## 2. Robustness Tests: Additional Estimation Results

## Source details

**Canonical URL:** [2. Robustness Tests: Additional Estimation Results](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1341.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1341.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1341.pdf.json)

---

### I. Introduction
- ASEAN-5 refers to Indonesia, Malaysia, Philippines, Thailand, and Vietnam.
- Stylized facts:
  - Infrastructure gaps: relatively bigger in Indonesia, the Philippines, and Vietnam than in Malaysia and Thailand.
  - Investment as a share of GDP has never fully recovered in most ASEAN-5 countries since the Asian crisis.
- Paper objective: assess empirically the impact of infrastructure (quality and quantity) and investment on income distribution, while analyzing the role of public policies in education, labor, and financial access.

### II. Infrastructure development in ASEAN-5 — stylized facts
- Country-specific observations:
  - Indonesia
    - GCI ranking on infrastructure improved over the past five years, but infrastructure investment lags regional peers.
    - Public investment remains low; constraints include project selection at local government levels and inefficiencies in procurement, budget preparation, and payment processes.
  - Malaysia
    - Overall better infrastructure; high rankings in GCI and IMD World Competitiveness Yearbook.
    - Comparative advantage in many basic infrastructure categories including energy.
    - Stagnated investment rate after the Asian crisis; Economic Transformation Program focuses on infrastructure, commodity, and construction sectors.
    - Room for improvement in ICT infrastructure, especially in less populous regions.
  - The Philippines
    - Rankings remain relatively low among ASEAN-5; transportation and electricity quality need improvement.
    - Governance, financial, and labor market limitations identified as main constraints.
  - Thailand
    - National-level infrastructure improved; significant regional disparities exist with concentration near Bangkok.
    - After late 2011 floods, government rolled out flood-prevention investment including a B 350 billion (3 percent of GDP) off-budget package to be implemented over three years.
  - Vietnam
    - Improvements in basic indicators: length of paved roads quadrupled over ten years; electricity production capacity increased by about 10 percent in two decades.
    - Challenges remain in quality of roads and ports and in public resource allocation efficiency.
- Post-Asian crisis dynamics:
  - Gross fixed investment declined significantly in most ASEAN-5 economies.
  - Decline driven by a sharp fall in private investment, including foreign direct investment.
  - Private investment commitments to infrastructure sectors dropped since the Asian crisis (measured in US dollars and as a share of GDP).

### III. Methodology and results
- Data and sample
  - Sample: 76 advanced and emerging market economies.
  - Period: 1980-2010.
  - Empirical setup: non-overlapping 5-year averages; pooled OLS with heteroscedasticity-consistent standard errors; variables de-meaned using country-specific means (equivalent to panel with fixed country effects); time dummies included.
- Construction of infrastructure indices (following Calderon and Serven)
  - Quantity index: first principal component of:
    - total telephone lines, cell subscriptions, and internet users per 100 people (communication),
    - electricity production per capita (in millions of KWh),
    - road density (km of road per 100 sq. km of land area).
  - Quantity index eigenvectors: 0.6377*Communication + 0.6336*Electricity + 0.4381*Roaddensity.
  - Quality index: first principal component of:
    - roads paved in percent of total roads,
    - electric power that is not lost in transmission and distribution in percent of total output.
  - Quality index eigenvectors: 0.7071*Pavedroads + 0.7071*Powerdistribution.
  - Note: waiting time for installation of main telephone lines excluded due to lack of data.
- Endogeneity treatment
  - Infrastructure indices enter regressions with one lag to address two-way causality.
- Main empirical findings (selected)
  - Infrastructure (both quality and quantity) is associated with reduced income inequality; coefficients negative and statistically significant across specifications.
  - Investment measures (total investment-to-GDP ratio, private investment-to-GDP ratio, public investment-to-GDP ratio), each entered with one lag, are negative but not statistically significant.
  - Addition of an index of institutional risks as a control for spending efficiency does not materially change baseline results.
- Estimated potential impacts (as reported)
  - Catching up to the 2010 average levels of infrastructure quantity and quality indices in advanced economies is estimated to reduce the Gini index by about 2 percentage points in Vietnam, the Philippines, and Indonesia (based on Column 3 of Table 1).
  - Estimated decline in the Gini index for Malaysia is 1 percentage point.
- Selected coefficient estimates from Table 1 (dependent variable: natural logarithm of Gini) — coefficients shown exactly as in source
  - constant: 11.48 ***, 10.07 ***, 9.51 ***, -6.12 ***, 13.08 ***, 13.30 ***, 11.89 ***, 14.36 ***
  - Inflation: 0.07 *, 0.05 *, 0.04, 0.10, 0.10 **, 0.09 **, 0.10 ***, 0.09 **
  - Infrastructure quantity (lag 1): -0.20 ***, -0.1767 ***
  - Infrastructure quality (lag 1): -0.22 ***, -0.19 ***
  - Investment-to-GDP ratio (lag 1): -0.14, -0.11
  - Private investment-to-GDP ratio (lag 1): -0.09
  - Public investment-to-GDP ratio (lag 1): -0.03, -0.03
  - Credit-to-GDP ratio: 0.08 ***, 0.09 ***, 0.09 ***, 0.07 **, 0.02, 0.01, 0.06 *, 0.01
  - Employment share in industry (lag 1): -0.19 **, -0.16, -0.11, -0.19 **, -0.18 *, -0.17 *, -0.26 **, -0.22 *
  - PPP real GDP per capita: 2.73 ***, 2.06 ***, 2.29 ***, 2.34 ***, 2.73 ***, 2.77 ***, 2.41 ***, 2.89 ***
  - PPP real GDP per capita squared: -0.01 ***, -0.01 ***, -0.01 ***, -0.13 ***, -0.01 ***, -0.01 ***, -0.01 ***, -0.01 ***
  - Education spending in percent of GDP (lag 1): -0.08, -0.11 **, -0.10 *, -0.08 **, -0.07, -0.05, -0.08, -0.06
  - Openness: 0.05, 0.08 *, 0.13 ***, 0.01, 0.03, 0.03, 0.02, 0.03
  - Institutional risks: -0.04, -0.14
  - Observations: 128, 126, 111, 149, 107, 106, 147, 105
  - Adjusted R-squared: 0.61, 0.60, 0.70, 0.48, 0.49, 0.48, 0.49, 0.49
  - Note: An * indicates significance at 10%, ** significant at 5%, and *** at 1%. All explanatory variables are in natural logarithm.
- Interpretation and robustness
  - Infrastructure quantity and quality indices likely capture different aspects and both matter for income distribution.
  - Investment-to-GDP measures are imperfect proxies for infrastructure development due to potential spending inefficiency and governance/institutional factors.
  - Results broadly consistent with Kuznets hypothesis: PPP real GDP per capita positive, its square negative.
  - Education-related variables (education spending and workforce with secondary education) have negative and statistically significant coefficients, suggesting greater access to education reduces income inequality.
  - Private credit in percent of GDP shows a significant negative effect on income distribution in some specifications, but interpretation depends on institutions and regulation quality.
  - Employment share in industry associated with more equal income distribution.
  - Openness tends to make income distribution more unequal.
- Robustness checks (appendix III summary)
  - Estimation approaches include fixed country effects and least-square estimations using variables that are not de-meaned.
  - Dropping one variable at a time: results remain broadly unchanged.
  - Replacing dependent variable with per capita income share held by the lowest 20 percent: results confirm that an improvement in infrastructure quality and quantity would increase the income share held by the poor.
  - General finding: In all models, infrastructure quantity and quality indices remain statistically significant with negative coefficients.

### IV. Conclusions and policy implications
- Key conclusion: Better infrastructure, both quantity and quality, improves income distribution and can contribute to poverty reduction and inclusive growth.
- Investment alone does not appear to promote income equality in the cross-country regressions; this may reflect a weak link between spending and accumulation of infrastructure assets and deficiencies in spending efficiency and institutions.
- Policy recommendations:
  - Improve institutional frameworks and spending efficiency (e.g., streamline procurement, improve coordination among institutions) to strengthen the productivity of infrastructure investment.
  - Promote education (increase education spending and access) to disproportionately raise earnings of lower-income groups and reduce inequality.
  - Encourage formal sector employment and expansion of industry, particularly in low-income regions, to provide more inclusive employment opportunities.
  - Advance financial development while improving institutions and broadening the poor’s access to finance to avoid disproportionate benefits accruing to the rich.
- Suggested area for future research: analyze how different types of infrastructure (for example, roads vs. broadband connection) affect inequality and growth differently.

### Appendix I: Description of data (selected)
- Gini index: From the World Bank’s World Development Indicators (WDI) database; natural logarithm used in regressions.
- Infrastructure quantity and quality: Authors’ principal-component constructions using WDI underlying series; natural logarithms used in regressions.
- Inflation: Year-on-year growth of consumer price index from IMF’s World Economic Outlook database; natural logarithm of inflation rate+100 used.
- Investment-to-GDP (total, private, public): Gross fixed capital formation series from World Economic Outlook; natural logarithms used.
- Credit-to-GDP: Claims on private sector from IMF IFS and MBRF2; natural logarithm used.
- Employment share in industry: WDI; natural logarithm used.
- PPP real GDP per capita: Penn World database; natural logarithm used.
- Education spending in percent of GDP: WDI; natural logarithm used.
- Openness: exports plus imports in percent of GDP from World Economic Outlook; natural logarithm used.
- Institutional risks: International Country Risk Guide (ICRG) database of the PRS Group; higher score indicates a lower risk environment.

- Country coverage: 24 industrial countries and 52 emerging markets listed in appendix II (source document).

*Source: IMF staff estimates and analysis in "2. Robustness Tests: Additional Estimation Results" (content unit).*

### References .............................................................................................................

### References

### Figures
- 1.   Infrastructure, Investment, and Regulatory Environment ....................................................4
- 2.   Poverty and Income Distribution .........................................................................................5

### Tables
- 1.   Income Inequality Regression Results ...............................................................................12

*Source: _wp1341 - References*

### 2.   Robustness   Tests:   Additional Estimation Results ..............................................................1

### 2. Robustness Tests: Additional Estimation Results

### I. Introduction
- ASEAN-5 refers to Indonesia, Malaysia, Philippines, Thailand, and Vietnam.
- Infrastructure gaps are relatively bigger in Indonesia, the Philippines, and Vietnam than in Malaysia and Thailand.
- Investment as a share of GDP has never fully recovered in most ASEAN-5 countries since the Asian crisis.
- Paper objective: assess empirically the impact of infrastructure (quality and quantity) and investment on income distribution, while analyzing the role of public policies in education, labor, and financial access.

### II. Infrastructure development in ASEAN-5 — stylized facts
- Indonesia
  - GCI ranking on infrastructure improved over the past five years, but infrastructure investment lags regional peers.
  - Public investment remains low; constraints include project selection at local government levels and inefficiencies in procurement, budget preparation, and payment processes.
- Malaysia
  - Overall better infrastructure; high rankings in GCI and IMD World Competitiveness Yearbook.
  - Comparative advantage in many basic infrastructure categories including energy.
  - Stagnated investment rate after the Asian crisis; Economic Transformation Program focuses on infrastructure, commodity, and construction sectors.
  - Room for improvement in ICT infrastructure, especially in less populous regions.
- The Philippines
  - Rankings remain relatively low among ASEAN-5; transportation and electricity quality need improvement.
  - Governance, financial, and labor market limitations identified as main constraints.
- Thailand
  - National-level infrastructure improved; significant regional disparities exist with concentration near Bangkok.
  - After late 2011 floods, government rolled out flood-prevention investment including a B 350 billion (3 percent of GDP) off-budget package to be implemented over three years.
- Vietnam
  - Improvements in basic indicators: length of paved roads quadrupled over ten years; electricity production capacity increased by about 10 percent in two decades.
  - Challenges remain in quality of roads and ports and in public resource allocation efficiency.
- Post-Asian crisis dynamics
  - Gross fixed investment declined significantly in most ASEAN-5 economies.
  - Decline driven by a sharp fall in private investment, including foreign direct investment.
  - Private investment commitments to infrastructure sectors dropped since the Asian crisis (measured in US dollars and as a share of GDP).

### III. Methodology and results
- Data and sample
  - 76 advanced and emerging market economies.
  - Period: 1980-2010.
  - Use of non-overlapping 5-year averages; pooled OLS with heteroscedasticity-consistent standard errors; variables de-meaned using country-specific means (equivalent to panel with fixed country effects); time dummies included.
- Infrastructure indices construction (following Calderon and Serven)
  - Quantity index: first principal component of
    - total telephone lines, cell subscriptions, and internet users per 100 people (communication),
    - electricity production per capita (in millions of KWh),
    - road density (km of road per 100 sq. km of land area).
    - Quantity index eigenvectors: 0.6377*Communication + 0.6336*Electricity + 0.4381*Roaddensity.
  - Quality index: first principal component of
    - roads paved in percent of total roads,
    - electric power that is not lost in transmission and distribution in percent of total output.
    - Quality index eigenvectors: 0.7071*Pavedroads + 0.7071*Powerdistribution.
  - Note: waiting time for installation of main telephone lines excluded due to lack of data.
- Endogeneity treatment
  - Infrastructure indices enter regressions with one lag to address two-way causality.
- Main empirical findings
  - Infrastructure (both quality and quantity) is associated with reduced income inequality; coefficients negative and statistically significant across specifications.
  - Investment measures (total investment-to-GDP ratio, private investment-to-GDP ratio, public investment-to-GDP ratio), each entered with one lag, are negative but not statistically significant.
  - Addition of an index of institutional risks as a control for spending efficiency does not materially change baseline results.
  - Estimated potential impacts:
    - Catching up to the 2010 average levels of infrastructure quantity and quality indices in advanced economies is estimated to reduce the Gini index by about 2 percentage points in Vietnam, the Philippines, and Indonesia (based on Column 3 of Table 1).
    - Estimated decline in the Gini index for Malaysia is 1 percentage point.
- Selected coefficient estimates from Table 1 (dependent variable: natural logarithm of Gini) — coefficients shown exactly as in source
  - constant: 11.48 ***, 10.07 ***, 9.51 ***, -6.12 ***, 13.08 ***, 13.30 ***, 11.89 ***, 14.36 *** (t-statistics in parentheses in original table)
  - Inflation: 0.07 *, 0.05 *, 0.04, 0.10, 0.10 **, 0.09 **, 0.10 ***, 0.09 ** (t-statistics in parentheses)
  - Infrastructure quantity (lag 1): -0.20 ***, -0.1767 ***
  - Infrastructure quality (lag 1): -0.22 ***, -0.19 ***
  - Investment-to-GDP ratio (lag 1): -0.14, -0.11
  - Private investment-to-GDP ratio (lag 1): -0.09
  - Public investment-to-GDP ratio (lag 1): -0.03, -0.03
  - Credit-to-GDP ratio: 0.08 ***, 0.09 ***, 0.09 ***, 0.07 **, 0.02, 0.01, 0.06 *, 0.01
  - Employment share in industry (lag 1): -0.19 **, -0.16, -0.11, -0.19 **, -0.18 *, -0.17 *, -0.26 **, -0.22 *
  - PPP real GDP per capita: 2.73 ***, 2.06 ***, 2.29 ***, 2.34 ***, 2.73 ***, 2.77 ***, 2.41 ***, 2.89 ***
  - PPP real GDP per capita squared: -0.01 ***, -0.01 ***, -0.01 ***, -0.13 ***, -0.01 ***, -0.01 ***, -0.01 ***, -0.01 ***
  - Education spending in percent of GDP (lag 1): -0.08, -0.11 **, -0.10 *, -0.08 **, -0.07, -0.05, -0.08, -0.06
  - Openness: 0.05, 0.08 *, 0.13 ***, 0.01, 0.03, 0.03, 0.02, 0.03
  - Institutional risks: -0.04, -0.14
  - Observations: 128, 126, 111, 149, 107, 106, 147, 105
  - Adjusted R-squared: 0.61, 0.60, 0.70, 0.48, 0.49, 0.48, 0.49, 0.49
  - Note: An * indicates significance at 10%, ** significant at 5%, and *** at 1%. All explanatory variables are in natural logarithm.
- Interpretation and robustness
  - Infrastructure quantity and quality indices likely capture different aspects and both matter for income distribution.
  - Investment-to-GDP measures are imperfect proxies for infrastructure development due to potential spending inefficiency and governance/institutional factors.
  - Results broadly consistent with Kuznets hypothesis: PPP real GDP per capita positive, its square negative.
  - Education-related variables (education spending and workforce with secondary education) have negative and statistically significant coefficients, suggesting greater access to education reduces income inequality.
  - Private credit in percent of GDP shows a significant negative effect on income distribution in some specifications, but interpretation depends on institutions and regulation quality.
  - Employment share in industry associated with more equal income distribution.
  - Openness tends to make income distribution more unequal.

### IV. Conclusions and policy implications
- Key conclusion: Better infrastructure, both quantity and quality, improves income distribution and can contribute to poverty reduction and inclusive growth.
- Investment alone does not appear to promote income equality in the cross-country regressions; this may reflect a weak link between spending and accumulation of infrastructure assets and deficiencies in spending efficiency and institutions.
- Policy recommendations and implications:
  - Improve institutional frameworks and spending efficiency (e.g., streamline procurement, improve coordination among institutions) to strengthen the productivity of infrastructure investment.
  - Policies to promote education (increase education spending and access) can disproportionately raise earnings of lower-income groups and reduce inequality.
  - Encourage formal sector employment and expansion of industry, particularly in low-income regions, to provide more inclusive employment opportunities.
  - Advance financial development while improving institutions and broadening the poor’s access to finance to avoid disproportionate benefits accruing to the rich.
- Suggested area for future research: analyze how different types of infrastructure (for example, roads vs. broadband connection) affect inequality and growth differently.

*Source: IMF staff estimates and analysis in "2. Robustness Tests: Additional Estimation Results" (content unit)._

### APPENDIX I: DESCRIPTION OF DATA

### APPENDIX I: DESCRIPTION OF DATA

### Variable definitions and data sources
- Gini index: From the World Bank’s World Development Indicators (WDI) database, and is based on nominal per capita income averages and distributions estimated from household survey data. Missing years are further gap-filled with data from various sources including Povcal, LIS, and OECD databases. Natural logarithm of the variable is used for the regressions.
- Infrastructure quantity: Authors’ calculations based on the first principal component of: total telephone lines, cell subscriptions, and internet users per 100 people; electricity production per capita (in millions of KWh); and road density (km of road per 100 sq. km of land area). The underlying data are obtained from the WDI.
- Infrastructure quality: Authors’ calculations based on the first principal component of: roads paved in percent of total roads and electric power that is not lost in transmission and distribution in percent of total output. The underlying data are obtained from the WDI.
- Inflation: Year-on-year growth of consumer price index, obtained from the IMF’s World Economic Outlook database. Natural logarithm of inflation rate+100 are used for the regressions.
- Investment-to-GDP: Gross fixed capital formation in percent of GDP, obtained from the World Economic Outlook database. Natural logarithm of investment in percent of GDP is used in the regressions.
- Private investment-to-GDP: Gross private fixed capital formation divided by GDP. Both series are obtained separately from the World Economic Outlook database. Natural logarithm of investment in percent of GDP is used in the regressions.
- Public investment-to-GDP: Gross public fixed capital formation divided by GDP. Both series are obtained separately from the World Economic Outlook database. Natural logarithm of investment in percent of GDP is used in the regressions.
- Credit-to-GDP: Claims on private sector, obtained mainly from the IMF’s International Financial Statistics database, and gap-filled with data from the IMF’s Money and Banking database (MBRF2). This variable is divided by the GDP, which is obtained from the World Economic Outlook database to compile the credit-to-GDP ratio. Natural logarithm of credit-to-GDP is used in the regressions.
- Employment share in industry: Obtained from the World Development Indicators (WDI) database, which calculates employment in the industry sector, in percent of total employment. Natural logarithm of the variable is used in the regressions.
- PPP real GDP per capita: The underlying data for PPP adjusted real GDP per capita are obtained from the Penn World database. Natural logarithm of the variable is used in the regressions.
- Education spending in percent of GDP: Public spending on education in percent of GDP, obtained from the World Development Indicators (WDI) database. Public expenditure on education in WDI consists of current and capital expenditure on education, which includes government spending on educational institutions (both public and private), education administration as well as subsidies for private entities (students/households and other privates entities).
- Openness: Estimated as the sum of exports and imports in percent of GDP. The underlying data are obtained from the World Economic Outlook database. Natural logarithm of the openness variable is used in the regressions.
- Institutional risks: Institutional risk ratings for all countries in the sample are obtained from the International Country Risk Guide (ICRG) database of the PRS Group. The PRG Group calculates the overall index by using all 17 risk components, 12 with an 18-month forecast horizon and five with a five-year forecast horizon. A higher score indicates a lower risk environment.

### Country coverage (appendix II)
- Industrial countries (24): Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Israel, Italy, Japan, Luxembourg, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, United Kingdom, and United States
- Emerging markets (52): Albania, Algeria, Argentina, Armenia, Belarus, Bosnia & Herzegovina, Brazil, Bulgaria, Chile, China, Costa Rica, Colombia, Croatia, Czech Republic, Dominican Republic, Ecuador, Estonia, El Salvador, Georgia, Guatemala, Hong Kong, Hungary, Indonesia, India, Jamaica, Jordan, Korea, Latvia, Lebanon, Lithuania, Malaysia, Mexico, Morocco, Pakistan, Panama, Peru, Philippines, Poland, Romania, Russia, Sri Lanka, Serbia, Singapore, South Africa, Taiwan, Thailand, Tunisia, Turkey, Ukraine, Uruguay, Venezuela, Vietnam

### Additional estimations and robustness checks (appendix III)
- Estimation approaches: Fixed country effects and least-square estimations using variables that are not de-meaned.
- Robustness checks performed:
  - Dropping one variable at a time; results remain broadly unchanged.
  - Replacing the dependent variable (Gini index) with per capita income share held by the lowest 20 percent; results confirm that an improvement in infrastructure quality and quantity would increase the income share held by the poor.
- General finding reported: In all models, infrastructure quantity and quality indices remain statistically significant with negative coefficients.

### Table 2 — Determinants of the Gini Coefficient (selected reported estimates)
- Dependent variable: natural logarithm of Gini
- Constant: -8.48 *** -4.28 ** -5.81 *** -6.12 *** -6.25 *** -8.19 *** -8.52 *** -8.43 *** 1.41
  - t-statistics: (-4.22)(-2.19)(-2.76)(-3.37)(-3.35)(-4.51)(-4.99)(-4.74)(0.41)
- Inflation: 0.07 * 0.05 * 0.04 0.10 *** 0.10 *** 0.10 ** 0.09 ** 0.09 ** 0.02
  - t-statistics: (1.81)(1.74)(0.92)(2.98)(2.81)(2.30)(2.24)(2.12)(0.59)
- Infrastructure quantity (lag 1): -0.20 *** -0.1767 *** -0.11***
  - t-statistics: (-6.80)(-6.82)(-4.04)
- Infrastructure quality (lag 1): -0.22 *** -0.19 *** -0.16***
  - t-statistics: (-6.16)(-4.56)(-2.74)
- Investment-to-GDP ratio (lag 1): -0.14 -0.11
  - t-statistics: (-1.34)(-1.06)
- Private investment-to-GDP ratio (lag 1): -0.09
  - t-statistic: (-1.06)
- Public investment-to-GDP ratio (lag 1): -0.03 -0.03
  - t-statistics: (-0.80)(-0.85)
- Credit-to-GDP ratio: 0.08 *** 0.09 *** 0.09 *** 0.07 ** 0.06 * 0.02 0.01 0.01 0.01
  - t-statistics: (2.68)(2.73)(2.66)(2.05)(1.68)(0.42)(0.22)(0.25)(0.40)
- Employment share in industry (lag 1): -0.19 ** -0.16 -0.11 -0.19 ** -0.26 ** -0.18 * -0.17 * -0.22 * 0.06
  - t-statistics: (-2.35)(-1.62)(-1.06)(-2.29)(-2.42)(-1.75)(-1.67)(-1.90)(0.70)
- PPP real GDP per capita: 2.73 *** 2.06 *** 2.29 *** 2.34 *** 2.41 *** 2.73 *** 2.77 *** 2.89 *** 0.69
  - t-statistics: (5.92)(4.55)(4.74)(5.47)(5.50)(6.31)(6.57)(6.84)(0.94)
- PPP real GDP per capita squared: -0.14 *** -0.12 *** -0.11 *** -0.13 *** -0.14 *** -0.15 *** -0.15 *** -0.16 *** -0.04
  - t-statistics: (-5.35)(-4.80)(-4.18)(-5.77)(-5.85)(-6.37)(-6.60)(-6.89)(-0.92)
- Education spending in percent of GDP (lag 1): -0.08 -0.11 ** -0.10 * -0.08 -0.08 -0.07 * -0.05 -0.06 * -0.03
  - t-statistics: (-1.40)(-2.07)(-1.86)(-1.34)(-1.43)(-1.18)(-0.86)(-0.93)(-0.61)
- Openness: 0.05 0.08 * 0.13 *** 0.01 0.02 0.03 0.03 0.03 0.07
  - t-statistics: (0.97)(1.87)(2.61)(0.27)(0.41)(0.60)(0.51)(0.67)(1.13)
- Institutional risks: -0.04 -0.14
  - t-statistics: (-0.18)(-0.63)
- Observations128126111149147107106105123
- Adjusted R-squared0.61 0.60 0.70 0.48 0.49 0.49 0.48 0.49 0.62
- Estimation notes: An * indicates significance at 10%, ** significant at 5%, and *** at 1%. t statistics are in parentheses. All explanatory variables are in natural logarithm. The equations are estimated by OLS. Underlying data are from World Bank: World Development Indicators and PovCal databases, IMF: World Economic Outlook and International Financial Statistics, and Penn World Table 7.0.

*Source: IMF staff estimates.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp1341.pdf_