## Appendix 1. Synthesis of Variables 22

## Source details

**Canonical URL:** [Appendix 1. Synthesis of Variables 22](https://www.imf.org/-/media/files/publications/tnm/2025/english/tnmea2025015.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/tnm/2025/english/tnmea2025015.pdf.md)
- [Structured JSON version](/-/media/files/publications/tnm/2025/english/tnmea2025015.pdf.json)

---

### Overview and motivation
- Tax noncompliance is not directly observable; exact measurement of the compliance gap is impossible, so the aim is reasonable, comparable, and consistent estimates.
- Central objective: derive an estimate that approximates or simulates the result of a standard RA-GAP estimation at scale, recognizing an inevitable margin of error.
- Reverse Method (RM) design goals:
  - Consistent with the RA-GAP framework.
  - Global, scalable, and cost-effective.
  - A complement to, not a replacement for, standard RA-GAP assessments.
- RM intended uses:
  - Global monitoring of VAT gap trends across many countries and years.
  - Cross-checking standard VAT gap assessments from production-based top-down techniques.
  - Generating indicative estimates for countries lacking standard RA-GAP assessments.

### Methodological foundation: C-efficiency and the accounting identity
- C-efficiency (E_C) definition: ratio of VAT revenue to the product of the standard rate and consumption (practical comparator of actual VAT raised vs. perfect enforcement at the standard rate).
- Relationship among C-efficiency, compliance gap (Γ), and policy gap (P):
  - Equation (1): E_C ≡ (1 − Γ)(1 − P)
  - Reverse identity used to obtain compliance gap as a residual:
    - Equation (2): 1 − Γ ≡ E_C / (1 − P)
- Illustration from RA-GAP assessments (38 countries, selected years):
  - Example observation: C-efficiency = 32.14 percent; VAT policy gap = 24.55 percent to potential; resulting compliance gap = 57.40 percent to potential (calculated as 1 − 0.3214/[1 − 0.2455]).
- RM employs the definition of C-efficiency to estimate the VAT compliance gap residually.

### Calibrated estimation model (accounting and estimation strategy)
- Policy gap decomposition (Hutton 2017): tax expenditure gap (TEG) and nontaxable goods and services gap (NTG).
  - TEG to potential: difference between potential VAT under normative tax policy structure (PV2) and VAT under current tax policy structure, expressed as a percentage of PV2.
  - NTG to potential: difference between potential VAT under comprehensive policy structure (PV3) and PV2, expressed as a percentage of PV3.
- Accounting identity transformed to actionable estimation:
  - Equation (3) expresses (1 − Γ) using TEG and NTG terms.
- Estimation strategy: operate in logarithmic terms (Equation (4)) and approximate three right-hand side variables:
  - E_C (C-efficiency): use IMF’s FAD C-efficiency data set (based on VAT collection data from WoRLD, final consumption from WEO, VAT rates from EY and KPMG sources).
  - TEG/PV2: approximate using the Global Tax Expenditure Dataset and an iterative uplift to approximate full-compliance provisions.
  - NTG/PV3: approximate using GDP shares of public administration, public education, and public health (GDP_O, GDP_P, GDP_Q) as proxies for final consumption of those sectors, multiplied by the standard VAT rate r.
- Proxies adjusted via constant calibration factors assumed constant across countries/time:
  - μ1 to adjust Log(E_C) proxy differences (Equation (5)): Log E_C = μ1 · Log E_C_FAD
  - μ2 to adjust Log(TEG/PV2) proxy differences (Equation (6)): Log TEG/PV2 = μ2 · [Log (sum VAT_p_ExpProv / (FC − AV0) r − NTG/PV3)] (iterative uplift applied to provisions under zero evasion)
  - μ3 to adjust Log(NTG/PV3) proxy differences (Equation (7)): Log NTG/PV3 = μ3 · [Log ((GDP_O + GDP_P + GDP_Q) r / (FC − AV0) r)]
- Calibration equation (Equation (8)):
  - Log(1 − Γ) = μ0 + μ1 · Log E_C_FAD + μ2 · Log[TET proxy terms] + μ3 · Log[NTG proxy terms] + ε
  - μ0 is intercept; μ1, μ2, μ3 are calibration coefficients; ε is consolidated error term.

### Calibration results and predictive specification
- Calibration sample: 208 VAT gap observations from 43 countries based on IMF RA-GAP direct top-down estimates for selected years during 2010–23.
- Estimated calibration coefficients (Table 2, four specifications):
  - Column (a) All (no constant):
    - μ1 = .62*** (standard error .03)
    - −μ2 = −.43** (standard error .20)
    - −μ3 = −.30* (standard error .16)
    - adj. R2 = .56
  - Column (b) All (with constant):
    - μ0 = −.50*** (standard error .03)
    - μ1 = .53*** (standard error .03)
    - −μ2 = −2.14*** (standard error .23)
    - −μ3 = −1.87*** (standard error .09)
    - adj. R2 = .84
  - Column (c) Non-EU (with constant) — chosen specification:
    - μ0 = −.40*** (standard error .04)
    - μ1 = .55*** (standard error .03)
    - −μ2 = −1.99*** (standard error .23)
    - −μ3 = −1.26*** (standard error .15)
    - adj. R2 = .80
    - No. obs. = 144; No. countries = 33
  - Column (d) EU (with constant) — chosen specification:
    - μ0 = −.64*** (standard error .07)
    - μ1 = .68*** (standard error .06)
    - −μ2 = −3.38*** (standard error .68)
    - −μ3 = −2.23*** (standard error .12)
    - adj. R2 = .92
    - No. obs. = 64; No. countries = 12
- Statistical significance: *p < .10; **p < .05; ***p < .01.
- Interpretation notes:
  - If proxies were exact, μ0 = 0 and μ1 = μ2 = μ3 = 1; observed deviations indicate systematic differences in proxies.
  - Separate calibrations for EU and non-EU produce higher predictive power, reflecting differences in VAT design (e.g., intra-community transactions).
- Final RM model selected using columns (c) and (d) (Non-EU and EU subsamples) to maximize predictive performance.

### Validation, out-of-sample predictions, and comparisons
- Visual comparisons:
  - Figure 2 compares RA-GAP direct estimates (blue) and RM calibrated estimates (green) across individual country-year observations.
  - Figure 3 presents out-of-sample predictions: RM calibrated results (green) vs. top-down estimates from countries, RA-GAP missions, and World Bank RA-GAP model results (orange).
  - Figure 4 compares RM with other indirect techniques (e.g., C-efficiency and import-to-consumption ratio models).
- Key validation points:
  - RM provides indicative compliance gap estimates for countries/years lacking RA-GAP assessments.
  - RM is used as a comparative exercise against the RA-GAP benchmark because the true compliance gap is unobservable.
  - Out-of-sample results demonstrate consistency and comparability of RM calibrated estimates relative to established top-down benchmarks.

### Applying the Reverse Method: scope, data, and sample
- RM was applied to 111 countries to produce annual estimates of the VAT compliance gap for the period 2010–23.
- Calibration model corresponds to columns (c) and (d) of Table 2.
- Data sources used:
  - C-efficiency data from IMF Fiscal Affairs Department data sets.
  - VAT expenditure provisions from the Global Tax Expenditure Dataset.
  - GDP series for public administration, education, and health from national accounts.
- Sample composition:
  - 10 advanced economies (AEs), 55 emerging market economies (EMEs), and 46 low-income developing countries (LIDCs).
- For 12 LIDC that do not report to the Global Tax Expenditure Dataset, the average VAT expenditure of five closely related global comparator countries was used.

### Results: selected country comparisons and aggregate trends
- Individual country comparisons:
  - Figure 5 compares individual compliance gap results (Reverse RA-GAP, green points) with standard RA-GAP assessment estimates (blue/orange points).
  - Consistency between Reverse Method and RA-GAP results is expected because the method was calibrated using RA-GAP mission results.
  - Other estimation techniques produce greater differences with RA-GAP in some cases.
  - Intended use: order-of-magnitude estimates and trends, not highly precise estimates for individual country-year pairs.
- VAT compliance gaps (Reverse RA-GAP), percent of GDP, average 2010−23:
  - AEs: 2.4 percent of GDP.
  - EMEs: 2.5 percent of GDP.
  - LIDCs: 3.1 percent of GDP.
- Compliance gap to potential: trends 2010–2023:
  - LIDCs: downward trend, about 44 percent in 2010 to approximately 40 percent by 2023.
  - EMEs: relatively stable, fluctuating slightly about 31 percent.
  - AEs: decline from about 26 percent in 2010 to 21 percent by 2023.
- VAT compliance gap to GDP, averaged over 2010–23 by region:
  - AFR (Africa, 38 countries): 2.9
  - APD (Asia and the Pacific, 15 countries): 2.2
  - EUR (Europe, 22 countries): 2.8
  - MCD (Middle East and Central Asia, 15 countries): 2.8
  - WHD (Western Hemisphere, 21 countries): 2.7
- Compliance gap to potential: trends 2010–2023 by region:
  - AFR: decreasing from 43 percent in 2010 to 41 percent in 2023.
  - APD: decreasing from 36 percent in 2010 to 33 percent in 2023.
  - EUR: decline from 26 percent in 2010 to 23 percent by 2023.
  - WHD: relatively stable at about 33 percent from 2010 to 2023.
  - MCD: increase from 32 percent in 2010 to 34 percent in 2023.

### Limitations, caveats, and practical considerations
- Fundamental limitation: compliance gap is not directly observable; RM produces estimates approximating a RA-GAP mirror, with an accepted margin of error.
- Proxy limitations and systematic differences:
  - E_C_FAD uses cash VAT collection (AV0) rather than accrued VAT collection that ideally should be used.
  - Final consumption (FC) adjustments may differ between national accounts and proxies.
  - Global Tax Expenditure Dataset reports vary across countries in coverage and benchmark assumptions; some provisions may be omitted or measured under current compliance rather than full compliance.
  - NTG proxies use GDP shares (GDP_O, GDP_P, GDP_Q) instead of final consumption by commodity due to data availability; introduces imprecision.
  - Country-specific VAT design features (administrative restrictions on input credits, limited exporter refunds, VAT on certain exports, treatment of financial services) can distort TEG and thus TEG-based estimates.
  - Differences in measurement dates across data sources (RA-GAP missions mostly between 2014 and 2020 vs. right-hand variables using updated 2025 figures) can affect coefficients.
- Calibration intent: adjust for systematic proxy differences via μ coefficients and intercept μ0 to maximize predictive power rather than retest strict accounting identities.
- Practical application guidance:
  - RM suitable for comparable large-scale monitoring, cross-checking, and indicative country-level estimates where RA-GAP is not feasible.
  - Not intended to replace detailed RA-GAP accounting assessments for country-specific policy or operational audit targeting.

### Next steps and methodological improvements
- (1) Add more and diverse RA-GAP estimation cases in the calibration process to enhance precision.
- (2) Incorporate increasingly complete and standardized tax expenditure estimates.
- (3) Keep estimates up to date by using the latest versions of global databases and, if possible, extend historical information before 2010.
- (4) Develop deep comparative analysis of results versus other techniques, requiring compilation of detailed information on those techniques.

### Appendix 1 — key variables and identity excerpts
- Key variables and sources:
  - E_C: C-efficiency—defined as the ratio of accrued actual VAT collection to the product of the standard rate and final consumption (net of VAT collection compiled in national accounts). Source: IMF’s FAD C-efficiency data set.
  - Γ: Compliance gap to potential PV1—proportion of VAT that is not perfectly enforced to potential. Estimated by the Reverse Method.
  - P: Policy gap to potential PV3—proportion of the VAT difference because of some goods and services not being taxable or being taxable at a different rate than the standard one. Not estimated directly.
  - PV3: Potential VAT revenue under full compliance and a comprehensive tax policy structure. Estimated by final consumption, FC, minus the official VAT collection, AV0, multiplied by the VAT standard rate, r.
  - TEG: Tax expenditure gap. VAT revenue forgone because of VAT expenditure provisions included in each country’s official tax expenditure report as an approximation.
  - NTG: Nontaxable goods and services gap. Equal to the difference between PV3 and PV2. Estimated using final consumption of three sectors: public administration, public education, and public health.
  - AV0: Official actual VAT collection. Source: IMF’s FAD C-efficiency data set.
  - r: Standard VAT rate. Source: IMF’s FAD C-efficiency data set.
  - FC: Final Consumption. Source: IMF’s WoRLD data set.
- Key identity derivations (exact expressions as presented):
  - AV = PV3 − (Γ × PV1) − (P × PV3)                            (1)
  - Ec = 1 − Γ (1−P) − P                                                                        (5)
  - Ec = (1 − Γ) x (1−P)                                                                        (6)

### Appendix 3 — applying the Reverse: country-specific accounting approach
- The Reverse Method’s country-specific application requires precise measurements of:
  - Ec (accrual C-efficiency): yAccrual value-added tax (VAT) collection; yNational Accounts VAT collection; yPrecise portion of the National Accounts VAT collection embedded in the Final Consumption.
  - VAT expenditure report based on a à la Keen benchmark (final consumption, destination principle, standardized rate) to measure TEG, including:
    - provisions not reported when a different VAT benchmark is applied (for example, financial services, final consumption related to dwelling ownership, negative provisions, and so on);
    - adjusting provisions to full compliance, if needed;
    - adjusting for joint effects (all provisions), if needed.
  - National accounts for final consumption for sectors O, P, and Q (typically available from Supply Use Tables) to measure NTG:
    - Sector O: Public administration and defense.
    - Sector P: Public education.
    - Sector Q: Public health.

### Appendix 4 — descriptive statistics (selected entries)
- Appendix Table 4.1 Calibration Model (Table 2 and Figure 1) — selected statistics (DEC indicates variables expressed in decimals; RA-GAP indicates figures from a RA-GAP tax gap assessment mission):
  - DECE_C: Count Mean Median Std Min Max — values as presented in source (table formatting retained in source).
  - DEC_NTG_PV2_1−(...): Count Mean Median Std Min Max — values as presented in source (table formatting retained in source).
  - DEC_TEG_PV2_1−(...): Count Mean Median Std Min Max — values as presented in source (table formatting retained in source).
  - DEC_(1 − Γ) RA-GAP: Count Mean Median Std Min Max — values as presented in source (table formatting retained in source).
- Appendix Table 4.2 Applying the Calibration Model to All Countries (Figures 2, 3, 4, 5, and 6) — selected statistics (DEC indicates variables expressed in decimals):
  - DECE_C: Count Mean Median Std Min Max — values as presented in source (table formatting retained in source).
  - DEC_NTG_PV2_1−(...): Count Mean Median Std Min Max — values as presented in source (table formatting retained in source).
  - DEC_TEG_PV2_1−(...): Count Mean Median Std Min Max — values as presented in source (table formatting retained in source).
  - DEC_(1 − Γ) ReverseMethod: Count Mean Median Std Min Max — values as presented in source (table formatting retained in source).

*The Reverse Method: A New Indirect Technique for Global VAT Compliance Gap Monitoring — TNM/2025/15, Appendix 1–4*

### Appendix 1. Synthesis of Variables 22

### Appendix 1. Synthesis of Variables 22

### Contained appendices and page references
- Appendix 1. Synthesis of Variables 22
- Appendix 2. Standard RA-GAP Assessment Mission 23
- Appendix 3. Applying the Reverse Method’s Accounting Approach under a Country-Specific Purpose 24
- Appendix 4. Descriptive Statistics 25

*Source: tnmea2025015 - Appendix 1. Synthesis of Variables 22*

### References                                                                                                              

### The Reverse Method: A New Indirect Technique for Global VAT Compliance Gap Monitoring (tnmea2025015 - References)

### Overview and Motivation
- Tax noncompliance is not directly observable; exact measurement of the compliance gap is impossible, so the aim is reasonable, comparable, and consistent estimates.
- Central objective: derive an estimate that approximates or simulates the result of a standard RA-GAP estimation at scale, recognizing an inevitable margin of error.
- RM (Reverse Method) is designed to be:
  - Consistent with the RA-GAP framework.
  - Global, scalable, and cost-effective.
  - A complement to, not a replacement for, standard RA-GAP assessments.
- RM intended uses:
  - Global monitoring of VAT gap trends across many countries and years.
  - Cross-checking standard VAT gap assessments from production-based top-down techniques.
  - Generating indicative estimates for countries lacking standard RA-GAP assessments.

### Methodological Foundation: C-efficiency and the Accounting Identity
- C-efficiency (E_C) definition: ratio of VAT revenue to the product of the standard rate and consumption (practical comparator of actual VAT raised vs. perfect enforcement at the standard rate).
- Relationship among C-efficiency, compliance gap (Γ), and policy gap (P):
  - Equation (1): E_C ≡ (1 − Γ)(1 − P)
  - Reverse identity used to obtain compliance gap as a residual:
    - Equation (2): 1 − Γ ≡ E_C / (1 − P)
- Illustration from RA-GAP assessments (38 countries, selected years):
  - Example observation: C-efficiency = 32.14 percent; VAT policy gap = 24.55 percent to potential; resulting compliance gap = 57.40 percent to potential (calculated as 1 − 0.3214/[1 − 0.2455]).
- RM employs the definition of C-efficiency to estimate the VAT compliance gap residually.

### Calibrated Estimation Model (Accounting and Estimation Strategy)
- Decomposition of policy gap (Hutton 2017): tax expenditure gap (TEG) and nontaxable goods and services gap (NTG).
  - TEG to potential: difference between potential VAT under normative tax policy structure (PV2) and VAT under current tax policy structure, expressed as a percentage of PV2.
  - NTG to potential: difference between potential VAT under comprehensive policy structure (PV3) and PV2, expressed as a percentage of PV3.
- Accounting identity transformed to actionable estimation:
  - Equation (3) expresses (1 − Γ) using TEG and NTG terms.
- Estimation strategy: operate in logarithmic terms (Equation (4)) and approximate three right-hand side variables:
  - E_C (C-efficiency): use IMF’s FAD C-efficiency data set (based on VAT collection data from WoRLD, final consumption from WEO, VAT rates from EY and KPMG sources).
  - TEG/PV2: approximate using the Global Tax Expenditure Dataset and an iterative uplift to approximate full-compliance provisions.
  - NTG/PV3: approximate using GDP shares of public administration, public education, and public health (GDP_O, GDP_P, GDP_Q) as proxies for final consumption of those sectors, multiplied by the standard VAT rate r.
- Proxies adjusted via constant calibration factors assumed constant across countries/time:
  - μ1 to adjust Log(E_C) proxy differences (Equation (5)): Log E_C = μ1 · Log E_C_FAD
  - μ2 to adjust Log(TEG/PV2) proxy differences (Equation (6)): Log TEG/PV2 = μ2 · [Log (sum VAT_p_ExpProv / (FC − AV0) r − NTG/PV3)] (iterative uplift applied to provisions under zero evasion)
  - μ3 to adjust Log(NTG/PV3) proxy differences (Equation (7)): Log NTG/PV3 = μ3 · [Log ((GDP_O + GDP_P + GDP_Q) r / (FC − AV0) r)]
- Substituting the three calibrated proxies into the logarithmic identity yields the calibration equation (Equation (8)):
  - Log(1 − Γ) = μ0 + μ1 · Log E_C_FAD + μ2 · Log[TET proxy terms] + μ3 · Log[NTG proxy terms] + ε
  - μ0 is intercept; μ1, μ2, μ3 are calibration coefficients; ε is consolidated error term.

### Calibration Results and Predictive Specification
- Calibration sample: 208 VAT gap observations from 43 countries based on IMF RA-GAP direct top-down estimates for selected years during 2010–23.
- Estimated calibration coefficients (Table 2, four specifications):
  - Column (a) All (no constant):
    - μ1 = .62*** (standard error .03)
    - −μ2 = −.43** (standard error .20)
    - −μ3 = −.30* (standard error .16)
    - adj. R2 = .56
  - Column (b) All (with constant):
    - μ0 = −.50*** (standard error .03)
    - μ1 = .53*** (standard error .03)
    - −μ2 = −2.14*** (standard error .23)
    - −μ3 = −1.87*** (standard error .09)
    - adj. R2 = .84
  - Column (c) Non-EU (with constant) — chosen specification:
    - μ0 = −.40*** (standard error .04)
    - μ1 = .55*** (standard error .03)
    - −μ2 = −1.99*** (standard error .23)
    - −μ3 = −1.26*** (standard error .15)
    - adj. R2 = .80
    - No. obs. = 144; No. countries = 33
  - Column (d) EU (with constant) — chosen specification:
    - μ0 = −.64*** (standard error .07)
    - μ1 = .68*** (standard error .06)
    - −μ2 = −3.38*** (standard error .68)
    - −μ3 = −2.23*** (standard error .12)
    - adj. R2 = .92
    - No. obs. = 64; No. countries = 12
- Statistical significance: *p < .10; **p < .05; ***p < .01.
- Interpretation notes:
  - If proxies were exact, μ0 = 0 and μ1 = μ2 = μ3 = 1; observed deviations indicate systematic differences in proxies.
  - Separate calibrations for EU and non-EU produce higher predictive power, reflecting differences in VAT design (e.g., intra-community transactions).
- Final RM model selected using columns (c) and (d) (Non-EU and EU subsamples) to maximize predictive performance.

### Validation, Out-of-Sample Predictions, and Comparisons
- Visual comparisons:
  - Figure 2 compares RA-GAP direct estimates (blue) and RM calibrated estimates (green) across individual country-year observations.
  - Figure 3 presents out-of-sample predictions: RM calibrated results (green) vs. top-down estimates from countries, RA-GAP missions, and World Bank RA-GAP model results (orange).
  - Figure 4 compares RM with other indirect techniques (e.g., C-efficiency and import-to-consumption ratio models).
- Key validation points:
  - RM provides indicative compliance gap estimates for countries/years lacking RA-GAP assessments.
  - RM is used as a comparative exercise against the RA-GAP benchmark because the true compliance gap is unobservable.
  - Out-of-sample results demonstrate consistency and comparability of RM calibrated estimates relative to established top-down benchmarks.

### Limitations, Caveats, and Practical Considerations
- Fundamental limitation: compliance gap is not directly observable; RM produces estimates approximating a RA-GAP mirror, with an accepted margin of error.
- Proxy limitations and systematic differences:
  - E_C_FAD uses cash VAT collection (AV0) rather than accrued VAT collection that ideally should be used.
  - Final consumption (FC) adjustments may differ between national accounts and proxies.
  - Global Tax Expenditure Dataset reports vary across countries in coverage and benchmark assumptions; some provisions may be omitted or measured under current compliance rather than full compliance.
  - NTG proxies use GDP shares (GDP_O, GDP_P, GDP_Q) instead of final consumption by commodity due to data availability; introduces imprecision.
  - Country-specific VAT design features (administrative restrictions on input credits, limited exporter refunds, VAT on certain exports, treatment of financial services) can distort TEG and thus TEG-based estimates.
  - Differences in measurement dates across data sources (RA-GAP missions mostly between 2014 and 2020 vs. right-hand variables using updated 2025 figures) can affect coefficients.
- Calibration intent: adjust for systematic proxy differences via μ coefficients and intercept μ0 to maximize predictive power rather than retest strict accounting identities.
- Practical application guidance:
  - RM suitable for comparable large-scale monitoring, cross-checking, and indicative country-level estimates where RA-GAP is not feasible.
  - Not intended to replace detailed RA-GAP accounting assessments for country-specific policy or operational audit targeting.

*Source: tnmea2025015 - References*

### 1. Morrow, Smart, and Swistak (2019)2. Reverse Method

### 1. Morrow, Smart, and Swistak (2019)2. Reverse Method

### Applying the Reverse Method
- The Reverse Method was applied to 111 countries to produce annual estimates of the value-added tax (VAT) compliance gap for the period 2010–23.
- Calibration model corresponds to columns (c) and (d) of Table 2 from the previous section.
- Data sources used:
  - C-efficiency data from IMF Fiscal Affairs Department data sets.
  - VAT expenditure provisions from the Global Tax Expenditure Dataset.
  - GDP series for public administration, education, and health from national accounts.
- Sample composition:
  - 10 advanced economies (AEs), 55 emerging market economies (EMEs), and 46 low-income developing countries (LIDCs).
- For 12 LIDC that do not report to the Global Tax Expenditure Dataset, the average VAT expenditure of five closely related global comparator countries was used.

### Results in Selected Countries
- Figure 5 compares individual compliance gap results (Reverse RA-GAP, green points) with standard RA-GAP assessment estimates (blue/orange points).
- The selection of displayed countries is random but includes countries with RA-GAP mission estimates.
- Observations:
  - Consistency between Reverse Method and RA-GAP results is expected because the method was calibrated using RA-GAP mission results.
  - Other estimation techniques produce greater differences with RA-GAP in some cases.
- Intended use:
  - Method primarily provides order-of-magnitude estimates and trends, not highly precise estimates for individual country-year pairs.
  - For high precision, countries should use well-prepared tax expenditure reports and either apply the identity to infer the compliance gap as a residual (see Appendix 3) or conduct a standardized RA-GAP assessment.

### VAT Compliance Gap Trends by Income Level
- Average VAT compliance gaps (Reverse RA-GAP), percent of GDP, average 2010−23:
  - AEs: 2.4 percent of GDP.
  - EMEs: 2.5 percent of GDP.
  - LIDCs: 3.1 percent of GDP.
- Compliance gap to potential: trends 2010–2023:
  - LIDCs: downward trend, about 44 percent in 2010 to approximately 40 percent by 2023.
  - EMEs: relatively stable, fluctuating slightly about 31 percent.
  - AEs: decline from about 26 percent in 2010 to 21 percent by 2023.
- Note: Indicator quantifies lost revenues to GDP because of noncompliance; compliance gap as a percentage of potential is a better approximation of compliance performance.

### VAT Compliance Gap Trends by Region
- Average VAT compliance gap to GDP, averaged over 2010–23:
  - AFR (Africa, 38 countries): 2.9
  - APD (Asia and the Pacific, 15 countries): 2.2
  - EUR (Europe, 22 countries): 2.8
  - MCD (Middle East and Central Asia, 15 countries): 2.8
  - WHD (Western Hemisphere, 21 countries): 2.7
- Compliance gap to potential: trends 2010–2023:
  - AFR: decreasing from 43 percent in 2010 to 41 percent in 2023.
  - APD: decreasing from 36 percent in 2010 to 33 percent in 2023.
  - EUR: decline from 26 percent in 2010 to 23 percent by 2023.
  - WHD: relatively stable at about 33 percent from 2010 to 2023.
  - MCD: increase from 32 percent in 2010 to 34 percent in 2023.

### Limitations and Caveats
- General limitations:
  - Estimates are not measurements; the compliance gap cannot be directly observed, introducing uncertainty.
  - C-efficiency concept has inherent limitations and may not capture VAT design deviations from the consumption macroeconomic base (e.g., restrictions on refunds, restrictions on credits).
  - Compiling tax expenditure provisions across countries with different concepts can misstate effects on the C-efficiency identity (for example, differences in treatment of financial services).
  - Errors are inherited from reliance on national accounts, which are subject to periodic revisions.
- Specific limitations noted:
  - Calibration captures some disparities but cannot entirely represent all differences in VAT design.
  - Advanced economies limitation: few AEs are included in the calibration process; standard RA-GAP assessments currently do not include most of the larger AEs, such as those belonging to the G-20.

### Conclusions and Advantages of the Reverse Method
- Purpose:
  - Approximate the result a standard RA-GAP assessment would have produced; serve as a leading indicator of the VAT compliance gap.
  - Not intended to replace standard RA-GAP assessments or to provide highly precise country-year estimates.
- Main contribution:
  - Introduces a new technique to produce VAT compliance gap estimates with broad comparability across countries.
- Advantages:
  - Provides VAT gap estimates for many years and countries using the same tool.
  - Based on public data sets, facilitating updates.
  - Allows rough estimation in countries with severely limited data sets.
  - Enables cross-checks of standard top-down VAT gap assessments.
- Next steps to refine methodology:
  - (1) Add more and diverse RA-GAP estimation cases in the calibration process to enhance precision.
  - (2) Incorporate increasingly complete and standardized tax expenditure estimates.
  - (3) Keep estimates up to date by using the latest versions of global databases and, if possible, extend historical information before 2010.
  - (4) Develop deep comparative analysis of results versus other techniques, requiring compilation of detailed information on those techniques.

### Appendix 1 — Synthesis of Variables (selected entries)
- Key variables and sources:
  - E_C: C-efficiency—defined as the ratio of accrued actual VAT collection to the product of the standard rate and final consumption (net of VAT collection compiled in national accounts). Source: IMF’s FAD C-efficiency data set.
  - Γ: Compliance gap to potential PV1—proportion of VAT that is not perfectly enforced to potential. Estimated by the Reverse Method.
  - P: Policy gap to potential PV3—proportion of the VAT difference because of some goods and services not being taxable or being taxable at a different rate than the standard one. Not estimated directly.
  - PV3: Potential VAT revenue under full compliance and a comprehensive tax policy structure. Estimated by final consumption, FC, minus the official VAT collection, AV0, multiplied by the VAT standard rate, r.
  - TEG: Tax expenditure gap. VAT revenue forgone because of VAT expenditure provisions included in each country’s official tax expenditure report as an approximation.
  - NTG: Nontaxable goods and services gap. Equal to the difference between PV3 and PV2. Estimated using final consumption of three sectors: public administration, public education, and public health.
  - AV0: Official actual VAT collection. Source: IMF’s FAD C-efficiency data set.
  - r: Standard VAT rate. Source: IMF’s FAD C-efficiency data set.
  - FC: Final Consumption. Source: IMF’s WoRLD data set.
- Key identity derivations (exact expressions as presented):
  - AV = PV3 − (Γ × PV1) − (P × PV3)                            (1)
  - Ec = 1 − Γ (1−P) − P                                                                        (5)
  - Ec = (1 − Γ) x (1−P)                                                                        (6)

### Appendix 2 — Standard RA-GAP Assessment Mission
- RA-GAP objective:
  - Estimate the gap between potential and actual tax revenues to identify areas of noncompliance and inefficiencies in tax policy and administration.
- What is an RA-GAP mission?
  - Structured engagement between IMF experts and national revenue authorities.
  - Focuses commonly on VAT, Corporate Income Tax, Personal Income Tax, and Excise Taxes.
  - Activities include data collection and analysis, preengagement meetings, intensive work sessions, and follow-up; can be in person or remote.
  - Uses an Assisted Self-Assessment model to foster local ownership and capacity building.
- How RA-GAP is applied:
  - Top-down approach estimating potential revenue using macroeconomic data (GDP, Supply Use Tables, Input-Output matrices) and comparing with actual tax collections.
  - Process elements:
    - Data-readiness assessment.
    - Sectoral analysis to identify compliance gaps across industries.
    - Capacity development through training and collaborative analysis.
    - Integration with tools such as the Tax Administration Diagnostic Assessment Tool, ISORA, and RA-FIT.
  - Tailored to country context and supports reform planning with empirical benchmarks.
- Advantages of RA-GAP missions:
  - Diagnostic Insight: quantifies compliance gaps and reveals systemic weaknesses.
  - Policy Guidance: informs tax policy and administrative reforms with evidence-based recommendations.
  - Capacity Building: enhances local expertise in data analysis and tax gap estimation.
  - Governance Support: provides transparent assessments of tax administration effectiveness.
  - Monitoring and Evaluation: facilitates tracking of reform outcomes over time.

*Source: Authors, based on Morrow, Smart, and Swistak (2019) and authors' application of the Reverse Method as presented in the PDF chapter.*

### APPENDIX 3. Applying the Reverse

### APPENDIX 3. Applying the Reverse

### Method’s Accounting Approach under a Country-Specific Purpose
- The methodological option of the Reverse Method requires precise measurements of Ec, the tax expenditure gap (TEG), PV2, the nontaxable gap (NTG), and PV3.
- With these values, it is possible to directly input them and obtain a reverse accounting estimate of the compliance gap.
- Especially key information inputs are listed below.

### Key information inputs required
1. A precise accrual measurement of the C-efficiency:
   - yAccrual value-added tax (VAT) collection
   - yNational Accounts VAT collection
   - yPrecise portion of the National Accounts VAT collection embedded in the Final Consumption

2. VAT expenditure report based on a à la Keen benchmark (final consumption, destination principle, standardized rate) to measure tax expenditure gap (TEG)
   - yIncluding provisions not reported when a different VAT benchmark is applied (for example, financial services, final consumption related to dwelling ownership, negative provisions, and so on)
   - yAdjusting provisions to full compliance, if needed
   - yAdjusting for joint effects (all provisions), if needed

3. National accounts for final consumption for the economic sectors O, P, and Q (typically available from Supply Use Tables) to measure nontaxable gap (NTG):
   - ySector O: Public administration and defense
   - ySector P: Public education
   - ySector Q: Public health

### Descriptive statistics (Appendix 4)

- Appendix Table 4.1 Calibration Model (Table 2 and Figure 1)
  - CountMeanMedianStdMinMax
  - DECE_  
    C  
      
    219.497101.482712.119 5 5. 2216911.112 3 8 4
  - DEC  
    NTG  
    PV2  
    1−  
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
    218.881607.886648.034936.77470 4.938301
  - DEC  
    TEG  
    PV2  
    1−  
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
    ﷟  
      
      
    208.879166.88128.027381.798921.94 4712
  - DEC_(1 − Γ) RA-GAP
    - 218.699076. 698 674.121945. 376757.955201
  - Source: Authors.
  - Note: DEC indicates variables expressed in decimals. RA-GAP indicates figures from a RA-GAP tax gap assessment mission.  
    RA-GAP = Revenue Administration Gap Analysis Program.

- Appendix Table 4.2 Applying the Calibration Model to All Countries (Figures 2, 3, 4, 5, and 6)
  - CountMeanMedianStdMinMax
  - DECE_  
    C  
      
    1,436.4684.462833.1510 97.0974791.23702
  - DEC  
    NTG  
    PV2  
    1−  
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
    1,436.886219.892201.040755.70 3 9 11.9 9176 2
  - DEC  
    TEG  
    PV2  
    1−  
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
    1,436. 8 8 6 511.885657.031604.784336.974 8 6 6
  - DEC_(1 − Γ) ReverseMethod
    - 1,436.654464. 6 6120 8.128 326.310491.997226
  - Source: Authors.
  - Note: DEC indicates variables expressed in decimals. RA-GAP = Revenue Administration Gap Analysis Program.

_The Reverse Method: A New Indirect Technique for Global VAT Compliance Gap Monitoring — TNM/2025/15, Appendix 3 and Appendix 4_

---


_Source: https://www.imf.org/-/media/files/publications/tnm/2025/english/tnmea2025015.pdf_
