## 1. Impact of Technical Assistance on Tax Revenues

## Source details

**Canonical URL:** [1. Impact of Technical Assistance on Tax Revenues](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021035-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021035-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021035-print-pdf.pdf.json)

---

### Introduction
- Objective: explore the role of capacity development (CD), in particular technical assistance (TA), in boosting tax revenues across a large cross-section of countries.
- Data and scope:
  - Sample: 115 countries over 2000–18.
  - TA measured in full-time equivalent (FTE), where one FTE comprises 240 working days.
  - Estimation frequency: 5-year non-overlapping averages for FE and FE-IV; logs and annual data for CS-ARDL.
- Motivation:
  - Domestic revenue mobilization (DRM) central to Financing for Development and achievement of the 2030 SDGs.
  - COVID-19 estimated to have pushed up to 100 million people into extreme poverty.
  - IMF CD accounts for about one-third of its budget.

### Literature and controls
- Standard determinants of tax-to-GDP included as controls:
  - GDP per capita, openness (exports+imports/GDP), sector composition (agriculture, mining), inflation, external debt, aid, institutional variables (corruption, government effectiveness).
- Empirical patterns cited as background:
  - GDP per capita: generally positive relationship with the tax ratio.
  - Agriculture: often negative impact on tax ratio in LICs due to subsistence and administrative costs.
  - Mining/extractive sector: mixed results across studies.
  - Openness: generally positive relationship with tax ratio.
  - Inflation (Tanzi effect): found to negatively impact revenues.
  - Corruption: higher corruption reduces tax revenues.
  - Aid: ambiguous effects on tax ratio across studies.

### Econometric approach and data sources
- Baseline panel model (conceptual specification preserved):
  - y_it = α_i + β_i TA_it−1 + θ X_it + ε_it
  - y_it: tax-to-GDP ratio; TA_it: technical assistance in revenue administration and tax policy (lagged); X_it: vector of controls.
- Estimation methods:
  - (i) Panel fixed effects (FE) with 5-year averages.
  - (ii) FE instrumental variables (FE-IV) using instruments: membership of a regional technical assistance center, participation in an IMF program, and lagged dependent variables.
  - (iii) Cross-section Auto Regressive Distributed Lag (CS-ARDL) using annual logged data; PMG considerations referenced.
- CS-ARDL advantages emphasized:
  - Accounts for cross-country heterogeneity and cross-sectional dependence.
  - Restricts long-run coefficients to be homogeneous while allowing heterogeneity in intercepts, short-run coefficients, and error variances.
  - Robust whether variables are I(0) or I(1).
- Data sources:
  - Tax-to-GDP ratio, agriculture value added/GDP, trade openness, debt-to-GDP, inflation: World Bank WDI, IMF WEO, IFS.
  - Country participation in IMF program: MONA database.
  - TA: IMF TA database.
  - Corruption and government effectiveness: World Bank Governance Indicators.
  - Conflict: UCDP/PRIO Armed Conflict Dataset.

### Main results
- FE (5-year non-overlapping averages):
  - Technical assistance aimed at increasing tax revenues is positive and significant at the five percent level.
  - Income per capita has the largest impact on tax revenues among controls.
  - Openness positively increases tax revenues.
  - Control of corruption improved the tax ratio; alternative measures of government effectiveness were not significant.
  - Agricultural sector coefficient not significant; secondary and tertiary sectors also not significant.
  - IMF program participation coefficient positive but insignificant in FE.
- FE-IV (instruments: regional TA center membership, IMF program participation, lagged dependent variables):
  - Technical assistance, trade openness, per capita income, and control of corruption contribute to positive improvement in tax revenues in one equation.
- CS-ARDL (long-run) — Long Run Coefficients (Table 3):
  - Full sample (No. of Countries: 111; Observations: 1,887)
    - Error Correction Term: -0.702*** (0.036)
    - TA: 0.0383*** (0.009)
    - Openness: 0.086*** (0.019)
    - GDP per capita: 0.145*** (0.036)
    - Constant: -0.541 (0.561)
  - Excluding Resource Rich (No. of Countries: 80; Observations: 1,360)
    - Error Correction Term: -0.710*** (0.0379)
    - TA: 0.0179* (0.01)
    - Openness: 0.052** (0.022)
    - GDP per capita: 0.201*** (0.039)
    - Constant: 0.318 (0.541)
  - LICs (No. of Countries: 19; Observations: 323)
    - Error Correction Term: -0.656*** (0.101)
    - TA: 0.123*** (0.019)
    - Openness: 0.283*** (0.033)
    - GDP per capita: -0.353*** (0.118)
    - Constant: -3.279** (1.302)
  - Interpretation:
    - Stable long-run relationship among income per capita, technical assistance, openness, and the tax ratio.
    - Error correction coefficients negative and significant indicating dynamic stability; speed of adjustment marginally slower for LICs (-0.656 vs. -0.702 full sample).
    - TA long-run coefficient largest in LICs: a 10 percent increase in technical assistance improves the tax ratio by 1.2 percent (derived from TA = 0.123 in LICs).
- Subsample and robustness:
  - Excluding resource rich countries: TA coefficient remains positive and significant.
  - LICs and fragile states: TA coefficient positive and significant though sample size is small.
- Selected coefficients from Table 2 (LICs / Fragile States; FE and FE-IV variants):
  - TA (t-1): 0.0555*** (0.0162); 0.0558*** (0.0167); 0.0349** (0.0152); 0.0178** (0.00733); 0.07* (0.04) across columns (1)–(5).
  - Trade Openness: 0.176*** (0.065); 0.196*** (0.0658); 0.189* (0.0980); 0.0559 (0.0405); 0.191 (0.215).
  - GDP per capita: 0.464*** (0.105); 0.490*** (0.110); 0.340*** (0.122); -0.0448 (0.0581); 1.36*** (0.373).
  - Control of Corruption: 0.155** (0.0639); -0.142** (0.0571) in different columns.
  - Observations and countries reported in Table 2: Observations = 126, 126, 180, 175, 34; Number of countries = 77, 77, 11, 10, 8, 13 (counts preserved as displayed).
  - R-squared examples: 0.544, 0.512, 0.325, 0.377, 0.71 across specifications.
- Additional dynamic check:
  - Bivariate CS-ARDL between TA and the tax ratio: error correction term = -0.57; long run coefficient on TA = 0.05.

### Conclusions and policy implications
- Core findings:
  - Technical assistance contributes positively and significantly to increasing tax revenues across the full sample and in key subsamples (excluding resource rich countries; LICs/fragile states).
  - A long-run relationship exists between technical assistance, income per capita, openness, and tax revenues; results robust across FE, FE-IV, and CS-ARDL estimators.
  - Technical assistance has the largest long-run impact on tax revenues in LICs (TA long-run coefficient = 0.123 in LICs).
  - Income per capita and openness consistently have significant positive impacts on the tax ratio.
  - Improvements in control of corruption and governance contribute to improved tax revenues.
- Policy relevance:
  - Scaling up TA in revenue administration and tax policy can help DRM, especially for LICs and fragile states that have large unmet capacity development needs.
  - Building institutional capacity takes time: dynamic results and error correction coefficients indicate gradual adjustment and the need for sustained TA interventions.
  - Complementary reforms to improve governance and reduce corruption are likely to magnify TA effectiveness on revenue mobilization.

*Source: IMF working paper section "1. Impact of Technical Assistance on Tax Revenues" (content unit: wpiea2021035-print-pdf).*

### 1. Impact of Technical Assistance on Tax Revenues _________________________________ 10

### 1. Impact of Technical Assistance on Tax Revenues

### Introduction
- Objective: explore the role of capacity development (CD), in particular technical assistance (TA), in boosting tax revenues across a large cross-section of countries.
- Data and scope:
  - Sample: 115 countries over 2000–18.
  - TA measured in full-time equivalent (FTE), where one FTE comprises 240 working days.
  - Estimation frequency: 5-year non-overlapping averages for FE and FE-IV; logs and annual data for CS-ARDL.
- Motivation:
  - Domestic revenue mobilization (DRM) central to Financing for Development and achievement of the 2030 SDGs.
  - COVID-19 estimated to have pushed up to 100 million people into extreme poverty.
  - IMF CD accounts for about one-third of its budget.

### Literature review (stylized findings used as controls)
- Standard determinants of tax-to-GDP include GDP per capita, openness (exports+imports/GDP), sector composition (agriculture, mining), inflation, external debt, aid, and institutional variables (corruption, government effectiveness).
- Empirical patterns highlighted:
  - GDP per capita: generally positive relationship with tax ratio (Leuthold (1991); Stotsky and Woldermariam (1997); Ghura (1998); Gupta (2007); Addison and Levin (2011)).
  - Agriculture: often negative impact on tax ratio in LICs due to subsistence and administrative costs.
  - Mining/extractive sector: mixed results; some studies find positive, others negative relationships.
  - Openness: generally positive relationship with tax ratio.
  - Inflation (Tanzi effect): found to negatively impact revenues (Ghura (1998)).
  - Corruption: higher corruption reduces tax revenues (Mauro (1996); Ghura (1998); Gupta (2007)).
  - Aid: ambiguous effects on tax ratio across studies.

### Econometric model and data
- Baseline panel model (equation preserved conceptually):
  - y_it = α_i + β_i TA_it−1 + θ X_it + ε_it
  - y_it: tax-to-GDP ratio; TA_it: technical assistance in revenue administration and tax policy (lagged); X_it: vector of controls.
- Estimation methods:
  - (i) Panel fixed effects (FE) with 5-year averages.
  - (ii) FE instrumental variables (FE-IV) using instruments: membership of a regional technical assistance center, participation in an IMF program, and lagged dependent variables.
  - (iii) Cross-section Auto Regressive Distributed Lag (CS-ARDL, Pesaran et al. (1999)) using annual logged data; pool mean group (PMG) considerations referenced.
- CS-ARDL advantages emphasized:
  - Accounts for cross-country heterogeneity and cross-sectional dependence.
  - Restricts long-run coefficients to be homogeneous while allowing heterogeneity in intercepts, short-run coefficients, and error variances.
  - Robust whether variables are I(0) or I(1).
- Data sources:
  - Tax-to-GDP ratio, agriculture value added/GDP, trade openness, debt-to-GDP, inflation: World Bank WDI, IMF WEO, IFS.
  - Country participation in IMF program: MONA database.
  - TA: IMF TA database.
  - Corruption and government effectiveness: World Bank Governance Indicators.
  - Conflict: UCDP/PRIO Armed Conflict Dataset.

### Results and discussion
- Main FE results (5-year non-overlapping averages):
  - Technical assistance aimed at increasing tax revenues is positive and significant at the five percent level.
  - Income per capita has the largest impact on tax revenues among controls.
  - Openness also positively increases tax revenues.
  - Control of corruption improved the tax ratio; alternative measures of government effectiveness were not significant.
  - Agricultural sector coefficient not significant; secondary and tertiary sectors also not significant.
  - IMF program participation coefficient positive but insignificant in FE.
- FE-IV results (instruments: regional TA center membership, IMF program participation, lagged dependent variables):
  - Technical assistance, trade openness, per capita income, and control of corruption contribute to positive improvement in tax revenues in one equation.
- CS-ARDL (long-run) results — Long Run Coefficients (Table 3):
  - Full sample (No. of Countries: 111; Observations: 1,887)
    - Error Correction Term: -0.702*** (0.036)
    - TA: 0.0383*** (0.009)
    - Openness: 0.086*** (0.019)
    - GDP per capita: 0.145*** (0.036)
    - Constant: -0.541 (0.561)
  - Excluding Resource Rich (No. of Countries: 80; Observations: 1,360)
    - Error Correction Term: -0.710*** (0.0379)
    - TA: 0.0179* (0.01)
    - Openness: 0.052** (0.022)
    - GDP per capita: 0.201*** (0.039)
    - Constant: 0.318 (0.541)
  - LICs (No. of Countries: 19; Observations: 323)
    - Error Correction Term: -0.656*** (0.101)
    - TA: 0.123*** (0.019)
    - Openness: 0.283*** (0.033)
    - GDP per capita: -0.353*** (0.118)
    - Constant: -3.279** (1.302)
  - Interpretation:
    - Existence of a stable long-run relationship among income per capita, technical assistance, openness, and the tax ratio.
    - Error correction coefficients negative and significant indicating dynamic stability; speed of adjustment marginally slower for LICs (-0.656 vs. -0.702 full sample).
    - TA long-run coefficient largest in LICs: a 10 percent increase in technical assistance improves the tax ratio by 1.2 percent (derived from TA = 0.123 in LICs).
- Subsample and robustness checks:
  - Excluding resource rich countries: TA coefficient remains positive and significant.
  - LICs and fragile states: TA coefficient positive and significant though sample size is small (Table 2).
- Selected coefficients and significance from Table 2 (LICS / Fragile States; FE and FE-IV variants):
  - TA (t-1): 0.0555*** (0.0162); 0.0558*** (0.0167); 0.0349** (0.0152); 0.0178** (0.00733); 0.07* (0.04) across columns (1)–(5).
  - Trade Openness: 0.176*** (0.065); 0.196*** (0.0658); 0.189* (0.0980); 0.0559 (0.0405); 0.191 (0.215).
  - GDP per capita: 0.464*** (0.105); 0.490*** (0.110); 0.340*** (0.122); -0.0448 (0.0581); 1.36*** (0.373).
  - Control of Corruption: 0.155** (0.0639); -0.142** (0.0571) in different columns.
  - Observations and countries reported in Table 2: Observations = 126, 126, 180, 175, 34; Number of countries = 77, 77, 11, 10, 8, 13 (table formatting preserves counts as displayed).
  - R-squared examples: 0.544, 0.512, 0.325, 0.377, 0.71 across specifications.
- Additional dynamic check:
  - Bivariate CS-ARDL between TA and the tax ratio: error correction term = -0.57; long run coefficient on TA = 0.05 (noted in text).

### Conclusions and policy implications
- Core findings:
  - Technical assistance contributes positively and significantly to increasing tax revenues across the full sample and in key subsamples (excluding resource rich countries; LICs/fragile states).
  - A long-run relationship exists between technical assistance, income per capita, openness, and tax revenues; results robust across FE, FE-IV, and CS-ARDL estimators.
  - Technical assistance has the largest long-run impact on tax revenues in LICs (TA long-run coefficient = 0.123 in LICs).
  - Income per capita and openness consistently have significant positive impacts on the tax ratio.
  - Improvements in control of corruption and governance contribute to improved tax revenues.
- Policy relevance:
  - Scaling up TA in revenue administration and tax policy can help DRM, especially for LICs and fragile states that have large unmet capacity development needs.
  - Building institutional capacity takes time: dynamic results and error correction coefficients indicate gradual adjustment and the need for sustained TA interventions.
  - Complementary reforms to improve governance and reduce corruption are likely to magnify TA effectiveness on revenue mobilization.

*Source: IMF working paper section "1. Impact of Technical Assistance on Tax Revenues" (content unit: wpiea2021035-print-pdf).*

### References

### wpiea2021035-print-pdf - References

### Key topics covered in the references
- Determinants of tax revenue and tax capacity
- Tax revenue mobilization episodes in emerging markets and low-income countries
- Tax revenues in fragile and conflict-affected states
- Trade liberalization, exchange rate changes, and tax revenue in Sub-Saharan Africa
- Domestic resource mobilization, aid allocation, and revenue response
- Institutions, corruption, and economic growth
- Econometric methods for dynamic heterogeneous panels and bias in panel estimators
- Classic contributions on taxation for development and public fiscal behavior in developing countries

### References on tax revenue, tax capacity, and mobilization
- Addison, T., and Levin, J., 2011, “The Determinants of Tax R evenue in Sub-Saharan Africa, Journal of International Development,” Vol. 23 pp. 936–58.
- Akitoby, B., J. Honda, and K. Primus, 2020, “Tax Revenues in Fragile and Conflict-Affected States—Why are Low and How Can We Raise Them? IMF Working Paper 20/143 (Washington: International Monetary Fund).
- Akitoby, Bernardin, Anja Baum, Clay Hackney, Olamide Harrison, Keyra Primus, and Veronique Salins, 2018, “Large Tax Revenue Mobilization Episodes in Emerging Markets and Low-Income Countries: Lessons from a New Dataset,” IMF Working Paper 18/234 (Washington: International Monetary Fund).
- Gupta, S. 2007, “Determinant s of Tax Revenue Efforts in Developing Countries,” IMF Working Paper 07/184 (Washington: International Monetary).
- Gaspar, V., L. Jaramillo, P. Wingender, 2016, “Tax Capacity and Growth: Is There a Tipping Point?” IMF Working Paper 16/234 (Washington: International Monetary Fund).
- Teera, J. M., and J. Hudson, 2004, “Tax Performance: A Comparative Study, Journal of International Development, Vol. 16(6), pp. 785–802.

### References on aid, fiscal behavior, and revenue responses
- Bhushan, A., and Y. Samy, 2014, “Fiscal Capacity and Aid Allocation Domestic Resource Mobilization and Foreign Aid in Developing Countries,” (UNRISD Working papers No. 7), Switzerland.
- Crivelli, E. and S. Gupta 2016, “ Does Conditionality Mitigate the Potential NegativeEffect of Aid on Revenues? IMF Working Paper 16/142 (Washington: International Monetary Fund).
- Gupta, S., B. Clements, A. Pivovarsky and E. Tiongson, 2003, “Foreign Aid and Revenue Response: Does the Composition of Aid Matter?” IMF Working Paper 03/176 (Washington: International Monetary Fund).
- Heller, P.S., 1975, “A Model of Public Fiscal Behavior in Developing Countries: Aid, Investment and Taxation,” The American Economic Review, Vol. 65, pp. 429–445.

### References on institutions, growth, and corruption
- Acemoglu, D., 2002, “Institutional Causes, Macroeconomic Symptoms: Volatility, Crises, and Growth,” Journal of Monetary Economics, Vol. 50 (January), pp. 49–123.
- Barro, R. and J. Lee., 2005, “IMF Programs: Who is Chosen and What are the Effects?” Journal of Monetary Economics, Vol. 52, pp. 1245–1269.
- Kaldor, N., 1963, “Taxation for Economic Development,” Journal of Modern African Studies, Vol. 1, pp. 7–23.
- Landes, David S., 1998, “The Wealth and Poverty of Nations: Why Some Are So Rich and Some So Poor” (New York: W.W. Norton).
- Mauro, P., 1996, “The Effects of Corruption on Growth, Investment, and Government Expenditures,” IMF Working Paper 96/98 (Washington: International Monetary Fund).
- North, Douglass C., 1990, “Institutions, Institutional Change and Economic Performance” (New York: Cambridge University Press).
- Tanzi, V., 1989, “The Impact of Macroeconomic Policies on the Level of Taxation and Fiscal Balances in Developing Countries,” Staff Papers, Vol. 36, pp.633–56 (Washington: International Monetary Fund).
- Tanzi, V., 1998, “Corruption Around the World, Causes, Consequences and Cures”, IMF Working Paper 98/63 (Washington: International Monetary Fund).
- World Economic Outlook, 2003, “Growth and Institutions” (Washington: International Monetary Fund).

### References on methodology and econometrics
- Buzzi, Samuel and Michael A. Clemens, 2013, “Blunt Instruments: Avoiding Common Pitfalls in Identifying the Causes of Economic Growth,” American Economic Journal: Macroeconomics, 5(2): 152–86.
- Cherif, A. and F. Hasanov, 2020, “Sharp Instrument: A Stab at Identifying the Causes of Economic Growth,” Forthcoming Empirical Economics.
- Nickell, S., 1981, “Biases in Dynamic Models with Fixed Effects,” Econometrica, Vol. 49, pp. 1417–1426.
- Pesaran, M.H, and R. Smith R, 1995, “Estimating Long-run Relationships from Dynamic Heterogeneous Panels,” Journal of Econometrics, Vol. 68, pp. 79–113.
- Pesaran, M.H., Y. Shin, and R. P. Smith, 1999, “Pooled Mean Group Estimation of Dynamic Heterogeneous Panels,” Journal of the American Statistical Association, Vol. 94. pp. 621–634.

*Source: wpiea2021035-print-pdf - References*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021035-print-pdf.pdf_