## _wp12257 - Section V the estimation results for the full-sample case, whereas results for each income group

## Source details

**Canonical URL:** [_wp12257 - Section V the estimation results for the full-sample case, whereas results for each income group](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12257.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12257.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12257.pdf.json)

---

### Dataset and Scope
- Final dataset used in estimations: 69 countries with at least 20 years of observations for tax and macro variables: 21 high-income countries, 23 middle-income countries, 25 low-income countries.
- Main sources: IMF Government Finance Statistics (GFS) yearbook (CCG level), OECD Revenue Statistics (CGG for 28 countries), UN public finance statistics for seven Latin-American countries, World Development Indicators for macro variables, Barro and Lee (2010) for average years of schooling.
- Time coverage and raw counts from Annex I:
  - Time coverage: 1970 through 2010 (historical data available in both GFSM1986 and GFSM2001).
  - Countries in raw dataset: 149.
  - Countries with >20 years of data: 91.

### Methodology and Empirical Strategy
- Objective: estimate long-run equilibrium relation between tax composition and growth (rate of growth of GDP per capita).
- Main estimator: Pooled Mean Group (PMG) procedure (Pesaran et al, 1999); Mean Group (MG) and Dynamic Fixed Effects (DFE) also reported for comparison.
- Dependent variable: g = growth rate of GDP per capita (log difference).
- Tax variables:
  - Total tax revenue expressed as a share of GDP (control to impose revenue-neutral interpretation).
  - Tax-composition variables expressed as shares of total tax revenue.
  - Aggregations: aggregate income taxes; aggregate consumption and property taxes; disaggregation into PIT, CIT, SSC, VAT & Sales, Trade, Property, Other consumption taxes.
- Identification: omit one tax component at a time so coefficients interpret the effect of shifting one percentage point of total tax revenue from the omitted component to the included component.
- Controls: population growth (n), investment ratio (I), average years of schooling (h), country-specific linear time trend and intercept.
- Error-correction / ARDL(1,1) framework used to estimate long-run coefficients and short-run dynamics (Annex III).

### Key Findings — Full Sample (PMG long-run estimates)
- Aggregate effects (income taxes vs. consumption & property taxes omitted):
  - A percentage point increase in the income-tax share offset by a reduction in consumption and property taxes is associated with a decrease in long-run growth of GDP per capita by 0.07 percentage points.
  - Income taxes long-run coefficient: -0.065*** (0.017).
  - Consumption and property taxes long-run coefficient: 0.044*** (0.017).
- Disaggregated income taxes:
  - Personal Income taxes (PIT): -0.140*** (0.025).
  - Social Security Contributions (SSC): -0.169*** (0.027).
  - Corporate income taxes (CIT): -0.005 (0.023) — effect weak and specification-sensitive.
  - Column (2) summary: PIT and SSC coefficients negative and significant; SSC associated with slowdown of 0.17 percentage points per percentage point increase; PIT associated with slowdown of 0.14 percentage points per percentage point increase.
- Disaggregated consumption and property taxes:
  - VAT & Sales taxes: 0.103*** (0.027); a percentage point increase in VAT & sales taxes with reduction in income taxes associated with 0.1 percentage point increase in long-run growth.
  - Consumption taxes (aggregate): 0.027 (0.018).
  - Trade taxes: -0.029 (0.029) — no significant relation in full sample.
  - Property taxes: 0.242*** (0.056) and 0.248*** (0.061) in specified models — positive and robust association with growth.
- Overall tax burden:
  - Overall tax burden long-run coefficients across models include: -0.062** (0.027); -0.073*** (0.027); -0.076*** (0.027); -0.061** (0.027); -0.098*** (0.030).
- Macroeconomic controls (example long-run estimates from Table 1):
  - Physical capital: 0.053*** (0.019) in model (1); other estimates: 0.026 (0.024), 0.040** (0.020), 0.031 (0.021), 0.057** (0.023).
  - Human capital: 0.024** (0.011); other estimates: 0.039** (0.016), 0.023** (0.011), 0.032*** (0.011), 0.013 (0.012).
  - Population growth: -1.549*** (0.147); other estimates: -1.572*** (0.148), -1.600*** (0.150), -1.708*** (0.153), -1.791*** (0.168).
- Robustness across estimators (Table 2):
  - PMG, MG, and DFE give generally consistent signs and significance; MG estimates larger and less precise in many cases.
  - Examples (long-run coefficients):
    - Income taxes: PMG -0.065*** (0.017); MG -0.170* (0.087); DFE -0.038** (0.017).
    - PIT: PMG -0.140*** (0.025); MG -0.440 (0.420); DFE -0.046 (0.029).
    - SSC: PMG -0.169*** (0.027); MG -0.925*** (0.316); DFE -0.083** (0.033).
    - VAT & Sales taxes: PMG 0.103*** (0.027); MG 0.863* (0.470); DFE 0.056** (0.023).
  - Hausman tests: reported p-values very high (examples: chi2(5) = 1.330, p-value = 0.931; chi2(7) = 1.330, p-value = 0.931), supporting PMG choice.

### Findings by Income Group (HICs, MICs, LICs)
- HICs and MICs: patterns similar to full sample; effects often larger in magnitude.
  - Income-tax share and growth (Table 3):
    - HICs: Income taxes -0.149*** (0.031); Physical capital 0.087** (0.036); Human capital 0.063** (0.031); Population growth -1.750*** (0.191); Observations 778.
    - MICs: Income taxes -0.105*** (0.036); Physical capital 0.044 (0.036); Human capital 0.097*** (0.029); Population growth -0.021*** (0.004); Observations 665.
    - LICs: Income taxes 0.016 (0.028) — not significant; Physical capital 0.064* (0.033); Human capital 0.009 (0.015); Population growth -0.001 (0.003); Observations 649.
  - Disaggregated drivers in HICs and MICs:
    - Negative relation driven by SSC and PIT rather than CIT; PIT and SSC negative, significant and larger in absolute value than CIT.
- Consumption-and-property-tax share and growth (Table 4):
  - HICs:
    - Consumption and property taxes 0.143*** (0.031); Property taxes 0.275*** (0.066); VAT & Sales taxes 0.218*** (0.045).
    - VAT & sales taxes: a percentage point shift from income taxes to VAT & sales taxes boosts growth by about 0.2 percentage points.
    - Wald test rejects equality of consumption and property tax coefficients (p-value = 0.015).
  - MICs:
    - Consumption and property taxes positive but non-significant at aggregate level.
    - VAT & Sales taxes 0.082* (0.043); Property taxes 0.419*** (0.161).
    - Trade taxes negative and significant: -0.230*** (0.070).
  - LICs:
    - Generally no significant associations between growth and most tax-structure variables.
    - Property taxes 0.138 (0.131) — not significant.
- After excluding countries with potential endogeneity concerns (Tables 7 and 8), main group results remain essentially unchanged:
  - Table 7 examples (post-exclusion, income-tax share):
    - HICs: Income taxes -0.119*** (0.036); Human capital 0.109*** (0.035); Population growth -1.663*** (0.223).
    - MICs: Income taxes -0.089** (0.036); Human capital 0.085*** (0.029); Population growth -2.131*** (0.359).
    - LICs: Income taxes 0.028 (0.034) — not significant.
  - Table 8 examples (post-exclusion, consumption-and-property share):
    - HICs: Consumption and property taxes 0.143*** (0.031); Property taxes 0.278*** (0.067); VAT & Sales taxes 0.257*** (0.057) in specified models.
    - MICs: Consumption taxes 0.103*** (0.033); Property taxes 0.427** (0.173); in some specifications Property taxes 1.091*** (0.304).
    - LICs: Consumption and property taxes -0.046 (0.031) — not significant; Property taxes 0.113 (0.172).

### Endogeneity Checks and Robustness
- Weak exogeneity testing (country-by-country, following Calderon et al (2011) in PMG context):
  - Test counts where error-correction term significant at 5% reported in Table 5.
  - Full-sample endogeneity counts (selected examples):
    - Income taxes Endo. # 14; Total # 69.
    - PIT vs. SSC Endo. # 18; Total # 47.
    - Trade vs. Prop. Endo. # 23; Total # 49.
  - HICs, MICs, LICs counts reported separately (see Table 5 for full breakdown).
- Robustness exercise:
  - Exclude countries with potential endogeneity and re-estimate (Table 6 and Tables 7–8 for groups).
  - Main results remain similar after exclusion:
    - Income taxes negative and significant (aggregate).
    - PIT and SSC remain negative and significant; CIT remains insignificant.
    - VAT & Sales taxes positive and significant.
    - Sizes slightly smaller in absolute value; standard errors larger due to loss of observations (e.g., loss of over 400 and 200 observations in some estimations).
- Limitations and further work:
  - Exclusion approach helpful but not definitive.
  - Authors plan further work using system GMM and IV methods to address endogeneity more comprehensively.

### Policy Implications and Conclusions
- Main empirical conclusions:
  - A shift from consumption and property taxes to income taxes is negatively associated with long-run growth.
  - SSC and PIT shares exhibit stronger negative association with growth than CIT; results robust across data sources and specifications.
  - Shifts from income to property taxes show stronger positive association with growth than shifts to consumption taxes.
  - Within consumption taxes, VAT & sales taxes robustly show a positive association with growth.
  - Results hold consistently for HICs and MICs; little evidence of significant effects in LICs.
- Policy-relevant implications reported:
  - Revenue-neutral tax shifts favoring consumption/property taxes (and within consumption, VAT & sales taxes) over income taxes—particularly SSC and PIT—are associated with higher long-run per capita growth in HICs and MICs.
  - For LICs, weak empirical associations may reflect administrative/enforcement constraints rather than absence of economic effects; caution warranted.
- Robustness to endogeneity checks:
  - After excluding countries that may violate weak exogeneity, main results remain essentially unchanged.

### Selected Key Numeric Results and Statistics (preserved exactly)
- Full-sample observations examples: 2,092; 1,478; 1,872; 1,504 (across models).
- Table 1 tax-structure long-run effects:
  - Income taxes: -0.065*** (0.017).
  - Personal Income taxes: -0.140*** (0.025).
  - Social Security Contributions: -0.169*** (0.027).
  - Corporate income taxes: -0.005 (0.023).
  - Consumption and property taxes: 0.044*** (0.017).
  - Consumption taxes: 0.027 (0.018).
  - VAT & Sales taxes: 0.103*** (0.027).
  - Trade taxes: -0.029 (0.029).
  - Other consumption taxes: 0.054* (0.030).
  - Property taxes: 0.242*** (0.056) and 0.248*** (0.061).
- Table 2 estimator comparisons examples:
  - Income taxes: PMG -0.065*** (0.017); MG -0.170* (0.087); DFE -0.038** (0.017).
  - SSC: PMG -0.169*** (0.027); MG -0.925*** (0.316); DFE -0.083** (0.033).
  - VAT & Sales taxes: PMG 0.103*** (0.027); MG 0.863* (0.470); DFE 0.056** (0.023).
- Table 3 group examples:
  - HICs Income taxes: -0.149*** (0.031); Observations 778.
  - MICs Income taxes: -0.105*** (0.036); Observations 665.
  - LICs Income taxes: 0.016 (0.028); Observations 649.
- Table 4 group examples (consumption/property):
  - HICs Consumption and property taxes: 0.143*** (0.031); Property taxes 0.275*** (0.066); VAT & Sales taxes 0.218*** (0.045).
  - MICs VAT & Sales taxes 0.082* (0.043); Property taxes 0.419*** (0.161); Trade taxes -0.230*** (0.070).
- Table 6 post-exclusion full-sample tax structure effects (examples):
  - Income taxes: -0.046** (0.021).
  - Personal Income taxes: -0.168*** (0.038).
  - Social Security Contributions: -0.173*** (0.046).
  - Consumption and property taxes: 0.036* (0.019).
  - VAT & Sales taxes: 0.169*** (0.042).
  - Property taxes: 0.201*** (0.059) and 0.293*** (0.091).
- Annex V summary-statistics examples:
  - Total tax revenue means:
    - LICs: Mean 14.64; Std.Dev. 5.31; Min 2.50; Max 42.04; Corr. w/ GDP p.c. 0.35***.
    - MICs: Mean 20.11; Std.Dev. 9.11; Min 4.38; Max 54.63; Corr. w/ GDP p.c. 0.55***.
    - HICs: Mean 34.75; Std.Dev. 7.84; Min 15.92; Max 52.26; Corr. w/ GDP p.c. 0.36***.
  - log(GDP per capita, 2000 constant USD prices) means:
    - LICs: Mean 7.19; Std.Dev. 0.75; Min 5.39; Max 8.67.
    - MICs: Mean 8.93; Std.Dev. 0.86; Min 6.97; Max 11.33.
    - HICs: Mean 10.21; Std.Dev. 0.41; Min 8.87; Max 11.06.
  - GDP growth rate (percent) means:
    - LICs: Mean 0.01; Std.Dev. 0.04; Min -0.21; Max 0.19.
    - MICs: Mean 0.02; Std.Dev. 0.04; Min -0.18; Max 0.20.
    - HICs: Mean 0.02; Std.Dev. 0.02; Min -0.09; Max 0.11.

*Source — _wp12257 (Section V and surrounding sections) — estimation results for the full-sample case and income-group analyses.*

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

### _wp12257 - References

### Dataset and Scope
- Main data sources: Government Finance Statistics (GFS) yearbook of the IMF, Revenue Statistics database of the OECD, and public finance statistics of the UN.
- Final dataset comprises a total of 69 countries covering the period 1970-2009, in which each country has at least 20 years of observations on total tax revenue.
- GFS is the main source; OECD or UN data are used when they have better coverage than GFS.
- Annexes describe:
  - I. Construction of the GFS Dataset
  - II. Definition of Tax Variables
  - III. The Underlying Error Correction Model
  - IV. Regression Analysis Considering the Output Level
  - V. Summary Statistics

### Research Question and Methodology
- Objective: Investigate the relation between tax structure and long-run growth using a comprehensive cross-country dataset for the longest possible time period.
- Estimation method: Pooled Mean Group (PMG) procedure as proposed by Pesaran et al (1999) to estimate the long-run equilibrium relation between tax composition and growth.
- Contrast with prior work:
  - Unlike Arnold et al (2011), which studies effects of tax composition on the long-run GDP per capita level, this paper studies effects on the rate of growth of GDP per capita.
  - Shares objectives with Kneller et al (1999) and Gemmell et al (2011) but uses a broader dataset and more detailed tax-variable analysis.
- Theoretical background references include Barro (1990); King and Rebelo (1990); Jones et al (1993); Mendoza et al (1997); Easterly and Rebelo (1993); Kneller et al (1999); Gemmell et al (2011); Arnold et al (2011).

### Key Findings
- Full-sample results:
  - Raising taxes on income while reducing consumption and property taxes, keeping the overall tax burden unchanged, is negatively associated with growth.
  - The negative association is more significant for social security contributions and personal income taxes relative to corporate income taxes.
  - A shift from income to property taxes has a robust and positive association with growth.
  - An increase in VAT and sales taxes compensated with a reduction in income taxes also shows a positive effect on growth.
- Results by income group:
  - High-income countries (HICs) and middle-income countries (MICs) show similar and consistent results to the full-sample case.
  - Low-income countries (LICs) do not show such robust results; the paper tentatively suggests this could be due to poorer quality of tax administration and tax enforcement in LICs, which may be reflected in the tax policy data being relatively more "disconnected" in their relation with growth.

### Empirical Content Inventory (Tables and Figures)
- Tables listed (titles only):
  - Table 1. Estimation Results, Full Sample
  - Table 2. Comparing Estimation Methods: PMG, MG, and DFE Estimates
  - Table 3. Estimation Results, The Income-Tax Share and Growth, HICs, MICs, and LICs
  - Table 4. Estimation Results, The Consumption-and-Property-Tax Share and Growth, HICs, MICs, and LICs
  - Table 5. Test of Weak Exogeneity, The Number of Countries with Potential Endogeneity Problem
  - Table 6. Estimation Results, Full Sample, After Excluding Countries with
  - Table 7. Estimation Results, The Income-Tax Share, HICs, MICs, and LICs, After Excluding Countries with Potential Endogeneity Problem
  - Table 8. Estimation Results, The Consumption-and-Property-Tax Share, HICs, MICs, LICs, After Excluding Countries with Potential Endogeneity Problem
- Figures listed (titles only):
  - Figure 1. Tax Revenue and Income Levels
  - Figure 2. Tax Revenue and Income Levels: Disaggregated Analysis
  - Figure 3. Long-Run Trends in Total Tax Revenue
  - Figure 4. Trends in Tax Revenue: Disaggregated Analysis
  - Figure 5. Trends in Tax Composition
- Annex Tables listed:
  - Annex Table 1. Tax Composition and Income Level, in Comparison with Arnold et al (2011)
  - Annex Table 2. Summary Statistics of Tax Variables
  - Annex Table 3. Summary Statistics of Other Variables

### Structure of the Paper (as provided)
- Section I: Introduction
- Section II: Dataset (described; see dataset and sources above)
- Section III: Stylized facts on relation between tax structures and level of development
- Section IV: Estimation method
- Subsequent sections: Report results and robustness analyses (tables and annexes detail estimation results and summary statistics)

*Source: _wp12257 - References (IMF PDF content unit).*

### Section V the estimation results for the full-sample case, whereas results for each income group

### _wp12257 - Section V the estimation results for the full-sample case, whereas results for each income group

### II. THE DATASET
- Main source: IMF’s Government Finance Statistics (GFS) yearbook; methodological change from GFSM1986 to GFSM2001 required mapping guidelines (see Wickens, 2002).
- Constructed panel: 80 countries from the GFS yearbook with at least 20 years of observations at the consolidated central government level (CCG) for the period 1970-2009.
- Supplementary sources:
  - Revenue Statistics database of the OECD for 28 countries to include consolidated general government (CGG) level for 1970-2009.
  - Economic Commission for Latin America and the Caribbean of the UN for seven Latin-American countries (CGG level for Argentina and Brazil; CCG for remaining countries); UN tax-revenue data for Latin America covers 1990-2009.
  - World Development Indicators (World Bank) for remaining macro variables; average years of schooling from Barro and Lee (2010).
- Sample grouping: divided into three groups by GDP per capita (PPP prices): low, middle, high income.
- Final dataset used in estimations: 69 countries with at least 20 years of observations for tax and macro variables: 21 high-income countries, 23 middle-income countries, 25 low-income countries.
- Notes on classification:
  - High income countries defined as 21 OECD economies (list provided in source).
  - For middle/low income classification, countries were compared to the cross-sectional median each year and allocated based on the number of times above or below the median.

### III. TAX STRUCTURE AND DEVELOPMENT — descriptive facts
- Aggregate patterns (decile medians plotted in Figures 1–2):
  - Overall tax-to-GDP increases with GDP per capita (Wagner’s law).
  - Tax structure shifts toward direct taxation as countries develop.
- Disaggregated facts:
  - Corporate income taxes (CIT) increase with income, but less than personal income taxes (PIT) or social security contributions (SSC).
  - For countries in the top three deciles: PIT ≈ 9.7 percent of GDP; SSC ≈ 8.8 percent of GDP.
  - For countries in the three lowest deciles: PIT ≈ 1.3 percent of GDP; SSC ≈ 0.7 percent of GDP.
  - VAT and sales taxes increase with income; differences across income groups are less pronounced than for PIT or SSC.
  - Trade taxes significantly decrease with development.
  - Property tax revenue trends upward with income but remains modest even for the highest income decile.
- Long-run trends (1970-2009 medians by decade, Figure 3):
  - LICs: taxes roughly 13 percent of GDP (about one-third of HICs).
  - HICs: taxes over 35 percent of GDP.
- Tax composition (Figure 5):
  - HICs rely much more on direct taxes: income taxes account for between a half and two-thirds of total tax revenue in MICs and HICs, respectively; only about one-third in LICs.
  - Components shown as shares of total tax revenue: SSC, PIT, CIT, VAT & Sales, Trade, Property, Other consumption.

### IV. EMPIRICAL STRATEGY
- Tax aggregation and decomposition:
  - Total tax revenue split into two broad categories: aggregate income taxes and aggregate consumption and property taxes.
  - Further decomposition into PIT, CIT, SSC, VAT & sales taxes, trade taxes, property taxes, other consumption taxes.
- Sample restrictions:
  - Only countries with at least 20 years of observations on total tax revenue included.
  - If aggregate income taxes available >20 years but sub-components not, country included for aggregate income-tax estimations but excluded from sub-component specifications.
- Control variables: population growth (n), investment ratio (I), average years of schooling (h) as proxy for human capital; country-specific linear time trend and intercept included.
- Econometric specification:
  - Error correction form estimating GDP per capita growth (g, log difference of GDP per capita).
  - Tax-composition variables (TC) expressed as share of total tax revenue.
  - Total tax revenue as share of GDP included as control to impose revenue neutrality interpretation for composition shifts.
- Estimation method:
  - Pooled Mean Group (PMG) estimator (Pesaran et al, 1999) as primary method: allows heterogeneous short-run dynamics and homogeneous long-run coefficients.
  - For completeness, Mean Group (MG) and Dynamic Fixed Effects (DFE) estimates also reported.
- Identification strategy for composition effects:
  - Omit one tax component at a time so coefficient interpretable as effect of shifting one percentage point of total tax revenue from omitted component to included component (revenue-neutral shift).

### V. TAX COMPOSITION AND GROWTH I: FULL SAMPLE (PMG long-run estimates)
- Aggregate income taxes (income taxes vs. consumption & property taxes omitted):
  - Column (1): a percentage point increase in the income-tax share offset by a reduction in consumption and property taxes is associated with a decrease in long-run growth of GDP per capita by 0.07 percentage points.
- Disaggregated income taxes (PIT, CIT, SSC):
  - Column (2): coefficients for PIT, CIT, SSC are negative; PIT and SSC coefficients significant.
  - A percentage point increase in SSC associated with slowdown in growth of 0.17 percentage points.
  - A percentage point increase in PIT associated with slowdown in growth of 0.14 percentage points.
  - CIT effect weaker and specification-sensitive.
- Aggregate consumption and property taxes (income taxes omitted):
  - Column (3): coefficient on aggregate consumption and property taxes positive and significant; a shift from income taxes to consumption and property taxes by one percentage point associated with increase in long-run per capita growth of 0.04-0.07 percentage points.
- Property taxes vs consumption taxes:
  - Column (5): increasing property-tax share while reducing income taxes associated with faster growth.
  - Aggregate consumption taxes effect weaker at aggregate level.
- Disaggregation of consumption taxes:
  - VAT and sales taxes (combined) coefficient positive and significant: a percentage point increase in VAT & sales taxes with reduction in income taxes associated with 0.1 percentage point increase in long-run growth.
  - No significant relation between growth and trade taxes in full sample.
- Model selection and robustness:
  - Hausman tests comparing MG and PMG suggest inability to reject homogeneity of long-run coefficients; p-values very high for all specifications — PMG preferred.
  - MG and DFE estimates generally consistent in sign and significance with PMG, though MG estimates larger and less precise.

### VI. TAX COMPOSITION AND GROWTH II: HIGH, MIDDLE AND LOW-INCOME COUNTRIES
- HICs and MICs (Table 3):
  - Negative and significant coefficients on income taxes for both HICs and MICs: raising income taxes while reducing consumption and property taxes is growth-reducing.
  - Effect size larger than full sample: a percentage point increase in income taxes associated with slowdown in growth by about 0.1 percentage points.
  - Negative relation driven by SSC and PIT rather than CIT; PIT and SSC coefficients negative, significant and larger in absolute value than CIT (CIT tends to be non-significant).
- HICs (Table 4 and disaggregation):
  - Positive and significant coefficient on consumption and property taxes when shifting from income taxes.
  - When separating consumption and property taxes: property taxes have stronger association with growth than consumption taxes.
  - VAT & sales taxes: a percentage point shift from income taxes to VAT & sales taxes boosts growth by about 0.2 percentage points over the long run.
  - Wald test rejects equality of consumption and property tax coefficients (p-value = 0.015).
- MICs:
  - Coefficient on consumption and property taxes positive but non-significant at aggregate level.
  - When separating property taxes from consumption taxes, property taxes show a strong relation with growth.
  - Detailed decomposition shows VAT & sales taxes positive and significant, trade taxes negative and significant — opposite signs may render aggregate consumption-tax effect ambiguous.
- LICs:
  - Generally no significant associations found between growth and most tax-structure variables in this subsample.
  - Tentative explanation: poorer tax administration and enforcement reduce measured empirical associations.

### VII. ENDOGENEITY CHECKS
- Concern: simultaneity between growth and tax variables (Wagner’s law) could bias causal interpretation.
- Approach:
  - Use country-by-country tests for weak exogeneity following Calderon et al (2011) in a PMG context.
  - Estimate marginal systems where the error correction term coefficients (δ_i) indicate whether tax variables react to deviations from long-run equilibrium (equation (2)); estimated by seemingly unrelated regression (Zellner, 1962).
  - Null hypothesis: coefficients on error correction terms jointly zero (Wald test at 5 percent). Rejection implies violation of weak exogeneity for that country.
- Robustness exercise:
  - Exclude countries that appear to violate weak exogeneity and re-estimate.
  - Main results remain similar after exclusion (Table 6):
    - Income-tax coefficient negative and significant (aggregate).
    - Consumption & property tax coefficient positive and significant (aggregate).
    - Sizes slightly smaller in absolute value; standard errors larger due to loss of over 400 and 200 observations respectively in estimations.
  - Granular decomposition after exclusion:
    - PIT and SSC negative and significant; CIT remains insignificant.
    - VAT & sales taxes positive and significant.
  - Similar robustness for country-group regressions (Tables 7 and 8): main results essentially unchanged after excluding problematic countries.
- Limitations acknowledged:
  - Exclusion approach useful but not definitive; authors plan further work using system GMM and IV methods to address endogeneity more comprehensively.

### VIII. CONCLUDING REMARKS — main conclusions and implications
- Study contribution:
  - Constructed a comprehensive dataset covering 69 countries with wide income range.
  - Estimated long-run effects of changes in tax composition on GDP per capita growth rates (using PMG and related methods).
- Key empirical findings:
  - A shift from consumption and property taxes to income taxes is negatively associated with long-run growth.
  - SSC and PIT shares exhibit stronger negative association with growth than CIT; results robust across data sources and specifications.
  - Shifts from income to property taxes show stronger positive association with growth than shifts to consumption taxes.
  - Within consumption taxes, VAT and sales taxes robustly show a positive association with growth.
  - Results hold consistently for high- and middle-income countries; little evidence of significant effects in low-income countries.
- Robustness to endogeneity checks:
  - After excluding countries that may violate weak exogeneity, main results remain essentially unchanged.
- Policy and research implications:
  - Revenue-neutral tax shifts favoring consumption/property taxes (and within consumption, VAT & sales taxes) over income taxes—particularly SSC and PIT—are associated with higher long-run per capita growth in HICs and MICs.
  - Caution for LICs: weak empirical association may reflect administrative/enforcement constraints rather than absence of economic effects.
  - Further research planned to address endogeneity more thoroughly using system GMM and IV methods.

*Italic: Source — _wp12257 (Section V and surrounding sections) — estimation results for the full-sample case and income-group analyses.*

### REFERENCES

### _wp12257 - REFERENCES

### References (selected entries from the source)
- Arnold, Jens, B. Brys, Ch. Heady, Å. Johansson, C. Schwellnus and L. Vartia, 2011, “Tax Policy For Economic Recovery and Growth,” The Economic Journal, 121, pp. F59-F80.
- Arnold, Jens, A. Bassanini and S. Scarpetta, 2007, “Solow or Lucas?: Teseting Growth Models Using Panel Data from OECD Countries,” OECD Economics Department Working Paper No. 592.
- Barro, Robert, 1990, “Government Spending in a Simple Model of Endogenous Growth,” Journal of Political Economy, 98:5, pp. S103–S125.
- Barro, Robert and J.-W. Lee, 2010, ”A New Data Set of Educational Attainment in the World, 1950-2010,” NBER Working Paper No. 15902.
- Baunsgaard, Thomas and M. Keen, 2010, “Tax Revenue and (or?) Trade Liberalization,” Journal of Public Economics, 94, pp. 563-577.
- Boswijk, H. Peter, 1995, “Efficient Inference on Cointegration Parameters in Structural Error Correction Models,” Journal of Econometrics 69, 133-158.
- Calderon, Cesar, E. Moral-Benito and L. Serven, 2011, “Is infrastructure capital productive? a dynamic heterogeneous approach,” Policy Research Working Paper Series 5682, The World Bank.
- Easterly, William and S. Rebelo, 1993, “Fiscal Policy and Economic Growth: An Empirical Investigation,” Journal of Monetary Economics, 32, 417-458.
- Gemmell, Norman, R. Kneller, and I. Sanz, 2011, “The Timing and Persistence of Fiscal Policy Impacts on Growth: Evidence from OECD Countries,” The Economic Journal, 121, pp. F33-F58.
- Gemmell, Norman, R. Kneller, and I. Sanz, 2012, “The Composition of Public Expenditure and Economic Growth,” Unpublished Manuscript.
- IMF, 2011, “Revenue Mobilization in Developing Countries,” IMF Policy Paper Series (Washington).
- Johansen, Soren, 1992, “Cointegration in Partial Systems and the Efficiency of Single-equation Analysis,” Journal of Econometrics, 52, 389-402.
- Jones, Larry, R. Manuelli and P. Rossi, 1993, “Optimal Taxation in Models of Endogenous Growth,” Journal of Political Economy, 101:3, pp. 485–517.
- Keele, Luke J. and S. DeBoef, 2008, "Taking Time Seriously: Dynamic Regression," American Journal of Political Science, 52:1, pp. 184-200.
- King, Robert and S. Rebelo, 1990, “Public Policy and Economic Growth: Developing Neoclassical Implications,” Journal of Political Economy, 98:5, pp. S126–S150.
- Kneller, Richard, M. Bleaney and N. Gemmell, 1999, “Fiscal Policy and Growth: Evidence from OECD Countries,” Journal of Public Economics, 74, pp. 171-190.
- Mendoza, Enrique, Milesi-Ferretti, G., and Asea, P, 1997, “On the Effectiveness of Tax Policy in Altering Long-Run Growth: Harberger’s Superneutrality Conjecture,” Journal of Public Economics, 66, pp. 99-126.
- Pesaran, M. Hashem, Y. Shin, and R. Smith, 1999, “Pooled Mean Group Estimation of Dynamic Heterogeneous Panels,” Journal of the American Statistical Association, 94:446, pp. 621-634.
- Ram, Rati, 1987, “Wagner’s Hypothesis in Time-Series and Cross-Section Perspectives: Evidence from “Real” Data for 115 Countries,” The Review of Economics and Statistics, 69, 2, pp. 194-204.
- Wickens, Tobias, 2002, “Classification of GFSM1986 Data to the GFSM2001 Framework,” IMF Government Finance Statistics Manual 2001 Companion Material (Washington).
- Xing, Jing, 2012, “Tax Structure and Growth: How Robust is the Empirical Evidence?,” Economics Letters, 117:1, pp. 379-382.
- Zellner, Arnold, 1962, “An Efficient Method of Estimating Seemingly Unrelated Regressions and Tests for Aggregation Bias,” Journal of the American Statistical Association, 57, pp. 348-368.

### Key empirical findings and long-run coefficient estimates (from tables)
- Table 1 (Full Sample, long-run coefficients; standard errors in parentheses; significance: * 10%; ** 5%; *** 1%):
  - Physical capital: 0.053*** (0.019) in model (1); other model estimates: 0.026 (0.024), 0.040** (0.020), 0.031 (0.021), 0.057** (0.023).
  - Human capital: 0.024** (0.011); other estimates include 0.039** (0.016), 0.023** (0.011), 0.032*** (0.011), 0.013 (0.012).
  - Population growth: -1.549*** (0.147); other estimates: -1.572*** (0.148), -1.600*** (0.150), -1.708*** (0.153), -1.791*** (0.168).
  - Overall tax burden: -0.062** (0.027); other estimates: -0.073*** (0.027), -0.076*** (0.027), -0.061** (0.027), -0.098*** (0.030).
  - Tax structure long-run effects:
    - Income taxes: -0.065*** (0.017).
    - Personal Income taxes: -0.140*** (0.025).
    - Social Security Contributions: -0.169*** (0.027).
    - Corporate income taxes: -0.005 (0.023).
    - Consumption and property taxes: 0.044*** (0.017).
    - Consumption taxes: 0.027 (0.018).
    - VAT & Sales taxes: 0.103*** (0.027).
    - Trade taxes: -0.029 (0.029).
    - Other consumption taxes: 0.054* (0.030).
    - Property taxes: 0.242*** (0.056) and 0.248*** (0.061) in specified models.
  - Observations: 2,092; 1,478; 2,092; 1,872; 1,504 (across models).
  - Omitted tax variable labels: Consumption and property taxes; Income taxes (depending on specification).
  - Note: Overall tax revenue is expressed as a share of GDP, and tax structure variables are expressed as a share of total tax burden. All equations include short-run dynamics, but only the long-run coefficient estimates are reported.

- Table 2 (Comparison of estimation methods: PMG, MG, and DFE; long-run coefficients reported with S.E.):
  - Income taxes: PMG -0.065*** (0.017); MG -0.170* (0.087); DFE -0.038** (0.017).
  - PIT: PMG -0.140*** (0.025); MG -0.440 (0.420); DFE -0.046 (0.029).
  - SSC: PMG -0.169*** (0.027); MG -0.925*** (0.316); DFE -0.083** (0.033).
  - CIT: PMG -0.005 (0.023); MG 0.034 (0.109); DFE -0.054** (0.025).
  - Consumption and property taxes: PMG 0.044*** (0.017); MG 0.170** (0.086); DFE 0.037** (0.017).
  - Consumption taxes: PMG 0.027 (0.018); MG 0.360* (0.193); DFE 0.038** (0.018).
  - VAT & Sales taxes: PMG 0.103*** (0.027); MG 0.863* (0.470); DFE 0.056** (0.023).
  - Trade taxes: PMG -0.029 (0.029); MG -0.728 (0.899); DFE 0.025 (0.023).
  - Other consumption taxes: PMG 0.054* (0.030); MG -0.271 (0.717); DFE 0.001 (0.027).
  - Property taxes: PMG 0.248*** (0.061); MG 3.103 (1.908); DFE 0.068 (0.066).
  - Hausman test results reported repeatedly: chi2(5) = 1.330, p-value = 0.931; chi2(7) = 1.330, p-value = 0.931; chi2(6)= 1.330, p-value = 0.931; chi2(8) = 1.330, p-value = 0.931.

- Table 3 (Income-tax share and growth by country group; long-run coefficients):
  - HICs (column 1): Physical capital 0.087** (0.036); Human capital 0.063** (0.031); Population growth -1.750*** (0.191); Overall tax revenue -0.031 (0.035); Income taxes -0.149*** (0.031).
  - MICs (column 2): Physical capital 0.044 (0.036); Human capital 0.097*** (0.029); Population growth -0.021*** (0.004); Overall tax revenue -0.076 (0.063); Income taxes -0.105*** (0.036).
  - LICs (column 3): Physical capital 0.064* (0.033); Human capital 0.009 (0.015); Population growth -0.001 (0.003); Overall tax revenue -0.057 (0.065); Income taxes 0.016 (0.028).
  - PIT / SSC / CIT columns report subgroup coefficients (selected): PIT in HICs -0.197*** (0.033); SSC in MICs -0.199*** (0.034); CIT in MICs 0.009 (0.048).
  - Observations: 778 (HICs), 665 (MICs), 649 (LICs) for columns (1)-(3); 722, 478, 278 for columns (4)-(6).
  - Omitted tax variable: Consumption and property taxes.

- Table 4 (Consumption-and-property-tax share and growth by country group; long-run coefficients):
  - HICs (col 1): Physical capital 0.087** (0.036); Human capital 0.061** (0.031); Population growth -1.742*** (0.192); Overall tax revenue -0.034 (0.035); Consumption and property taxes 0.143*** (0.031); Property taxes 0.275*** (0.066).
  - MICs (col 2): Physical capital 0.018 (0.040); Human capital 0.087*** (0.030); Population growth -0.025*** (0.004); Overall tax revenue -0.140** (0.064); VAT& Sales taxes 0.082* (0.043); Property taxes 0.419*** (0.161).
  - LICs (col 3): Physical capital 0.044 (0.034); Human capital 0.010 (0.016); Population growth -0.001 (0.003); Overall tax revenue -0.013 (0.069); Property taxes 0.138 (0.131).
  - VAT & Sales taxes: 0.218*** (0.045) in HICs; Trade taxes significant and negative for MICs and LICs: -0.230*** (0.070) and -0.107*** (0.039) respectively.
  - Observations vary by specification (examples): 778, 665, 649; 778, 610, 484; 698, 459, 347.
  - Omitted tax variable: Income taxes.

- Table 5 (Test of Weak Exogeneity — number of countries with potential endogeneity problem; counts where the error correction term coefficient is significant at 5% in marginal models):
  - Full-sample Endogeneity counts:
    - Income taxes Endo. # 14; Total # 69.
    - PIT vs. SSC Endo. # 18; Total # 47.
    - Con. and prop. vs. CIT Endo. # 9; Total # 69.
    - Con. VAT & Sales vs. prop. Endo. # 10; Total # 60.
    - Trade vs. Prop. Endo. # 23; Total # 49.
  - HICs:
    - Income taxes Endo. # 2; Total # 21.
    - PIT vs. SSC Endo. # 4; Total # 20.
    - Con. and prop. vs. CIT Endo. # 0; Total # 21.
    - Con. VAT & Sales vs. prop. Endo. # 1; Total # 21.
    - Trade vs. Prop. Endo. # 10; Total # 21.
  - MICs:
    - Income taxes Endo. # 1; Total # 23.
    - PIT vs. SSC Endo. # 4; Total # 17.
    - Con. and prop. vs. CIT Endo. # 2; Total # 23.
    - Con. VAT & Sales vs. prop. Endo. # 2; Total # 21.
    - Trade vs. Prop. Endo. # 6; Total # 16.
  - LICs:
    - Income taxes Endo. # 3; Total # 25.
    - PIT vs. SSC Endo. # 4; Total # 10.
    - Con. and prop. vs. CIT Endo. # 4; Total # 25.
    - Con. VAT & Sales vs. prop. Endo. # 6; Total # 18.
    - Trade vs. Prop. Endo. # 6; Total # 12.
  - Note: First row of each category reports the number of countries in which the coefficient on error correction term is statistically significant at 5 percent level in marginal models.

- Table 6 (Full sample after excluding countries with potential endogeneity problem; long-run coefficients):
  - Physical capital: 0.028 (0.023); 0.041 (0.035); 0.029 (0.024); 0.008 (0.025); 0.078** (0.035).
  - Human capital: 0.027** (0.012); 0.067*** (0.024); 0.025** (0.012); 0.001 (0.016); -0.033 (0.023).
  - Population growth: -1.357*** (0.190); -1.660*** (0.244); -1.508*** (0.161); -1.609*** (0.161); -1.695*** (0.228).
  - Overall tax burden: -0.034 (0.033); -0.040 (0.040); -0.067** (0.029); -0.068** (0.030); -0.044 (0.034).
  - Tax structure long-run effects (post-exclusion):
    - Income taxes: -0.046** (0.021).
    - Personal Income taxes: -0.168*** (0.038).
    - Social Security Contributions: -0.173*** (0.046).
    - Corporate income taxes: -0.023 (0.033).
    - Consumption and property taxes: 0.036* (0.019).
    - VAT & Sales taxes: 0.169*** (0.042).
    - Trade taxes: 0.005 (0.039).
    - Other consumption taxes: 0.048 (0.042).
    - Property taxes: 0.201*** (0.059) and 0.293*** (0.091).
  - Observations: 1,628; 817; 1,813; 1,607; 796 (across models).
  - Omitted tax variable labels: Consumption and property taxes; Income taxes (depending on specification).

- Tables 7 and 8 (Results by income group after excluding countries with potential endogeneity problem):
  - Table 7 (Income-tax share by HICs, MICs, LICs after exclusions):
    - HICs (col 1): Physical capital 0.033 (0.040); Human capital 0.109*** (0.035); Population growth -1.663*** (0.223); Overall tax revenue -0.037 (0.043); Income taxes -0.119*** (0.036).
    - MICs (col 2): Physical capital 0.045 (0.036); Human capital 0.085*** (0.029); Population growth -2.131*** (0.359); Overall tax revenue -0.105* (0.063); Income taxes -0.089** (0.036).
    - LICs (col 3): Physical capital 0.033 (0.038); Human capital 0.002 (0.016); Population growth -0.053 (0.341); Overall tax revenue -0.005 (0.069); Income taxes 0.028 (0.034).
    - PIT / SSC / CIT subgroup notes: PIT in HICs -0.207*** (0.034); PIT in MICs -0.327*** (0.058); SSC in HICs -0.221*** (0.035); SSC in MICs -0.180** (0.077); CIT in MICs -0.084** (0.039).
    - Observations examples: 702, 647, 560 (columns 1-3); 608, 357, 154 (columns 4-6).
    - Omitted tax variable: Consumption and property taxes.
  - Table 8 (Consumption-and-property-tax share by HICs, MICs, LICs after exclusions):
    - HICs (col 1): Physical capital 0.087** (0.036); Human capital 0.061** (0.031); Population growth -1.742*** (0.192); Overall tax revenue -0.034 (0.035); Consumption and property taxes 0.143*** (0.031); Property taxes 0.278*** (0.067).
    - MICs (col 2): Physical capital 0.014 (0.040); Human capital 0.073** (0.030); Population growth -0.024*** (0.004); Overall tax revenue -0.168*** (0.064); Consumption taxes 0.103*** (0.033); Property taxes 0.427** (0.173).
    - LICs (col 3): Physical capital 0.008 (0.039); Human capital 0.002 (0.017); Population growth 0.000 (0.003); Overall tax revenue 0.045 (0.072); Consumption and property taxes -0.046 (0.031); Property taxes 0.113 (0.172).
    - VAT & Sales taxes notable in HICs: 0.257*** (0.057) in specified models; Property taxes sometimes large and significant (e.g., HICs 0.300*** (0.092); MICs 1.091*** (0.304) in certain specifications).
    - Observations vary by specification (examples): 778, 629, 535; 740, 564, 304; 365, 317, 155.
    - Omitted tax variable: Income taxes.

### Interpretation-related notes included in the source tables
- Significance notation: *Significant at 10% level; ** at 5% level; *** at 1% level.
- Standard errors are reported in brackets next to coefficient estimates.
- Overall tax revenue is expressed as a share of GDP; tax structure variables are expressed as a share of total tax burden.
- All equations include short-run dynamics, but tables report only long-run coefficient estimates.

*Content derived directly from the source PDF "_wp12257 - REFERENCES".*

### ANNEX I. CONSTRUCTION OF THE GFS DATASET

### ANNEX I. CONSTRUCTION OF THE GFS DATASET

### Dataset compilation and scope
- Historical data available in both GFSM1986 and GFSM2001 methodologies starting in 1970 through 2010 were retrieved for expenditure and revenue for all countries reporting to the IMF’s GFS yearbook during that period.
- In raw form, the dataset covers 149 countries that reported at least five years of data on total revenue and expenditure for the whole 1970-2010 period.
- The number of countries decreases to 91 when considering only those that reported data for more than 20 years.
- Although information is generally available from 1972 onwards, for advanced economies different key variables have been reported starting in 1970.

### Conversion and mapping rules
- The conversion guideline stated in Wickens (2002) was applied to map the different public finance concepts under GFSM1986 and GFSM2001.
- GFSM1986 expenditure and revenue items were re-classified according to the GFSM2001 nomenclature.
- Example: GFSM1986 functional classification provides 14 categories whereas GFSM2001 provides 10 categories; IMF (2002) guidelines map these 14 categories into the 10 GFSM2001 categories.

### Reporting-basis rules (accrual vs. cash)
- Given differences in reporting—mostly accrual in GFSM2001 yet only cash basis in GFSM1986—the following rule was applied: whenever an observation was reported on an accrual basis that data point was included; otherwise cash-basis information was taken when available.

### Government level and dataset reference
- Since GFSM1986 reports information for the CCG level but not for the CGG level, the newly-created GFS dataset refers only to the CCG level.

### Macroeconomic variables and external sources
- Remaining macroeconomic variables used in compiling the dataset (e.g., GDP and exchange rates) were taken from either the World Economic Outlook (WEO) or the International Financial Statistics (IFS) databases of the IMF.

### Key numeric facts (preserved)
- Time coverage: 1970 through 2010 (historical data available in both GFSM1986 and GFSM2001).
- Countries in raw dataset: 149.
- Countries with >20 years of data: 91.

### Annex II — Definition of tax variables (GFSM2001 codes in parenthesis)
- Total tax revenue: total tax revenue (11) included as a percent of GDP.
- Income taxes: includes taxes on income, profits, and capital gains (111), Taxes on payroll and workforce (112), and social contributions (12).
- Personal income taxes (PIT): taxes on income, profits, and capital gains payable by individuals (1111).
- Corporate income taxes (CIT): taxes on income, profits, and capital gains payable by corporations and other enterprises (1112).
- Social security contributions: includes social contributions (12) and taxes on payroll and workforce (112).
- Consumption and property taxes: includes taxes on property (113), taxes on goods and services (114), taxes on international trade and transactions (115), and other taxes (116).
- Consumption taxes: includes taxes on goods and services (114), taxes on international trade and transactions (115), and other taxes (116). Taxes on goods and services include: general taxes on goods and services (1141), excises (1142), profits on fiscal monopolies (1143), taxes on specific services (1144), and taxes on use of goods and on permission to use goods or perform activities (1145).
- VAT and sales taxes: includes general taxes on goods and services (1141), which include value-added-taxes (11411), sales taxes (11412), and turnover and other general taxes on goods and services (11413).
- Property taxes: includes taxes on property (113).

### Annex III — Underlying error correction model (methodology)
- Base ARDL(1,1) specification (equation (1)):
  - 011011ttttt YYXXααββ ε −− = +  +  +  + ,
  - Assuming |α1|<1, long-run relation between X and Y at steady state is given by equation (2).
  - Long-run effect of X on Y captured by coefficient (β01 β1)/(1+α1−1) as shown in equation (3).
- ECM reparameterization (equation (4)) derived directly from ARDL(1,1); ECM contains same information and can be estimated with OLS.
- Advantages of ECM:
  - Readily estimates long-run equilibrium coefficients as coefficient on X_{t−1}.
  - Simultaneously and separately estimates short-term and long-term coefficients and the speed of adjustment to equilibrium (1−α1).
  - Explicitly models possibility that some fiscal variables have only short-run effects while others have long-run growth effects.
  - First-differencing of dependent variable and short-run dynamics reduces spurious-regression concerns.

### Annex IV — Regression analysis considering the output level (comparison with Arnold et al (2011))
- Arnold et al (2011) — 21 OECD countries, 1970-2004 — main findings summarized:
  - An increase in consumption and property taxes, while reducing income taxes, has a positive effect on the long-run GDP level.
  - Tax and growth ranking: recurrent taxes on immovable property least harmful, followed by consumption taxes, personal income taxes, and corporate income taxes.
- This paper’s replication and extension (21 OECD countries, 1970-2009; different data sources and linear time trend) — main findings:
  - At the aggregate level, a shift from consumption and property taxes to income taxes negatively affects the long-run output level (negative coefficient on income taxes in column (2); positive coefficient on consumption and property taxes in column (6)).
  - Did not find that CIT is the most detrimental tax to long-run output.
  - Did not find that recurrent taxes on immovable property are the least harmful; positive effect of property taxes is driven by the other property taxes category. Similar result found in Xing (2012).
- Annex Table 1 reports detailed regression coefficients and significance levels for baseline and updated specifications (columns (1)–(10)); note on significance: *Significant at 10% level; ** at 5% level; *** at 1% level.

### Annex V — Summary statistics (selected)
- Annex Table 2: Summary statistics of tax variables by country income group (LICs, MICs, HICs). Examples (values preserved as in source):
  - Total tax revenue:
    - LICs: Obs 739; Mean 14.64; Std.Dev. 5.31; Min 2.50; Max 42.04; Corr. w/ GDP p.c. 0.35***; Corr. w/ GDP p.c. growth 0.02
    - MICs: Obs 727; Mean 20.11; Std.Dev. 9.11; Min 4.38; Max 54.63; Corr. w/ GDP p.c. 0.55***; Corr. w/ GDP p.c. growth 0.05
    - HICs: Obs 821; Mean 34.75; Std.Dev. 7.84; Min 15.92; Max 52.26; Corr. w/ GDP p.c. 0.36***; Corr. w/ GDP p.c. growth -0.06
  - Income taxes:
    - LICs: Obs 7394.94.943.360.2130.57; Corr. w/ GDP p.c. 0.26***; Corr. w/ GDP p.c. growth -0.01
    - MICs: Obs 7279.636.131.2929.49; Corr. w/ GDP p.c. 0.53***; Corr. w/ GDP p.c. growth 0.01
    - HICs: Obs 821; Mean 22.16; Std.Dev. 6.01; Min 7.81; Max 37.49; Corr. w/ GDP p.c. 0.47***; Corr. w/ GDP p.c. growth -0.10**
  - (Full table lists many tax subcategories: Corporate income taxes, Personal income taxes, Social security contributions, Consumption taxes, VAT and sales taxes, Trade taxes, Property taxes, Recurrent taxes on immovable property, Other property taxes — with Obs, Mean, Std.Dev., Min, Max, Corr. w/ GDP p.c., Corr. w/ GDP p.c. growth preserved as in source.)
- Note: ***p<0.001; general government data for OECD countries, Argentina and Brazil; central government data for remaining countries.

- Annex Table 3: Summary statistics of other variables (examples preserved exactly):
  - log(GDP per capita, 2000 constant USD prices):
    - LICs: Obs 739; Mean 7.19; Std.Dev. 0.75; Min 5.39; Max 8.67
    - MICs: Obs 727; Mean 8.93; Std.Dev. 0.86; Min 6.97; Max 11.33
    - HICs: Obs 821; Mean 10.21; Std.Dev. 0.41; Min 8.87; Max 11.06
  - GDP growth rate (percent):
    - LICs: Obs 700; Mean 0.01; Std.Dev. 0.04; Min -0.21; Max 0.19
    - MICs: Obs 701; Mean 0.02; Std.Dev. 0.04; Min -0.18; Max 0.20
    - HICs: Obs 800; Mean 0.02; Std.Dev. 0.02; Min -0.09; Max 0.11
  - Gross fixed capital formation (percent of GDP):
    - LICs: Obs 722; Mean 0.20; Std.Dev. 0.07; Min 0.03; Max 0.49; Corr. w/ GDP p.c. 0.36***; Corr. w/ GDP p.c. growth 0.32***
    - MICs: Obs 716; Mean 0.22; Std.Dev. 0.06; Min 0.07; Max 0.46; Corr. w/ GDP p.c. -0.02; Corr. w/ GDP p.c. growth 0.25***
    - HICs: Obs 820; Mean 0.20; Std.Dev. 0.04; Min 0.12; Max 0.32; Corr. w/ GDP p.c. -0.25***; Corr. w/ GDP p.c. growth 0.18***
  - Average years of schooling (pop. aged 25-64):
    - LICs: Obs 739; Mean 4.15; Std.Dev. 2.14; Min 0.35; Max 10.52; Corr. w/ GDP p.c. 0.46***; Corr. w/ GDP p.c. growth 0.03
    - MICs: Obs 727; Mean 7.23; Std.Dev. 2.48; Min 1.31; Max 12.32; Corr. w/ GDP p.c. 0.63***; Corr. w/ GDP p.c. growth 0.09
    - HICs: Obs 821; Mean 9.64; Std.Dev. 2.06; Min 3.72; Max 13.45; Corr. w/ GDP p.c. 0.65***; Corr. w/ GDP p.c. growth -0.14***
  - Population growth rate:
    - LICs: Obs 739; Mean 0.03; Std.Dev. 0.01; Min -0.01; Max 0.14; Corr. w/ GDP p.c. 0.13***; Corr. w/ GDP p.c. growth -0.07
    - MICs: Obs 727; Mean 0.02; Std.Dev. 0.01; Min 0.00; Max 0.07; Corr. w/ GDP p.c. -0.35***; Corr. w/ GDP p.c. growth -0.08*
    - HICs: Obs 821; Mean 0.01; Std.Dev. 0.01; Min -0.04; Max 0.04; Corr. w/ GDP p.c. -0.17***; Corr. w/ GDP p.c. growth -0.09**

*Source: _wp12257 - ANNEX I. CONSTRUCTION OF THE GFS DATASET (https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12257.pdf)*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12257.pdf_