## _wp11104 - 2014. The dataset up to 2009 is based on realized historical data, and this section is used for

## Source details

**Canonical URL:** [_wp11104 - 2014. The dataset up to 2009 is based on realized historical data, and this section is used for](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11104.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11104.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11104.pdf.json)

---

### Data sources and coverage
- Macroeconomic variables: World Economic Outlook (International Monetary Fund, 2009).
- Demographics: World Bank’s World Development Indicators database.
- Effective exchange rate and trade weight: IMF Information Notice System.
- Net foreign assets: IMF Balance of Payments database.
- Institutional indicators: Worldwide Governance Indicators dataset (World Bank, 2009).
- Armed conflict data: Uppsala Conflict Data Program (2009).
- Time coverage: dataset up to 2009 uses realized historical data; remaining five years are IMF Staff’s projections used for forecasting in Section VI.
- Number of observations drawn from the common sample: 1543.

### Dataset adjustments and data quality rules
- Treated as data errors and excluded from estimation sample:
  - Negative values for nominal GDP, GDP at constant prices, government consumption, exports, imports, population, employment, exchange rate, and terms of trade.
  - Absolute values greater than 100 percent for dependency ratio, population growth, and for fiscal balance, government spending, current account balance and trade balance as a percent of GDP.
- Trade weight data are replaced by data reported by country authorities whenever there is a large discrepancy.
- Stationarity: panel unit root test statistics (Appendix B) indicate none of the variables have a unit root that cannot be rejected by a majority of test results.
- Multicollinearity: pairwise correlation coefficients (Appendix C) show high and significant correlation across governance indicators, and between governance indicators and income.

### Governance indicators (period 1996–2008)
- Six measures examined; country values range from -2.6 to 1.5 (higher = better governance):
  - Control of Corruption: extent public power is exercised for private gain.
  - Government Effectiveness: quality of public services and policy formulation, capacity and independence of civil service.
  - Political Stability: likelihood of government destabilization by unconstitutional or violent means.
  - Regulatory Quality: capacity to provide sound policies and regulations promoting private sector development.
  - Rule of Law: confidence in law, contract enforcement, property rights, police, and courts.
  - Voice and Accountability: citizens’ capacity to select government, freedom of expression, association, and free media.

### Classification of resource-rich countries
- Resource-rich classification based on main exports in IMF’s World Economic Outlook (Appendix A).
- Of 150 low- and middle-income countries (LIMC), 45 are classified as resource-rich.
  - 26 countries classified as fuel exporters: around half from the Middle East, a little less than one third from Africa.
  - 19 countries export other primary commodities: more than half representing African countries.

### Summary statistics (common estimation sample, selected)
- Per Capita Growth: Min. -0.33, Mean 0.05, Max. 0.86, Std. Dev. 0.05, Skewness 2.21, Kurtosis 41.22.
- Resource (0-1): Min. 0.000, Mean 0.30, Max. 1.000, Std. Dev. 0.46, Skewness 0.89, Kurtosis 1.79.
- Fuel (0-1): Min. 0.000, Mean 0.13, Max. 1.000, Std. Dev. 0.34, Skewness 1.79, Kurtosis 4.20.
- Initial Income (logs, in thousands of US dollar): Min. -1.20, Mean 1.22, Max. 4.45, Std. Dev. 1.10, Skewness -0.03, Kurtosis 2.28.
- REER (in logs; ratioed by 100): Min. -1.49, Mean -0.01, Max. 1.47, Std. Dev. 0.24, Skewness -1.24, Kurtosis 14.60.
- Imports/GDP: Min. 0.000, Mean 0.47, Max. 6.28, Std. Dev. 0.29, Skewness 6.45, Kurtosis 115.62.
- Terms of Trade (in logs; ratioed by 100): Min. -0.95, Mean 0.00, Max. 0.87, Std. Dev. 0.22, Skewness -0.13, Kurtosis 6.10.
- Fiscal Balance/GDP: Min. -0.49, Mean -0.02, Max. 0.43, Std. Dev. 0.07, Skewness 1.29, Kurtosis 15.33.
- Control of Corruption: Min. -1.76, Mean -0.35, Max. 1.51, Std. Dev. 0.66, Skewness 0.54, Kurtosis 2.72.
- Government Effectiveness: Min. -1.88, Mean -0.32, Max. 1.49, Std. Dev. 0.66, Skewness 0.31, Kurtosis 2.55.
- Political Stability: Min. -2.61, Mean -0.27, Max. 1.40, Std. Dev. 0.83, Skewness -0.35, Kurtosis 2.57.
- Regulatory Quality: Min. -2.44, Mean -0.25, Max. 1.58, Std. Dev. 0.75, Skewness -0.15, Kurtosis 2.76.
- Rule of Law: Min. -1.88, Mean -0.36, Max. 1.40, Std. Dev. 0.71, Skewness 0.30, Kurtosis 2.28.
- Voice and Accountability: Min. -2.24, Mean -0.32, Max. 1.46, Std. Dev. 0.84, Skewness -0.01, Kurtosis 2.07.
- Armed Conflict (0-1): Min. 0.000, Mean 0.17, Max. 1.000, Std. Dev. 0.37, Skewness 1.78, Kurtosis 4.16.
- Old Age Dependency: Min. 0.01, Mean 0.09, Max. 0.26, Std. Dev. 0.05, Skewness 1.51, Kurtosis 4.41.
- Population Growth: Min. -0.17, Mean 0.02, Max. 0.26, Std. Dev. 0.02, Skewness 2.13, Kurtosis 39.92.
- Notes: Per Capita Growth calculated from the log-difference. Resource, Fuel, and Armed Conflict are 0-1 indicator variables. REER and terms of trade ratioed by 100, and income in thousands of US dollar. Old Age Dependency = age 65+ / ages 16 to 65.

### Comparative growth findings: resource-rich vs diversified LMICs
- Average growth rate for LIMC: 5 percent (Mean per capita growth).
- High variance across countries: 5 percent standard deviation; range from -33 percent to 86 percent.
- Initial real income per capita for LIMC: around 3,400 US dollar (Mean initial income), standard deviation around 3,000 US dollars.
- Stylized relationships (1990–2008):
  - Poorer diversified LMICs tend to grow faster than richer ones (inverse relationship between trend growth and initial income).
  - Resource-rich LMICs:
    - At low income levels (per capita incomes < $1,000 in PPP), per capita income growth tends to fall below diversified peers.
    - At higher income levels, resource-rich LMICs grow in line with or faster than diversified peers.
- Table 2: Average per capita income growth, 1990-2008 (by initial income in 1973-1989):
  - Initial income < $500: Diversified 5.7, Resource-rich 2.7, Difference -3.0.
  - $500 to < $1,000: Diversified 3.5, Resource-rich 3.3, Difference -0.2.
  - $1,000 to < $5,000: Diversified 3.9, Resource-rich 4.5, Difference +0.6.
  - Over $5,000: Diversified 4.1, Resource-rich 3.5, Difference -0.6.
- Interpretation: evidence consistent with a resource-based poverty trap where low-income resource-rich countries grow more slowly than diversified peers; difference of 3 percentage points per annum for incomes < $500.

### Empirical methodology
- Estimation approach: panel least squares (Panel LS) and dynamic generalized method of moments (Dynamic Panel).
- Core growth equations (notation preserved from source):
  - G_i,t = logarithmic growth rate of real income per capita.
  - R_i = 0-1 indicator (1 if resource-rich).
  - Y_i,t-1 = initial income (lagged log real income per capita).
  - I_{YY* <= } = 0-1 indicator taking value 1 if initial income < Y*.
  - Z_it and EG_it are indicator variables for Zimbabwe and Equatorial Guinea (outliers).
- Hypotheses:
  - Resource curse: β1 < 0.
  - Poverty trap for poor resource-rich economies: β3 < 0 (coefficient on interaction/resource indicator with low initial income).
- Choice of Y*: equations solved for six levels of Y* from $500 to $3,000 (increments of $500); optimal model around Y* = $2,000 based on error variances (note: only results for Y* = 2000 reported in table).

### Estimation results (selected from Table 3; Y* = 2000)
- Models and number of observations:
  - Panel LS (Annual Data): 5044.
  - Dynamic Panel (Annual Data): 4744.
  - Panel LS (Non-overlapping 9-Year Averages): 431.
  - Dynamic Panel (Non-overlapping 9-Year Averages): 284.
- Key coefficient estimates (point estimate; std. error in parentheses; statistical significance indicated in source table):
  - Resource:
    - Panel LS (Annual): -0.004 (0.003) *
    - Dynamic Panel (Annual): -0.002 (0.002)
    - Panel LS (9-Year): -0.003 (0.005)
    - Dynamic Panel (9-Year): -0.005 (0.007)
  - Initial Income pc:
    - Panel LS (Annual): -0.007 (0.001) ****
    - Dynamic Panel (Annual): -0.003 (0.001) ****
    - Panel LS (9-Year): -0.004 (0.002) **
    - Dynamic Panel (9-Year): -0.005 (0.004) *
  - Resource I (y* > y) (y*-y):
    - Panel LS (Annual): -0.009 (0.003) ****
    - Dynamic Panel (Annual): -0.004 (0.002) ***
    - Panel LS (9-Year): -0.008 (0.006) *
    - Dynamic Panel (9-Year): -0.019 (0.011) **
  - Growth (-1) (lagged growth) present in Dynamic Panels:
    - Dynamic Panel (Annual): 0.542 (0.075) ****
    - Dynamic Panel (9-Year): -0.747 (0.302) ***
  - Zimbabwe dummy:
    - Panel LS (Annual): -0.134 (0.014) ****
    - Dynamic Panel (Annual): -0.073 (0.011) ****
    - Panel LS (9-Year): -0.125 (0.022) ****
    - Dynamic Panel (9-Year): -0.191 (0.015) ****
  - Equatorial Guinea dummy:
    - Panel LS (Annual): 0.080 (0.012) ****
    - Dynamic Panel (Annual): 0.036 (0.006) ****
    - Panel LS (9-Year): 0.099 (0.022) ****
    - Dynamic Panel (9-Year): 0.239 (0.026) ****
- Goodness-of-fit and residuals:
  - Adjusted R Squared: 0.0306 (Panel LS annual), 0.1553 (Dynamic annual), 0.1033 (Panel LS 9-year), 0.0017 (Dynamic 9-year).
  - Std Error of Regression: 0.0726, 0.0672, 0.0366, 0.0423 respectively.
  - Sum of Squared Residuals: 26.532, 21.421, 0.5699, 0.496 respectively.

### Main empirical conclusions
- Evidence supports both an underlying resource curse and a low-income poverty trap effect; the poverty-trap effect dominates.
- Resource curse is significantly different from zero only at the 20 percent confidence level in the static model.
- Convergence effect is weak:
  - In the static model, a country with initial per capita income of $2,000 would grow 0.5 percent per annum faster than a country with per capita income of $4,000 (implying a full convergence period of 350 years).
- Estimated coefficients on the interaction term (resource × low initial income) are negative and often statistically significant, consistent with poorer resource-rich countries experiencing slower growth relative to higher-income peers.

### Estimated growth by initial income and resource intensity (Table 4 interpretation)
- Table 4 reported estimated growth at initial incomes $500, $1,000, $2,000, $4,000 for diversified and resource-rich LMICs; reported illustrative partial effects:
  - Diversified LMICs growth (at initial income): values shown as 6.5 6.0 5.5 5.0 for $500, $1,000, $2,000, $4,000 respectively (table formatting).
  - Resource-rich LMICs growth (at initial income): values shown as 4.8 5.0 5.1 4.6 for $500, $1,000, $2,000, $4,000 respectively.
  - Resource-rich dummy: -0.4 (applied across incomes).
  - Growth convergence effect: -0.7 * ln(Income) with reported partial effects 0.5 0.0 -0.5 -1.0 (for $500, $1,000, $2,000, $4,000).
  - Low-income poverty trap effect: -0.9 * ln(2000-Income) with reported partial effects -1.3 -0.7 0.0 0.0 (for $500, $1,000, $2,000, $4,000).
- Interpretative findings:
  - A resource-rich country with per capita incomes of $2,000 would grow at broadly the same rate as a diversified economy with per capita incomes of $4,000.
  - Resource-rich country with initial income of $1,000 would grow 0.7 percent slower than otherwise.
  - Resource-rich country with initial income of $500 would grow 1.3 percent slower than otherwise.
  - The income divergence effect fades as country income approaches the $2,000 threshold; for a country with initial income of $500, reaching $2,000 would take around 30 years.

### Dynamic model and persistence versus long-run convergence
- Dynamic model (annual):
  - Lagged growth coefficient estimate (α) is significant and positive; roughly half of growth this year is due to last year’s growth performance (reported 0.5α = 0.5 in text).
  - Persistence/business-cycle interpretation: positive lagged coefficient implies short-run persistence.
- Long-run reinterpretation:
  - Re-writing the dynamic model in the long-run yields equivalent coefficient estimates to static panel when multiplying by 2 given 0.5α = 0.5.
  - Using non-overlapping 9-year averaged data:
    - Coefficients from annual and 9-year averaged data are similar in magnitude except for the lagged growth term.
    - Lagged growth coefficient becomes negative 0.75 in 9-year averaged dynamic panel GMM (indicating long-run convergence to potential).

### Robustness checks
- Inclusion of an uninteracted poverty-trap term for the poor only was statistically insignificant at the 20-percent confidence level; other coefficients robust (results available upon request).
- 9-year averaging rationale and caveats:
  - 9-year averages assumed long enough to capture business cycles and remove impact on coefficients.
  - Caveats: (1) start dates/durations of business cycles vary across countries and may not align with 9-year windows; (2) 9-year averaging excludes countries with few time-series observations and reduces variation, decreasing efficiency of estimates.

### Impact of macroeconomic and structural policies (Model (4), Table 5)
- Model (4) adds macroeconomic (X_i,t) and structural (S_i,t) variables:
  - Macroeconomic variables: Real Effective Exchange Rate (REER), Terms of Trade (TOT), Fiscal balance/GDP, Imports/GDP.
  - Structural variables: Old Age Dependency, Population Growth.
- Key coefficient signs and magnitudes (reported in text and Table 5 highlights):
  - REER (depreciation improves competitiveness) expected negative coefficient for growth.
  - Imports/GDP positive and significant: open economies have higher growth rates.
  - Terms of Trade positive effect when favorable.
  - Fiscal balance/GDP positive: fiscal austerity leads to higher growth rates.
  - Old Age Dependency positive: better demographics associated with higher growth.
  - Population Growth negative.
- Important finding:
  - In the more richly specified model, the resource-rich dummy becomes positive and significant: resource-rich countries are calculated to grow around 1 percent per annum faster than more diversified economies, other factors equal.
  - However, poverty-trap coefficient for poor resource-rich countries remains significant at the 1-percent confidence interval for annual data and weakly significant at the 20-percent level for 9-year averaged data — institutional and macro-structural variables do not overturn the poverty-trap explanatory power.

### Decomposition of macroeconomic and structural differences (Table 6)
- Table 6 compares means across Resource-rich LMICs and Diversified LMICs and estimates growth impact (Model (4)). Selected entries (column E shows estimated growth impact):
  - Real effective exchange rate: difference 0.014, coefficient -0.005, impact -0.01%.
  - Import/GDP: difference -0.068, coefficient 0.037, impact -0.25%.
  - Terms of trade: difference -0.009, coefficient 0.011, impact -0.01%.
  - Fiscal balance/GDP: difference 0.029, coefficient 0.190, impact 0.56%.
  - Old age dependency: difference -0.032, coefficient 0.203, impact -0.65%.
  - Population growth: difference 0.008, coefficient -0.436, impact -0.35%.
- Net effect reported: Macro and structural explanatory variables help to explain on average 0.7 percent per annum slower growth for resource-rich countries; when these variables are included the resource dummy sign flips from negative to positive.

### Institutions matter (Model (5), Table 7)
- Model (5) adds fuel-exporter dummy (F_i,t) and institutional indicators Z_i,t (Control of Corruption, Government Effectiveness, Political Stability, Regulatory Quality, Rule of Law, Voice and Accountability, Armed Conflict).
- Estimation approach: 9-year averaged data with cross-sectional random effects; columns (1)–(7) include one institutional variable at a time to reduce collinearity; (1')–(7') report reduced-form models.
- Main results:
  - Strong institutions have a significant and positive impact on growth: improvement in indicators by one point increases growth by 0.5 to 2.3 percent (text summary).
  - Control of Corruption, Government Effectiveness, and Rule of Law show the largest positive impacts on growth.
  - Once institutional factors are included, the low-income poverty-trap falls in size and becomes insignificantly different from zero — suggesting the poverty trap for poor resource-rich countries is associated with weak institutions.
  - Fuel exporters coefficient: positive and significant across Table 7 columns (examples reported: 0.028 ****, 0.032 ****, 0.024 ***) indicating fuel-exporting resource-rich economies exhibit higher growth conditional on controls.

### Application to Ghana: where Ghana stands and projections
- Comparative performance (1996–2008):
  - Ghana grew on average 4.6 percent during the last two decades, less than averages for both resource-rich and diversified LMICs.
  - Ghana’s initial income and per capita income: Ghana had less than one third of average per capita income of both resource-rich and diversified LMICs during last two decades.
  - REER: Ghana’s real exchange rate appreciated on average 17 percent during whole sample, whereas LMIC exchange rates depreciated on average by 1 percent.
  - Imports/GDP: Ghana more open than peers; import-to-GDP rose from average 43 percent (late 1990s) to 64 percent (late 2000s).
  - Terms of Trade: Ghana benefited from positive TOT shocks during the last two decades.
  - Fiscal stance: Ghana had significantly worse fiscal accounts than peers, on average by around 7 percent of GDP; fiscal deficit declined from 12.6 percent in late 1990s to 6.6 percent in late 2000s; during 2005–2008 gap to resource-rich peers rose to 12 percent despite being a cocoa and gold exporter.
- Governance and structural indicators (Tables 9 and 10):
  - Ghana performs better than both resource-rich and diversified LMICs on governance indicators and has been improving over time.
  - Structural: no armed conflict (1996–2008); Ghana receives more aid inflows as ratio of GDP; old age dependency lower and population growth higher relative to peers.
- Growth projections for Ghana (based on Table 7 regression using Government Effectiveness and applying model to Ghana):
  - Forecasted growth (selected years and scenarios):
    - 2009: 4.5%.
    - 2010: 4.7%.
    - 2011: growth without oil 7.7% / growth with oil 5.8%.
    - 2014: growth without oil 8.2% / growth with oil 6.3%.
  - Text decomposition of drivers:
    - Growth rate projected to increase from 4.5 percent in 2009 to 8.2 percent in 2014.
    - Two main drivers:
      1. Improvement in fiscal accounts: negative impact of fiscal deficit fades, yielding around 1.5 percent additional growth.
      2. Oil production with strong institutions: Ghana will benefit from an additional 2 percent of growth as an oil exporter.
  - Concluding applied result: Ghana will achieve an additional growth rate of 2 percent in the medium-term as an oil-exporting economy (conditional on fiscal improvement and institutional quality).

### Paper-wide conclusion
- Dataset: 150 low and middle income countries, 1973–2008, analyzed with static and dynamic panel techniques.
- Main findings:
  - There is a poverty trap for poor resource-rich countries driven by low institutional quality.
  - For countries with good governance and strong macroeconomic management, resource wealth (oil) can be utilized to achieve higher economic growth.
  - Institutional quality is crucial: improving institutions can eliminate the observed low-income resource-trap effect.
- Application: Based on results, Ghana is expected to gain an additional 2 percent medium-term growth as an oil-exporting economy, conditional on fiscal improvement and institutional quality.

*Source: _wp11104 (excerpt) — dataset description, summary statistics, methodology, regression results, decompositions, and Ghana application.*

### 2014. The dataset up to 2009 is based on realized historical data, and this section is used for

### _wp11104 - 2014. The dataset up to 2009 is based on realized historical data, and this section is used for

### Data sources and coverage
- Macroeconomic variables: World Economic Outlook (International Monetary Fund, 2009).
- Demographics: World Bank’s World Development Indicators database.
- Effective exchange rate and trade weight: IMF Information Notice System.
- Net foreign assets: IMF Balance of Payments database.
- Institutional indicators: Worldwide Governance Indicators dataset (World Bank, 2009).
- Armed conflict data: Uppsala Conflict Data Program (2009).
- Time coverage: dataset up to 2009 uses realized historical data; remaining five years are IMF Staff’s projections used for forecasting in Section VI.
- Number of observations drawn from the common sample: 1543.

### Dataset adjustments and data quality rules
- Treated as data errors and excluded from estimation sample:
  - Negative values for nominal GDP, GDP at constant prices, government consumption, exports, imports, population, employment, exchange rate, and terms of trade.
  - Absolute values greater than 100 percent for dependency ratio, population growth, and for fiscal balance, government spending, current account balance and trade balance as a percent of GDP.
- Trade weight data are replaced by data reported by country authorities whenever there is a large discrepancy.
- Stationarity: panel unit root test statistics (Appendix B) indicate none of the variables have a unit root that cannot be rejected by a majority of test results.
- Multicollinearity: pairwise correlation coefficients (Appendix C) show high and significant correlation across governance indicators, and between governance indicators and income.

### Governance indicators (period 1996–2008)
- Six measures examined; country values range from -2.6 to 1.5 (higher = better governance):
  - Control of Corruption: extent public power is exercised for private gain.
  - Government Effectiveness: quality of public services and policy formulation, capacity and independence of civil service.
  - Political Stability: likelihood of government destabilization by unconstitutional or violent means.
  - Regulatory Quality: capacity to provide sound policies and regulations promoting private sector development.
  - Rule of Law: confidence in law, contract enforcement, property rights, police, and courts.
  - Voice and Accountability: citizens’ capacity to select government, freedom of expression, association, and free media.

### Classification of resource-rich countries
- Resource-rich classification based on main exports in IMF’s World Economic Outlook (Appendix A).
- Of 150 low- and middle-income countries (LIMC), 45 are classified as resource-rich.
  - 26 countries classified as fuel exporters: around half from the Middle East, a little less than one third from Africa.
  - 19 countries export other primary commodities: more than half representing African countries.

### Summary statistics (common estimation sample, selected)
- Per Capita Growth: Min. -0.33, Mean 0.05, Max. 0.86, Std. Dev. 0.05, Skewness 2.21, Kurtosis 41.22.
- Resource (0-1): Min. 0.000, Mean 0.30, Max. 1.000, Std. Dev. 0.46, Skewness 0.89, Kurtosis 1.79.
- Fuel (0-1): Min. 0.000, Mean 0.13, Max. 1.000, Std. Dev. 0.34, Skewness 1.79, Kurtosis 4.20.
- Initial Income (logs, in thousands of US dollar): Min. -1.20, Mean 1.22, Max. 4.45, Std. Dev. 1.10, Skewness -0.03, Kurtosis 2.28.
- REER (in logs; ratioed by 100): Min. -1.49, Mean -0.01, Max. 1.47, Std. Dev. 0.24, Skewness -1.24, Kurtosis 14.60.
- Imports/GDP: Min. 0.000, Mean 0.47, Max. 6.28, Std. Dev. 0.29, Skewness 6.45, Kurtosis 115.62.
- Terms of Trade (in logs; ratioed by 100): Min. -0.95, Mean 0.00, Max. 0.87, Std. Dev. 0.22, Skewness -0.13, Kurtosis 6.10.
- Fiscal Balance/GDP: Min. -0.49, Mean -0.02, Max. 0.43, Std. Dev. 0.07, Skewness 1.29, Kurtosis 15.33.
- Control of Corruption: Min. -1.76, Mean -0.35, Max. 1.51, Std. Dev. 0.66, Skewness 0.54, Kurtosis 2.72.
- Government Effectiveness: Min. -1.88, Mean -0.32, Max. 1.49, Std. Dev. 0.66, Skewness 0.31, Kurtosis 2.55.
- Political Stability: Min. -2.61, Mean -0.27, Max. 1.40, Std. Dev. 0.83, Skewness -0.35, Kurtosis 2.57.
- Regulatory Quality: Min. -2.44, Mean -0.25, Max. 1.58, Std. Dev. 0.75, Skewness -0.15, Kurtosis 2.76.
- Rule of Law: Min. -1.88, Mean -0.36, Max. 1.40, Std. Dev. 0.71, Skewness 0.30, Kurtosis 2.28.
- Voice and Accountability: Min. -2.24, Mean -0.32, Max. 1.46, Std. Dev. 0.84, Skewness -0.01, Kurtosis 2.07.
- Armed Conflict (0-1): Min. 0.000, Mean 0.17, Max. 1.000, Std. Dev. 0.37, Skewness 1.78, Kurtosis 4.16.
- Old Age Dependency: Min. 0.01, Mean 0.09, Max. 0.26, Std. Dev. 0.05, Skewness 1.51, Kurtosis 4.41.
- Population Growth: Min. -0.17, Mean 0.02, Max. 0.26, Std. Dev. 0.02, Skewness 2.13, Kurtosis 39.92.
- Notes: Per Capita Growth calculated from the log-difference. Resource, Fuel, and Armed Conflict are 0-1 indicator variables. REER and terms of trade ratioed by 100, and income in thousands of US dollar. Old Age Dependency = age 65+ / ages 16 to 65.

### Comparative growth findings: resource-rich vs diversified LMICs
- Average growth rate for LIMC: 5 percent (Mean per capita growth).
- High variance across countries: 5 percent standard deviation; range from -33 percent to 86 percent.
- Initial real income per capita for LIMC: around 3,400 US dollar (Mean initial income), standard deviation around 3,000 US dollars.
- Figure 1 (1990–2008): poorer diversified LMICs tend to grow faster than richer ones (inverse relationship between trend growth and initial income).
- Resource-rich LMICs:
  - At low income levels (per capita incomes < $1,000 in PPP), per capita income growth tends to fall below diversified peers.
  - At higher income levels, resource-rich LMICs grow in line with or faster than diversified peers.
- Table 2. Average per capita income growth, 1990-2008 (by initial income in 1973-1989):
  - Initial income < $500: Diversified 5.7, Resource-rich 2.7, Difference -3.0.
  - $500 to < $1,000: Diversified 3.5, Resource-rich 3.3, Difference -0.2.
  - $1,000 to < $5,000: Diversified 3.9, Resource-rich 4.5, Difference +0.6.
  - Over $5,000: Diversified 4.1, Resource-rich 3.5, Difference -0.6.
- Interpretation: evidence consistent with a resource-based poverty trap where low-income resource-rich countries grow more slowly than diversified peers; difference of 3 percentage points per annum for incomes < $500.

### Empirical methodology
- Estimation approach: panel least squares (Panel LS) and dynamic generalized method of moments (Dynamic Panel).
- Core growth equations (notation preserved from source):
  - G_i,t = logarithmic growth rate of real income per capita.
  - R_i = 0-1 indicator (1 if resource-rich).
  - Y_i,t-1 = initial income (lagged log real income per capita).
  - I_{YY* <= } = 0-1 indicator taking value 1 if initial income < Y*.
  - Z_it and EG_it are indicator variables for Zimbabwe and Equatorial Guinea (outliers).
- Hypotheses:
  - Resource curse: β1 < 0.
  - Poverty trap for poor resource-rich economies: β3 < 0 (coefficient on interaction/resource indicator with low initial income).
- Choice of Y*: equations solved for six levels of Y* from $500 to $3,000 (increments of $500); optimal model around Y* = $2,000 based on error variances (note: only results for Y* = 2000 reported in table).

### Estimation results (selected from Table 3; Y* = 2000)
- Models: Panel LS (Annual Data), Dynamic Panel (Annual Data), Panel LS (Non-overlapping 9-Year Averages), Dynamic Panel (Non-overlapping 9-Year Averages).
- Number of Observations: 5044 (Panel LS annual), 4744 (Dynamic annual), 431 (Panel LS 9-year averages), 284 (Dynamic 9-year).
- Key coefficient estimates (point estimate; statistical significance indicated in table via ****, ***, **, *):
  - Resource:
    - Panel LS (Annual): -0.004 * (std. error 0.003)
    - Dynamic Panel (Annual): -0.002 (std. error 0.002)
    - Panel LS (9-Year): -0.003 (std. error 0.005)
    - Dynamic Panel (9-Year): -0.005 (std. error 0.007)
  - Initial Income pc:
    - Panel LS (Annual): -0.007 **** (std. error 0.001)
    - Dynamic Panel (Annual): -0.003 **** (std. error 0.001)
    - Panel LS (9-Year): -0.004 ** (std. error 0.002)
    - Dynamic Panel (9-Year): -0.005 * (std. error 0.004)
  - Resource I (y* > y) (y*-y):
    - Panel LS (Annual): -0.009 **** (std. error 0.003)
    - Dynamic Panel (Annual): -0.004 *** (std. error 0.002)
    - Panel LS (9-Year): -0.008 * (std. error 0.006)
    - Dynamic Panel (9-Year): -0.019 ** (std. error 0.011)
  - Growth (-1) (lagged growth) present in Dynamic Panels:
    - Dynamic Panel (Annual): 0.542 **** (std. error 0.075)
    - Dynamic Panel (9-Year): -0.747 *** (std. error 0.302)
  - Zimbabwe dummy:
    - Panel LS (Annual): -0.134 **** (std. error 0.014)
    - Dynamic Panel (Annual): -0.073 **** (std. error 0.011)
    - Panel LS (9-Year): -0.125 **** (std. error 0.022)
    - Dynamic Panel (9-Year): -0.191 **** (std. error 0.015)
  - Equatorial Guinea dummy:
    - Panel LS (Annual): 0.080 **** (std. error 0.012)
    - Dynamic Panel (Annual): 0.036 **** (std. error 0.006)
    - Panel LS (9-Year): 0.099 **** (std. error 0.022)
    - Dynamic Panel (9-Year): 0.239 **** (std. error 0.026)
- Goodness-of-fit and residuals:
  - Adjusted R Squared: 0.0306 (Panel LS annual), 0.1553 (Dynamic annual), 0.1033 (Panel LS 9-year), 0.0017 (Dynamic 9-year).
  - Std Error of Regression: 0.0726, 0.0672, 0.0366, 0.0423 respectively.
  - Sum of Squared Residuals: 26.532, 21.421, 0.5699, 0.496 respectively.

### Main empirical conclusions
- Evidence supports both an underlying resource curse and a low-income poverty trap effect; the poverty-trap effect dominates.
- Resource curse is significantly different from zero only at the 20 percent confidence level in the static model.
- Convergence effect is weak:
  - In the static model, a country with initial per capita income of $2,000 would grow 0.5 percent per annum faster than a country with per capita income of $4,000 (implying a full convergence period of 350 years).
- Estimated coefficients on the interaction term (resource × low initial income) are negative and often statistically significant, consistent with poorer resource-rich countries experiencing slower growth relative to higher-income peers.

*Source: _wp11104 (excerpt) — dataset description, summary statistics, methodology, and main regression results.*

### 0.4 percent per annum more slowly than their diversified peers, though, as noted above, this

### _wp11104 - 0.4 percent per annum more slowly than their diversified peers, though, as noted above, this

### Estimated growth by initial income and resource intensity
- Table 4: Estimated growth at different initial income levels (based on regression coefficients from equation (1)):
  - Diversified LMICs growth (at initial income): $500: 6.5 6.0 5.5 5.0 (table formatting shows values for $500, $1,000, $2,000, $4,000)
  - Resource-rich LMICs growth (at initial income): $500: 4.8 5.0 5.1 4.6
  - Resource-rich dummy: -0.4 (applied across incomes)
  - Growth convergence effect: -0.7 * ln(Income) with reported partial effects at incomes: 0.5 0.0 -0.5 -1.0 (for $500, $1,000, $2,000, $4,000)
  - Low-income poverty trap effect: -0.9 * ln(2000-Income) with reported partial effects at incomes: -1.3 -0.7 0.0 0.0 (for $500, $1,000, $2,000, $4,000)
- Interpretation:
  - A resource-rich country with per capita incomes of $2,000 would grow at broadly the same rate as a diversified economy with per capita incomes of $4,000.
  - The poverty-trap term captures that the poorest resource-rich countries grow the least; coefficient estimates are highly significant.
  - Examples from Table 4:
    - Resource-rich country with initial income of $1,000 would grow 0.7 percent slower than otherwise.
    - Resource-rich country with initial income of $500 would grow 1.3 percent slower than otherwise.
  - The income divergence effect fades as country income approaches the $2,000 threshold; for a country with initial income of $500, reaching $2,000 would take around 30 years.

### Dynamic model and persistence versus long-run convergence
- Dynamic model (second column of Table 3):
  - Lagged growth coefficient estimate (α) is significant and positive; roughly half of growth this year is due to last year’s growth performance (reported 0.5α = 0.5 in text).
  - Persistence/business-cycle interpretation: positive lagged coefficient implies short-run persistence.
- Robustness and long-run reinterpretation:
  - Re-writing the dynamic model in the long-run yields equivalent coefficient estimates to static panel when multiplying by 2 given 0.5α = 0.5.
  - Using non-overlapping 9-year averaged data:
    - Coefficients from annual and 9-year averaged data are similar in magnitude except for the lagged growth term.
    - Lagged growth coefficient becomes negative 0.75 in 9-year averaged dynamic panel GMM (last column of Table 3), indicating long-run convergence to potential.

### Robustness checks
- Inclusion of an uninteracted poverty-trap term for the poor only was statistically insignificant at the 20-percent confidence level; other coefficients robust (results available upon request).
- 9-year averaging rationale and caveats:
  - 9-year averages assumed long enough to capture business cycles and remove impact on coefficients.
  - Caveats: (1) start dates/durations of business cycles vary across countries and may not align with 9-year windows; (2) 9-year averaging excludes countries with few time-series observations and reduces variation, decreasing efficiency of estimates.

### Impact of macroeconomic and structural policies (Model (4), Table 5)
- Model (4) adds macroeconomic (X_i,t) and structural (S_i,t) variables:
  - Macroeconomic variables: Real Effective Exchange Rate (REER), Terms of Trade (TOT), Fiscal balance/GDP, Imports/GDP.
  - Structural variables: Old Age Dependency, Population Growth.
- Key coefficient signs and magnitudes (Table 5 highlights and text):
  - REER (depreciation improves competitiveness) expected negative coefficient for growth.
  - Imports/GDP positive and significant: open economies have higher growth rates.
  - Terms of Trade positive effect when favorable.
  - Fiscal balance/GDP positive: fiscal austerity leads to higher growth rates.
  - Old Age Dependency positive: better demographics associated with higher growth.
  - Population Growth negative.
- Important finding:
  - In the more richly specified model, the resource-rich dummy becomes positive and significant: resource-rich countries are calculated to grow around 1 percent per annum faster than more diversified economies, other factors equal.
  - However, poverty-trap coefficient for poor resource-rich countries remains significant at the 1-percent confidence interval for annual data and weakly significant at the 20-percent level for 9-year averaged data — institutional and macro-structural variables do not overturn the poverty-trap explanatory power.

### Decomposition of macroeconomic and structural differences (Table 6)
- Table 6 compares means of variables across Resource-rich LMICs and Diversified LMICs and estimates growth impact (Model (4)):
  - Selected entries (column E shows estimated growth impact):
    - Real effective exchange rate: difference 0.014, coefficient -0.005, impact -0.01%
    - Import/GDP: difference -0.068, coefficient 0.037, impact -0.25%
    - Terms of trade: difference -0.009, coefficient 0.011, impact -0.01%
    - Fiscal balance/GDP: difference 0.029, coefficient 0.190, impact 0.56%
    - Old age dependency: difference -0.032, coefficient 0.203, impact -0.65%
    - Population growth: difference 0.008, coefficient -0.436, impact -0.35%
  - Net effect: Macro and structural explanatory variables help to explain on average 0.7 percent per annum slower growth for resource-rich countries; when these variables are included the resource dummy sign flips from negative to positive.

### Institutions matter (Model (5), Table 7)
- Model (5) adds fuel-exporter dummy (F_i,t) and institutional indicators Z_i,t (Control of Corruption, Government Effectiveness, Political Stability, Regulatory Quality, Rule of Law, Voice and Accountability, Armed Conflict).
- Estimation approach: 9-year averaged data with cross-sectional random effects; columns (1)–(7) include one institutional variable at a time to reduce collinearity; (1')–(7') report reduced-form models.
- Main results:
  - Strong institutions have a significant and positive impact on growth: improvement in indicators by one point increases growth by 0.5 to 2.3 percent (text summary).
  - Control of Corruption, Government Effectiveness, and Rule of Law show the largest positive impacts on growth.
  - Once institutional factors are included, the low-income poverty-trap falls in size and becomes insignificantly different from zero — suggesting the poverty trap for poor resource-rich countries is associated with weak institutions.
  - Fuel exporters coefficient:
    - Fuel Exporters positive and significant across Table 7 columns (e.g., 0.028 ****, 0.032 ****, 0.024 *** etc.), indicating fuel-exporting resource-rich economies exhibit higher growth conditional on controls.

### Application to Ghana: where Ghana stands and projections
- Comparative performance (1996–2008) highlights:
  - Ghana grew on average 4.6 percent during the last two decades, less than averages for both resource-rich and diversified LMICs.
  - Ghana’s initial income and per capita income: Ghana had less than one third of average per capita income of both resource-rich and diversified LMICs during last two decades.
  - REER: Ghana’s real exchange rate appreciated on average 17 percent during whole sample, whereas LMIC exchange rates depreciated on average by 1 percent.
  - Imports/GDP: Ghana more open than peers; import-to-GDP rose from average 43 percent (late 1990s) to 64 percent (late 2000s).
  - Terms of Trade: Ghana benefited from positive TOT shocks during the last two decades.
  - Fiscal stance: Ghana had significantly worse fiscal accounts than peers, on average by around 7 percent of GDP; fiscal deficit declined from 12.6 percent in late 1990s to 6.6 percent in late 2000s; during 2005–2008 gap to resource-rich peers rose to 12 percent despite being a cocoa and gold exporter.
- Governance and structural indicators (Tables 9 and 10):
  - Ghana performs better than both resource-rich and diversified LMICs on governance indicators and has been improving over time.
  - Structural: no armed conflict (1996–2008); Ghana receives more aid inflows as ratio of GDP; old age dependency lower and population growth higher relative to peers.
- Growth projections for Ghana (based on Table 7 regression using Government Effectiveness as institutional variable and applying model to Ghana):
  - Forecasted growth: 2009: 4.5%; 2010: 4.7%; 2011: growth without oil 7.7% / growth with oil 5.8% (figure shows both lines); 2014: growth without oil 8.2% / growth with oil 6.3% (figure labels show growth with and without oil lines).
  - Text summary of decomposition:
    - Growth rate projected to increase from 4.5 percent in 2009 to 8.2 percent in 2014.
    - Two main drivers:
      1. Improvement in fiscal accounts: negative impact of fiscal deficit fades, yielding around 1.5 percent additional growth.
      2. Oil production with strong institutions: Ghana will benefit from an additional 2 percent of growth as an oil exporter.
  - Concluding applied result: Ghana will achieve an additional growth rate of 2 percent in the medium-term as an oil-exporting economy.

### Conclusion (paper-wide)
- Dataset: 150 low and middle income countries, 1973–2008, analyzed with static and dynamic panel techniques.
- Main findings:
  - There is a poverty trap for poor resource-rich countries driven by low institutional quality.
  - For countries with good governance and strong macroeconomic management, resource wealth (oil) can be utilized to achieve higher economic growth.
  - Institutional quality is crucial: improving institutions can eliminate the observed low-income resource-trap effect.
- Application: Based on results, Ghana is expected to gain an additional 2 percent medium-term growth as an oil-exporting economy, conditional on fiscal improvement and institutional quality.

*Italic: Source – Excerpts and tables from the provided PDF content unit.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11104.pdf_