.011)
- Sample sizes and fit (Table 3 columns):
  - Observations: Column [1]: 163; [2]: 163; [3]: 156; [4]: 101; [5]: 163; [6]: 163; [7]: 163
  - R-squared: Column [1]: 0.423; [2]: 0.438; [3]: 0.427; [4]: 0.638; [5]: 0.417; [6]: 0.447; [7]: 0.408
- Methodological caveat:
  - Estimates are OLS with lagged explanatory variables to mitigate simultaneity; authors note further research with longer time coverage required for robust causal inference.

### Appendix and robustness findings (exact statistics and tests)
- Descriptive statistics for QGI variants (Obs = 372):
  - QGI1 — Mean: 0.604; Std.: 0.140; Min: 0.258; Max: 0.849
  - QGI2 — Mean: 0.550; Std.: 0.137; Min: 0.201; Max: 0.797
  - QGI3 — Mean: 0.591; Std.: 0.116; Min: 0.299; Max: 0.836
  - QGI4 — Mean: 0.616; Std.: 0.167; Min: 0.217; Max: 0.868
  - QGI5 — Mean: 0.584; Std.: 0.105; Min: 0.319; Max: 0.829
- Correlation among alternative QGIs (Appendix 5; p-values in brackets):
  - QGI1 vs QGI2: 0.9868*** (0.000)
  - QGI1 vs QGI3: 0.9937*** (0.000)
  - QGI1 vs QGI4: 0.9635*** (0.000)
  - QGI1 vs QGI5: 0.9845*** (0.000)
  - QGI2 vs QGI3: 0.9624*** (0.000)
  - QGI2 vs QGI4: 0.9941*** (0.000)
  - QGI2 vs QGI5: 0.9452*** (0.000)
  - QGI3 vs QGI4: 0.9274*** (0.000)
  - QGI3 vs QGI5: 0.9965*** (0.000)
  - QGI4 vs QGI5: 0.9051*** (0.000)
- Spearman’s rank-order correlation test (Arithmetic vs Geometric mean QGI):
  - Number of Observations: 372
  - Spearman's rho: 0.995
  - Prob > |t| = 0.0000
- Region-specific benchmark comparisons (Spearman’s rho and sample sizes):
  - QGIAPC1 vs QGIAPC2 — rho: 0.988; Observations: 48; P-value: 0.000
  - QGICEEC1 vs QGICEEC2 — rho: 0.98; Observations: 64; P-value: 0.000
  - QGILAC1 vs QGILAC2 — rho: 0.98; Observations: 68; P-value: 0.000
  - QGIMENA1 vs QGIMENA2 — rho: 0.961; Observations: 40; P-value: 0.000
  - QGISSA1 vs QGISSA2 — rho: 0.999; Observations: 144; P-value: 0.000
- Appendix 3 (geometric mean–based QGI by subperiods) — selected entries preserved:
  - 1990–94 top: Malaysia 0.804; China 0.757; Thailand 0.748; bottom: Mali 0.153; Ethiopia 0.002.
  - 1995–99 top: Malaysia 0.802; China 0.781; Poland 0.780; bottom: Chad 0.168; Niger 0.195.
  - 2000–04 top: China 0.843; Latvia 0.790; Vietnam 0.785; bottom: Central African Rep. 0.278; Niger 0.289.
  - 2005–11 top: China 0.836; Bulgaria 0.826; Argentina 0.823; bottom: Chad 0.220; Burkina Faso 0.393.

### Stylized facts, conclusions, and policy relevance
- Purpose and construction recap:
  - QGI captures intrinsic nature (growth fundamentals) and social dimension (health and education) of growth; composite index ranges between 0 and 1.
  - Growth fundamentals: strength, stability, diversification, outward orientation.
  - Social outcomes: life expectancy at birth, infant survival rate at birth, primary school completion rate.
- Main stylized facts:
  - QGI correlates with UN HDI, poverty rates, income inequality, and income per capita.
  - QGI improved over time overall but varies markedly across regions and income levels; LIC and SSA countries lag behind.
  - Fragility and resource endowment associated with lower QGI.
  - Evidence of convergence in QGI over time, though at a slow pace.
- Drivers of quality of growth:
  - Positive associations: political stability, quality of bureaucracy, aggregated institutional quality; public social spending; price stability (lower inflation volatility); financial development and inclusion (credit to private sector); FDI.
  - Negative association (in one specification and subsamples): foreign aid-to-GDP ratio (e.g., -1.184*** (0.209) in column [3]).
- Policy relevance:
  - QGI can serve as a benchmarking tool to guide policies for inclusive growth.
  - Potential extensions: include labor market and inequality measures as data become available to improve the inclusiveness dimension.

*Source: _wp14172 - 1. Full Sample QGI-based Ranking (PDF chapter) based on the provided content unit.*

### 1. Full Sample QGI-based Ranking .......................................................................................

### 1. Full Sample QGI-based Ranking

### I. Introduction: purpose and conceptualization
- Objective: introduce a Quality of Growth Index (QGI) that encompasses both the intrinsic nature of growth and its social dimensions to assess the “quality of growth” across developing and emerging countries over 1990–2011.
- Conceptual definition: “good quality growth” = high, durable, and socially-friendly growth; emphasizes that how growth is generated (path from level L1 to L2 of income) matters for social outcomes.
- Positioning vs. existing indices:
  - QGI goes beyond the Human Development Index (HDI) by concentrating on the nature of growth rather than just income levels accumulated over long periods.
  - QGI differs from the Social Progress Index by incorporating growth fundamentals (pace, stability, sectoral breadth, outward-orientation) alongside social outcomes.
- Key research questions:
  - How has the quality of growth evolved over time and across regions?
  - Is there convergence in QGI dynamics or a “growth quality trap”?
  - How is QGI related to other development indicators?
- Main findings (as stated in the source):
  - The quality of growth has been improving in the majority of countries over the past two decades.
  - The rate of convergence is relatively slow.
  - There are considerable cross-country variations across income levels and regions.
  - Political stability, public pro-poor spending, macroeconomic stability, financial development, institutional quality and external factors such as FDI are associated with higher QGI.

### II. Methodology and index structure
- QGI composition:
  - QGI is a composite index aggregating two building blocks: the intrinsic nature of growth sub-index (“growth fundamentals”) and the social dimension sub-index (“social outcomes”).
  - Formal expression: QGI = α·(growth fundamentals) + β·(social dimension)
- Growth fundamentals sub-index: four dimensions
  - Strength (γ1): measured by annual change in real GDP per capita (GDP per capita used to align with pro-poor concept).
  - Stability (γ2): measured by the inverse of the coefficient of variation (CV) of growth using a five-year rolling window (CV = standard deviation / average).
  - Diversification of sources (γ3): proxied by 1 − Herfindahl-Hirschman index (HHI) using exports data (export diversification used due to missing value issues in sectoral value added).
  - Outward-orientation (γ4): proxied by share of net external demand (exports − imports), both as percent of GDP.
- Social dimension sub-index: two components
  - Health (δ1): aggregation of the reverse of infant mortality rate and life expectancy at birth.
  - Education (δ2): primary school completion rate (chosen due to data availability; missing observations are addressed per Appendix 2).
- Notes on interpretation:
  - A higher CV implies lower inverse-CV and lower stability; instability is harmful to poverty and equity via hysteresis.
  - Outward orientation can raise productivity but may increase vulnerability to external shocks; volatility accounted for separately.

### III. Standardization, weighting, and aggregation
- Standardization approaches discussed:
  - Z-score (centered-reduced normalization): Z = (X − μ) / σ; sensitive to outliers and unbounded.
  - Min-Max: Z' = (X − Xmin) / (Xmax − Xmin); bounded between 0 and 1; this paper uses Xmin and Xmax observed in the sample.
- Weighting choices (benchmark equal-weighting):
  - α = 50 percent and β = 50 percent (equal weight between growth fundamentals and social dimension).
  - Within growth fundamentals: γ1 = γ2 = γ3 = γ4 = 25 percent.
  - Within social dimension: δ1 = δ2 = 50 percent; health sub-index sub-components also equally weighted at 50 percent each.
  - Rationale: simplicity and transparency; noted arbitrariness of equal weights.
  - Robustness: alternative weights and aggregation methods explored in robustness section.
- Aggregation:
  - Benchmark aggregation: arithmetic mean (simple averaging) for the two main sub-indices and their sub-components.
  - Alternative (robustness): geometric averaging to account for potential complementarities among components.
  - Formal arithmetic formulation given: QGI = Social^(β) + Fundamentals^(α) (text presents additive notation and geometric alternative).
- Caveats:
  - Equal weighting implies substitutability among components; geometric averaging addresses possible complementarities (e.g., human capital ↔ productivity).

### IV. Data and sample
- Panel coverage: 93 developing countries over 1990–2011.
  - Breakdown: 57 middle-income countries and 36 low-income countries.
- Temporal aggregation: five-year averages to smooth short-term fluctuations:
  - Periods: 1990–94, 1995–99, 2000–04, and 2005–11.
- Data sources: IMF World Economic Outlook, World Bank WDI, COMTRADE, International Country Risk Guide, Barro and Lee (2010), Xala-i-Martin (2006). Detailed sources and definitions in Appendix 1.
- Missing data handling:
  - Primary school completion rate has missing observations; assumptions and imputation approach described in Appendix 2.
  - Export diversification proxied using exports data due to missing sectoral value added data.

### V. Robustness and sensitivity
- Sensitivity tests performed:
  - Alternative weights for α, β, γi, δi.
  - Alternative aggregation: geometric mean vs. arithmetic mean.
  - Alternative standardization choices discussed; benchmark uses sample observed min/max for Min-Max scaling.
- Justification for chosen approaches:
  - Min-Max with sample observed extrema reduces subjectivity compared with hypothetical ideal extrema.
  - Arithmetic mean preferred for simplicity and comparability with other indices (HDI, EVI), but geometric mean used to check sensitivity to assumed substitutability/complementarity.

### VI. Key empirical takeaways (explicit figures and specifications)
- Sample size: 93 developing countries.
- Income group composition: 57 middle-income countries; 36 low-income countries.
- Time coverage: 1990–2011, averaged over five-year windows (1990–94, 1995–99, 2000–04, 2005–11).
- Weighting scheme (benchmark):
  - α = 50 percent; β = 50 percent.
  - γ1 = γ2 = γ3 = γ4 = 25 percent.
  - δ1 = δ2 = 50 percent.
  - Health sub-index: two sub-components weighted at 50 percent each.
- Standardization: Min-Max using sample observed Xmin and Xmax.
- Aggregation: arithmetic mean (benchmark); geometric mean used for robustness.

*Source: _wp14172 - 1. Full Sample QGI-based Ranking (PDF chapter) based on the provided content unit.*

### 1. Appendix 2 elaborates on the specific case of dealing with missing observations in the

### _wp14172 - 1. Appendix 2 elaborates on the specific case of dealing with missing observations in the

### III. RESULTS — Overview
- The section presents: stylized facts of the QGI; ranking and categorization of countries by QGI; assessment of convergence in the QGI; comparison of the QGI with existing development and living-standard indicators; and exploration of potential drivers of the QGI.

### A. Some stylized facts
- Ranking highlights (period 2005–2011):
  - Top performers: Bulgaria 0.843, China 0.842, Argentina 0.830.
  - Poorest performers: Chad 0.334, Central African Republic 0.402, Niger 0.415.
- Aggregate QGI statistics and distributional shifts:
  - Average value of the QGI stands at 0.604.
  - Minimum QGI: 0.258 for Niger over 1990–94.
  - Maximum QGI: 0.849 for China over 2000–04.
  - QGI increases from 0.556 in 1990–94 to 0.656 in 2005–2011.
  - Density plots indicate the QGI distribution shifts rightward over time and becomes narrower, indicating some convergence.
- Regional and income-level patterns:
  - Regional: LA countries exhibit the highest QGI scores; sub-Saharan Africa (SSA) exhibits the lowest QGI and the flattest/leftmost density with thick tails, implying greater inequality in QGI scores and that roughly more than 60 percent of SSA observations lie below the full sample average score (0.604).
  - Income-level ordering: The QGI is positively correlated with countries’ income level. Upper-middle income countries record the highest QGI, followed by lower-middle income countries and low-income countries.
- Fragility and resource endowment:
  - Fragile countries underperformed the sample average by almost 16 percentage point.
  - Resource-rich countries have QGI scores slightly lower than non-resource rich peers; density comparisons suggest natural resource endowment tended to be more of a curse than a blessing for achieving higher QGI.

### B. Convergence hypothesis
- Pooled OLS evidence (Table 2):
  - Lagged QGI (one period) coefficient: -0.066*** (standard error reported as (0.0156) in source).
  - Initial QGI (1990-94) coefficient: -0.068*** (standard error reported as (0.016) in source).
  - Observations: 279.
  - R-squared: 0.072 (column 1) and 0.074 (column 2).
  - Note: Robust standard errors in brackets; *, **, and *** indicate significance at 10 percent, 5 percent, and 1 percent.
- Interpretation:
  - Past QGI performance is negatively associated with subsequent QGI growth rates, suggesting convergence: lower-performing countries tend to catch up to better performers over time.
- Country categorization (1990–94 vs 2005–2011):
  - Region 2 (“hopefuls”): QGI below sample average in both periods but improved — largely low-income and/or fragile countries, mostly from SSA and some from MENA.
  - Region 3 (“contenders”): Improved from below to above the sample average — includes countries from Asia Pacific, SSA, MENA, CEE, and some from LA (examples: Bangladesh, Laos, Nepal, Ghana, Tanzania, Zambia, Algeria, Iran, Morocco, Azerbaijan, Tajikistan, Uzbekistan, Guatemala, Nicaragua).
  - Region 4 (“club of best performers”): Above sample average in both initial and final periods and improved — chiefly upper-middle and lower-middle income countries; includes a handful of SSA countries (Kenya, Namibia, South Africa).
  - Region 5 (“club of superior performers”): Above sample average in both periods but experienced a mild drop — comprises Botswana and Malaysia.
  - Virtually all countries improved QGI over the two decades.

### C. Putting the QGI into perspective with existing development indicators
- Pairwise correlations (Figure 5 summary):
  - QGI is positively correlated with the UN HDI (strong positive relationship reported: y = 0.7754x + 0.1993, R² = 0.7843).
  - QGI is positively associated with real GDP per capita (non-linear; reported fit: y = 0.0019x + 0.5323, R² = 0.2003).
  - QGI is negatively correlated with the poverty rate (reported fit: y = -0.7372x + 0.6614, R² = 0.4271).
  - QGI shows weak/near-zero correlation with income inequality measured by Gini (reported fit: y = -0.001x + 0.6762, R² = 0.0046).
- Implication:
  - The QGI can serve as a complementary indicator for gauging countries’ progress toward inclusive growth alongside HDI, GDP per capita, poverty, and inequality measures.

### D. Drivers of the QGI: an appraisal
- Pairwise correlation findings (Figures 6.a–6.d):
  - Politico-institutional factors:
    - Institutional quality measures (quality of bureaucracy, rule of law, control of corruption) are positively associated with QGI; the association is strongest for bureaucracy.
    - Greater government stability correlates with higher QGI.
  - Domestic macroeconomic environment:
    - Volatility of inflation is negatively correlated with QGI.
    - Credit to the private sector is positively correlated with QGI.
  - Social spending:
    - Public spending in education and health sectors is positively linked with QGI.
  - External environment:
    - Weak association between QGI and FDI and remittances in pairwise plots.
    - Correlation between QGI and foreign aid is markedly negative (authors note aid is mostly allocated to low-income countries).
- Econometric (conditional) analysis — OLS with lagged explanatory variables (summary of Table 3 narrative):
  - Robust conditional results reported in column 1 (baseline domestic macroeconomic factors):
    - More public resources to social sectors improve QGI.
    - Stable government (government stability) is conducive to better QGI.
    - Volatility of inflation coefficient is negative and significant — macroeconomic stability is important for pro-poor growth outcomes.
    - Financial development has a positive impact on QGI — greater financial development, especially with higher access to credit, helps achieve better quality growth.
    - Institutional quality (average of quality of bureaucracy, rule of law, control of corruption) matters; quality of bureaucracy appears most relevant.
  - External-condition columns (2 to 4) findings:
    - FDI: positive relationship with QGI.
    - Remittances: positive but not significant.
    - Foreign aid: significantly negative effect (authors note interpretation caveat: may reflect allocation of aid to poorer countries rather than aid ineffectiveness).
- Methodological caveat and next steps:
  - The analysis is based on partial correlations and OLS estimates with lagged explanatory variables to mitigate simultaneity and endogeneity concerns.
  - Authors note that further research with longer time coverage of the QGI is warranted before drawing robust causal inferences.

*Data source: Authors' calculations (as reported in the supplied content).*

### Appendix 9 depicts signs and significance levels of the QGI’s drivers for different sub-sampling.

### _wp14172 - Appendix 9 depicts signs and significance levels of the QGI’s drivers for different sub-sampling

### Determinants of Quality of Growth (Table 3: 1990–2011)
- Dependent variable: QGI (columns [1] through [7])
- Key estimated coefficients (one period lag) with robust standard errors in parentheses; *, **, and *** indicate significance at 10%, 5%, and 1%:
  - Social spending-to-GDP ratio:
    - Column [1]: 1.340*** (0.341)
    - Column [2]: 1.292*** (0.339)
    - Column [3]: 1.318*** (0.348)
    - Column [4]: 1.363*** (0.464)
    - Column [5]: 1.451*** (0.333)
    - Column [6]: 1.296*** (0.338)
    - Column [7]: 1.573*** (0.336)
  - Government stability:
    - Column [1]: 0.017*** (0.005)
    - Column [2]: 0.014*** (0.005)
    - Column [3]: 0.014*** (0.005)
    - Column [4]: 0.007 (0.006)
    - Column [5]: 0.015*** (0.005)
    - Column [6]: 0.017*** (0.005)
    - Column [7]: 0.017*** (0.005)
  - Log of inflation volatility:
    - Column [1]: -0.017*** (0.005)
    - Column [2]: -0.016*** (0.005)
    - Column [3]: -0.017*** (0.005)
    - Column [4]: -0.020*** (0.006)
    - Column [5]: -0.018*** (0.005)
    - Column [6]: -0.016*** (0.005)
    - Column [7]: -0.018*** (0.005)
  - Credit to private sector-to-GDP ratio:
    - Column [1]: 0.106*** (0.027)
    - Column [2]: 0.108*** (0.026)
    - Column [3]: 0.107*** (0.028)
    - Column [4]: 0.061 (0.043)
    - Column [5]: 0.114*** (0.026)
    - Column [6]: 0.092*** (0.027)
    - Column [7]: 0.118*** (0.026)
  - Aggregated institutional quality (included in columns [1]–[4]):
    - Column [1]: 0.029** (0.014)
    - Column [2]: 0.029** (0.014)
    - Column [3]: 0.029** (0.014)
    - Column [4]: 0.036** (0.014)
  - FDI-to-GDP ratio (column [2] only):
    - 0.628** (0.260)
  - Remittances-to-GDP ratio (column [3] only):
    - 0.196 (0.160)
  - Foreign Aid-to-GDP ratio (column [3] only):
    - -1.184*** (0.209)
  - Rule of law (column [4] only):
    - 0.013 (0.008)
  - Quality of bureaucracy (column [4] only):
    - 0.038*** (0.012)
  - Control of corruption (column [4] only):
    - -0.005 (0.011)
- Sample sizes and fit:
  - Observations: Column [1]: 163; [2]: 163; [3]: 156; [4]: 101; [5]: 163; [6]: 163; [7]: 163
  - R-squared: Column [1]: 0.423; [2]: 0.438; [3]: 0.427; [4]: 0.638; [5]: 0.417; [6]: 0.447; [7]: 0.408
- Notes:
  - Robust standard errors reported in brackets.
  - Intercept included but not reported.

### Robustness of the QGI to alternative specifications (Section IV)
- Alternative index weighting and aggregation:
  - Four alternative QGIs considered: QGI2, QGI3, QGI4, QGI5 (different α and β weightings between intrinsic nature and social dimension).
  - QGI4: α = 3/4 and β = 1/4 for intrinsic nature and social dimension, respectively; QGI5 is the converse.
- Key robustness findings:
  - Correlation among five QGIs (Appendix 5) ranges from 0.91 to 0.99, indicating high and significant correlation across weighting schemes.
  - Geometric averaging (instead of arithmetic) does not substantially alter rankings (Appendix 3).
    - Spearman’s rank-order correlation test between arithmetic mean–based and geometric mean–based country rankings: p-value = zero; test statistic = 0.995.
  - Region-specific Min-Max rebasing for normalization produces rankings significantly comparable to the benchmark (Appendix 7 and Appendix 8).
- On stock vs. flow variables for the social sub-component:
  - Authors explain they refrained from using flow variables for the social dimension because:
    - Dynamics of social variables show marked convergence (Figure 7).
    - Flow-based social changes could overwhelm growth fundamentals and bias QGI scores and rankings.
  - Figure 7 illustrates convergence: fitted log relationships reported in text:
    - y =  -19.22ln(x) - 1.9901
    - y =  -7.127ln(x) +  1.9857

### Stylized facts and conclusions (Section V)
- Purpose and construction of QGI:
  - QGI captures intrinsic nature (growth fundamentals) and social dimension (health and education) of growth; composite index ranges between 0 and 1.
  - Growth fundamentals sub-index covers: strength of growth, stability of growth, diversification of sources, and outward orientation.
  - Social outcomes sub-index uses life expectancy at birth, infant survival rate at birth, and primary school completion rate.
- Main stylized facts:
  - QGI correlates well with other welfare measures such as UN HDI, poverty rates, income inequality, and income per capita.
  - QGI has improved over time overall but varies markedly across regions and income levels; LIC and sub-Saharan African countries lag behind.
  - Structural factors such as fragility and resource endowment are associated with lower QGI.
  - Evidence of convergence in quality of growth over time, though at a slow pace; possibility of a “poor growth quality trap” explored.
- Drivers of quality of growth (empirical findings):
  - Institutions and policies associated with higher QGI: political stability, quality of bureaucracy, aggregated institutional quality.
  - Sound macroeconomic policies contributing to better QGI: share of social spending, price stability, financial development and inclusion (credit to private sector), and FDI.
  - Foreign aid-to-GDP ratio estimated coefficient in one specification: -1.184*** (0.209) (see Determinants table).
- Policy relevance:
  - QGI can serve as a benchmarking tool to guide policies for inclusive growth.
  - Potential future extensions: include labor market and inequality measures as data become available to improve the inclusiveness dimension.

### Data sources, definitions, and handling missing data (Appendices 1–2)
- Selected variable definitions and sources (excerpt):
  - Quality of Growth Index (QGI): composite index between 0 and 1; authors’ own calculations.
  - Poverty rate: percent living with less than one dollar a day (Sala-i-Martin (2006)).
  - HDI: UN HDI database.
  - Inflation: CPI-based (World Economic Outlook).
  - GDP per capita growth rate: WDI.
  - Life expectancy at birth; Infant mortality rate: standard definitions (WDI).
  - Credit to private sector: percent of GDP.
  - FDI, Remittances, Primary school completion rate: WDI and authors’ estimations for missing observations.
  - Index of diversification of export products: complement of HHI (COMTRADE).
  - Foreign Official Aid: percent of GDP (Guillaumont and Tapsoba (2012)).
  - Public health and education spending: percent of GDP (IMF dataset).
  - Quality of bureaucracy: ICRG (ranging 0–4).
  - Rule of law and Control of corruption: indices ranging 0–6.
  - Government stability: index ranging 0–12.
- Method to impute missing primary school completion rates (Appendix 2):
  - Step 1: Compute Complest = average years of primary schooling (Barro and Lee, 2010) divided by average duration of primary school (WDI, 2010).
  - Step 2: If actual completion rate (Complact) exists for country-year, use it; if missing, use Complest adjusted by the average deviation between Complest and Complact in a neighborhood spanning 5 years before to 5 years after the missing year:
    - ititit estComplComplλ_ + = (notation from source)
    - Neighborhood average: 4,4),__(+−=−==ttjestComplactComplmean iitj (text as in source).
  - Note: “5” corresponds to the interval over which data are averaged in the calculation of the QGI.

### Country ranking (Appendix 3: Geometric mean–based QGI by subperiods)
- Selected top and bottom entries across subperiod rankings (values preserved as in source):
  - 1990–94 top: Malaysia 0.804; China 0.757; Thailand 0.748; …; bottom: Mali 0.153; Ethiopia 0.002.
  - 1995–99 top: Malaysia 0.802; China 0.781; Poland 0.780; …; bottom: Chad 0.168; Niger 0.195.
  - 2000–04 top: China 0.843; Latvia 0.790; Vietnam 0.785; …; bottom: Central African Rep. 0.278; Niger 0.289.
  - 2005–11 top: China 0.836; Bulgaria 0.826; Argentina 0.823; …; bottom: Chad 0.220; Burkina Faso 0.393.
- Full country-by-country rankings and values presented in Appendix 3 (tabulated per subperiod).

*Source: _wp14172 - Appendix 9 depicts signs and significance levels of the QGI’s drivers for different sub-sampling (IMF working paper content provided).*

### Appendix 4. QGI for the Full Sample over 1990–2011: Descriptive Statistics

### Appendix 4. QGI for the Full Sample over 1990–2011: Descriptive Statistics

### Descriptive statistics for QGI variants
- Variables and sample size (Obs = 372 for all entries)
  - QGI1 — Mean: 0.604; Std.: 0.140; Min: 0.258; Max: 0.849
  - QGI2 — Mean: 0.550; Std.: 0.137; Min: 0.201; Max: 0.797
  - QGI3 — Mean: 0.591; Std.: 0.116; Min: 0.299; Max: 0.836
  - QGI4 — Mean: 0.616; Std.: 0.167; Min: 0.217; Max: 0.868
  - QGI5 — Mean: 0.584; Std.: 0.105; Min: 0.319; Max: 0.829
- Note on group labels and QGI definitions:
  - APC = Asian and Pacific Countries; CEEC = Central and Eastern European Countries; LAC = Latin American Countries; MENA = Middle East and North Africa; SSA = Sub-Saharan Africa; LIC = Low-income countries; LMIC = Lower-Middle income countries; UMIC = Upper-Middle income countries; OIL = Oil exporting countries; NON-OIL = Non-oil exporting countries.
  - QGI1 corresponds to the QGI obtained using 2/12 = 1/6 for αα (as presented: 2/121).
  - QGI2 corresponds to (3/12, 3/21 =?) per source notation (presented as (3/12,3/21)).
  - QGI3 corresponds to (3/22, 3/11) per source notation (presented as (3/22,3/11)).
  - QGI4 corresponds to (4/12, 4/31) per source notation (presented as (4/12,4/31)).
  - QGI5 corresponds to (4/32, 4/11) per source notation (presented as (4/32,4/11)).

### Appendix 5 — Correlation matrix of alternative QGIs
- Pairwise correlations (significance and p-values in brackets where provided)
  - QGI1 vs QGI2: 0.9868*** (0.000)
  - QGI1 vs QGI3: 0.9937*** (0.000)
  - QGI1 vs QGI4: 0.9635*** (0.000)
  - QGI1 vs QGI5: 0.9845*** (0.000)
  - QGI2 vs QGI3: 0.9624*** (0.000)
  - QGI2 vs QGI4: 0.9941*** (0.000)
  - QGI2 vs QGI5: 0.9452*** (0.000)
  - QGI3 vs QGI4: 0.9274*** (0.000)
  - QGI3 vs QGI5: 0.9965*** (0.000)
  - QGI4 vs QGI5: 0.9051*** (0.000)
- Note: *** indicates the significance level of 1 percent; P-value in brackets.

### Appendix 6 — Spearman’s rank order correlation test (Arithmetic Mean-based versus Geometric Mean-based QGI)
- Number of Observations: 372
- Spearman's rho: 0.995
- Ho: Arithmetic Mean-based QGI and Geometric mean-based QGI are independent
- Prob > |t| = 0.0000

### Appendix 7 — Spearman’s rank order correlation test by region (benchmark vs region-specific)
- Spearman's rho and sample sizes
  - QGIAPC1 vs QGIAPC2 — rho: 0.988; Number of Observations: 48; P-value: 0.000
  - QGICEEC1 vs QGICEEC2 — rho: 0.98; Number of Observations: 64; P-value: 0.000
  - QGILAC1 vs QGILAC2 — rho: 0.98; Number of Observations: 68; P-value: 0.000
  - QGIMENA1 vs QGIMENA2 — rho: 0.961; Number of Observations: 40; P-value: 0.000
  - QGISSA1 vs QGISSA2 — rho: 0.999; Number of Observations: 144; P-value: 0.000
- Note: subscript 1 refers to the benchmark QGI score by region while subscript 2 refers to the region-specific one.
- Hypothesis tested: Ho: The two sets of QGI indices are independent.

### Appendix 8 — Correlation matrix between benchmark QGI and region-specific QGI
- Pairwise correlations (P-value in brackets)
  - QGIAPC1 vs QGIAPC2: 0.993*** (0.000)
  - QGICEEC1 vs QGICEEC2: 0.984*** (0.000)
  - QGILAC1 vs QGILAC2: 0.982*** (0.000)
  - QGIMENA1 vs QGIMENA2: 0.964*** (0.000)
  - QGISSA1 vs QGISSA2: 0.999*** (0.000)
- Note: subscript 1 refers to the benchmark QGI score by region while subscript 2 refers to the region-specific one; P-value in brackets.

### Appendix 9 — Determinants of Quality of Growth on different subsamples (summary)
- Dependent variable: QGI
- Subsamples (columns): [1] LIC, [2] MIC, [3] 1990-99, [4] 2000-11, [5] SSA, [6] Non-Fragile, [7] Resource-rich
- Reported sign and significance for explanatory variables (notation: (+) positive effect, (-) negative effect; significance: * = 10%, ** = 5%, *** = 1%)
  - Social spending-to-GDP ratio: positive in most subsamples; significance varies (e.g., (+)*** in [3], (+)*** in [5], (+)* in [1] and [7])
  - Government stability: generally positive; significance includes (***) in several subsamples and a negative sign in one column (reported as (-) in [4])
  - Log of inflation volatility: negative across subsamples with significance at ** or *** (e.g., (-)** in [1], (-)*** in [3])
  - Financial depth: positive and significant across subsamples (noted as (+)*** in multiple columns)
  - Aggregated institutional quality: positive with varying significance (e.g., (+)* in [1], (+)*** in [3], (+)** in [6])
  - FDI-to-GDP ratio: generally positive; significance appears in some columns (e.g., (+)* in [3], (+)* in [4])
  - Remittances-to-GDP ratio: mixed signs across subsamples; significance varies (e.g., (+)** in [1], negative in [2] and [3])
  - Foreign Aid-to-GDP ratio: negative and significant in several subsamples ((-)*** reported)
  - Rule of law: generally positive; significance present in some columns (e.g., (+)** in [2])
  - Quality of bureaucracy: positive and often significant ((+)*** in multiple columns)
  - Control of corruption: mixed signs and significance across subsamples (e.g., (-) in [1], (-)* in [2], (+) in some others)
- Observations and R-squared (selected reported values in table)
  - Observations: column [1] = 71; [2] = 158; [3] = 96; [4] = 133; [5] = 78; [6] = 159; [7] = 52
  - R-squared: column [1] = 0.558; [2] = 0.218; [3] = 0.509; [4] = 0.370; [5] = 0.541; [6] = 0.432; [7] = 0.566
- Note: (+) and (-) indicate a positive and negative effect respectively, while *, **, and *** indicate the significance level of 10%, 5%, and 1% respectively.

*Source: Appendix material from the IMF working paper chapter provided.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14172.pdf_