## wpiea2026031-source-pdf

## Source details

**Canonical URL:** [wpiea2026031-source-pdf](https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026031-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2026/english/wpiea2026031-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2026/english/wpiea2026031-source-pdf.pdf.json)

---

### Introduction and key findings
- Fiscal context: "the average tax-to-GDP ratio in emerging economies hovers around 15 percent".
- Role of LTOs: "less than 5 percent of economic entities—predominantly, though not exclusively, corporations—account for more than 40–50 percent of total tax revenues."
- Georgia case timeline:
  - LTO re-established in September 2021, initially covering 203 firms.
  - In January 2024, GRS revised thresholds and expanded coverage to a total of 264 firms.
- Main empirical finding: "We find significant and persistent revenue gains—about 0.4–0.7 percent of GDP annually, despite no change in statutory tax rates, indicating stronger compliance."
- Distribution of gains:
  - "The gains are largest in value-added and withholding taxes."
  - "They are also the largest in sectors with low cash usage and more paper trail."
- Spillovers: "suggestive evidence that shifting scrutiny toward monitoring large firms has a modest impact on reducing compliance among smaller taxpayers."
- Identification strategy: administrative records for all corporate taxpayers from 2017 to 2024 and a weighted difference-in-differences framework exploiting staggered implementation and revised eligibility thresholds (2021 rollout and 2024 threshold revision).

### Institutional context and LTO design in Georgia
- Tax parameters:
  - "distributed profit tax (DPT) ... taxed only when distributed as dividends, at a flat rate of 15 percent (20 percent for financial institutions, and also for the gambling sector since 2025)."
  - "The PIT is a flat rate of 20 percent, the VAT is 18 percent"
- GRS coverage: "The GRS managed about 1.7 million taxpayers as of 2024, including around 150,000 firms subject to the DPT and CIT."
- History and coverage:
  - LTO dismantled in 2010; re-established September 15, 2021.
  - Between September 15, 2021, and January 1, 2024, designation used several thresholds; "over 75 percent of firms qualified on [the VAT] basis."
  - 2024 threshold revision: "expanding coverage by more than 25 percent (a total of 264 firms)."
    - "Specifically, 161 firms remained in the LTO, 103 firms entered, and 42 firms exited."
    - "Firms were given no intimation about the revision of the criteria before its announcement."
- LTO functions and risk framework:
  - Dual mission: assist large taxpayers via enhanced services and strengthen enforcement via data-driven risk analysis.
  - Risk classification: "low, medium, or high risk—or as key taxpayers—based on more than 50 risk indicators."
  - 2023 activity: "identified 67 high-risk companies and flagged potential tax losses exceeding 259 million GEL, primarily due to underreporting."
  - Survey: "98 percent of large taxpayers were satisfied with LTO services, while 80 percent reported that the assigned officer improved their understanding of tax obligations."

### Data, sample, and empirical strategy
- Data universe: "anonymized microdata provided by the GRS, covering the universe of 1.45 million active taxpayers registered in Georgia from 2017 to 2024."
- Focus: firm taxpayers only.
- Cohort definitions:
  - "first cohort" = firms assigned in 2021.
  - "second cohort" = firms assigned under 2024 revision.
- Variables included:
  - LTO selection criteria: VAT turnover, exports, net declared property value, recognized tax assessments, number of employees; sector flags for banking and gambling.
  - Tax assessment data: initial, additional, and final tax assessments (DPT, VAT, WHT).
  - Taxpayer characteristics: sector (NACE), number of establishments, number of legal founders.
  - Enforcement records: audits, required corrections, underpayment letters, tax collection letters.
- Deflators and appendix: "We use the Georgian CPI index to deflate nominal real variables to 2021... The list of all variables used in this analysis is presented in Appendix Table B1."
- Sample restrictions and final analysis sample:
  - Retained only taxpayers with positive VAT turnover; removed non-VAT-eligible or inactive taxpayers.
  - Restricted to LTO and non-LTO taxpayers in the top 5 percent of the VAT turnover distribution (above the 95th percentile).
  - Resulting dataset: "25,292 taxpayer-year observations and 4,215 unique taxpayers, of which 190 and 235 are in the LTO cohorts of 2021 and 2024, respectively."
- Empirical approach:
  - Covariate Balancing Propensity Score (CBPS) to estimate treatment probability and balance covariates.
  - Weighted difference-in-differences (WDID) using CBPS weights to estimate causal effects.
  - WDID regression: y_i,t^k = alpha_i + delta_t + beta^k × (LTO_i × POST_t) + theta X_i,t + epsilon_i,t.
  - Common support restriction: drop taxpayers below 2.5th and above 99.9th percentiles of VAT turnover distribution in treated or control groups.

### Descriptive statistics and LTO sample characteristics
- Size and concentration:
  - VAT turnover for LTO taxpayers is, on average, 16 times larger than for non-LTO taxpayers.
  - Large taxpayers account for 44 percent of total VAT turnover while representing only 4.5 percent of taxpayers in the selected sample.
  - LTO taxpayers’ final total tax assessment is, on average, 10 times higher than for non-LTO taxpayers and accounts for 26 percent of total tax assessments of the selected sample.
- Compliance indicators:
  - LTO taxpayers show lower rates of late declarations.
  - LTO taxpayers show lower rates of correction for the initial VAT assessments.
  - When corrections are requested, the average amount requested by the GRS is either negative for LTO taxpayers—indicating overreported tax liabilities—or lower than for non-LTO taxpayers.
- Sectoral concentration (Appendix Figure A1):
  - Wholesale and retail trade: 39 percent
  - Manufacturing: 18 percent
  - Financial and insurance: 8.9 percent
  - Transportation and storage: 7.4 percent
  - Construction: 6.4 percent
  - Electricity and gas: 5.9 percent
- Selection criteria distribution:
  - VAT turnover is the most distinguishing criterion.
  - "77 percent of LTO taxpayers are classified based on VAT turnover alone."
  - Post-LTO shows rightward shifts in VAT turnover, number of employees, tax assessments, and accrual tax paid among LTO firms.

### Empirical findings — Impact on tax revenues (first cohort)
- WDID benchmark impacts (average effect per taxpayer per year, LTO_i × POST_t):
  - Final total tax assessment: 1.39*** million GEL (standard error (0.10))
  - Final DPT: 0.18*** million GEL (0.03)
  - Final VAT: 0.67*** million GEL (0.09)
  - Final WHT: 0.56*** million GEL (0.03)
  - Initial total tax assessment: 1.31*** million GEL (0.16)
  - Initial DPT: 0.17*** million GEL (0.03)
  - Initial VAT: 0.59*** million GEL (0.09)
  - Initial WHT: 0.58*** million GEL (0.13)
- Component contributions to average increase in final tax assessments:
  - VAT: about 0.67 million GEL
  - WHT: about 0.56 million GEL
  - DPT: about 0.17 million GEL
- Aggregate extrapolation (multiplying per-taxpayer effect by LTO taxpayers in initial database; conservative lower bound):
  - Total effect on final tax assessments: 263.79 million GEL (Total) = 0.43 percent of GDP.
  - By tax type (final assessments):
    - DPT: 33.24 million GEL = 0.05 percent of GDP
    - VAT: 126.48 million GEL = 0.21 percent of GDP
    - WHT: 106.02 million GEL = 0.17 percent of GDP
  - Initial assessments aggregate (million GEL):
    - Total: 249.45; DPT: 32.42; VAT: 111.92; WHT: 109.48
  - Initial assessments aggregate (Percent of GDP):
    - Total: 0.41; DPT: 0.05; VAT: 0.18; WHT: 0.18
- Administrative corroboration:
  - Collection letters to LTO taxpayers decreased from 113 in 2020 to 25 in 2021.
  - Higher tax assessments with fewer collection letters indicate improved compliance.
- LTO administrative costs and net fiscal return:
  - 47 staff in 2021 with average monthly salary of 4,572 GEL.
  - Total personnel costs: about 2.6 million GEL.
  - Assuming other administrative and operating expenses twice personnel costs, total annual costs reach 7.8 million GEL.
  - Total annual costs are equivalent to about 3 percent of the total revenue gain.
- Initial versus final assessments:
  - Average impact on total initial tax assessments: 1.31 million GEL per LTO taxpayer per year.
  - Initial vs final: 1.31 million GEL (initial) vs 1.39 million GEL (final)—most gain captured at initial assessment stage.
  - Difference between final and initial assessments most significant for VAT.

### Role of weighting and selection correction
- Unweighted vs CBPS-weighted estimates:
  - Unweighted total final assessment effect: 1.73 million GEL per taxpayer per year (about 25 percent higher than weighted 1.39 million GEL).
  - Unweighted initial total assessment effect: 1.82 million GEL (nearly 40 percent higher than weighted 1.31 million GEL).
  - In percent of GDP, unweighted total effects are higher by 0.1–0.15 percentage point of GDP than weighted estimates.
- Conclusion: CBPS weighting corrects selection bias and yields more accurate attribution of revenue gains to LTO assignment.

### Combined first and second cohort impacts (treating all firms ever assigned as treated)
- Average effect per taxpayer per year (LTO_i × POST_t), combined cohorts:
  - Final total tax assessment: 1.51*** million GEL (0.09)
  - Final DPT: 0.26*** million GEL (0.02)
  - Final VAT: 0.63*** million GEL (0.08)
  - Final WHT: 0.63*** million GEL (0.03)
  - Initial total tax assessment: 1.54*** million GEL (0.13)
  - Initial DPT: 0.26*** million GEL (0.03)
  - Initial VAT: 0.65*** million GEL (0.08)
  - Initial WHT: 0.65*** million GEL (0.11)
- Aggregate magnitudes (combined cohorts):
  - Total Effect (million GEL), final assessments: 413.83 (Total), 70.98 (DPT), 172.992 (VAT), 173.626 (WHT)
  - Total Effect (million GEL), initial assessments: 424.05 (Total), 70.34 (DPT), 178.80 (VAT), 178.92 (WHT)
  - Total Effect (Percent of GDP), final assessments: 0.67 (Total), 0.12 (DPT), 0.28 (VAT), 0.28 (WHT)
  - Total Effect (Percent of GDP), initial assessments: 0.69 (Total), 0.11 (DPT), 0.29 (VAT), 0.29 (WHT)
- Key pattern: Impact strengthened as LTO matured; difference between final and initial assessments narrowed (from 0.08 million GEL in first cohort to -0.03 million GEL in combined cohorts), indicating gains shifted toward filing-stage compliance.

### Dynamic and subgroup responses to threshold revisions
- Firms that remained in LTO (Panel A; year-by-year final total and initial total in million GEL):
  - Year=2022: Final total 1.52***; Initial total 1.54***
  - Year=2023: Final total 1.73***; Initial total 1.73***
  - Year=2024: Final total 3.48***; Initial total 3.371***
  - Pattern: large and persistent effects with a surge in 2024.
- Drop-out firms (Panel B):
  - Year=2022: Final total 0.86***; Initial total 0.76* 
  - Year=2023: Final total 0.65**; Initial total 0.53
  - Year=2024: Final total 0.53*; Initial total 0.40
  - Interpretation: exit associated with partial erosion of reporting, with estimated revenue loss of about 0.1 percent of GDP relative to continued LTO behavior.
- Drop-in firms (Panel C; newly added in 2024):
  - Year=2022: Final total 0.37; Initial total 0.58
  - Year=2023: Final total 0.981***; Initial total 1.25***
  - Year=2024: Final total 3.01***; Initial total 3.29***
  - Interpretation: entry triggers large, immediate adjustments (total final assessments rising to 3.01 million GEL per taxpayer per year).

### Mechanisms: enforcement, voluntary compliance, and administrative efficiency
- Conceptual channels: deterrence via higher perceived audit/detection probability and penalties; facilitation via advisory services and dedicated officers lowering compliance costs.
- Empirical indicators:
  - Accrual tax paid rises significantly for LTO taxpayers.
  - Probability of receiving collection letters increases in first years after assignment, but number of collection orders fell from 113 in 2020 to 25 in 2021.
  - Probability of filing additional (corrective) assessments falls significantly, indicating more accurate upfront reporting.
  - Probability of audit falls significantly after LTO assignment, even as reported liabilities and payments rise.
- Net effect: higher revenue collected with more targeted enforcement and lower marginal enforcement costs—compliance improved while traditional audits became less frequent.

### Triple-difference estimators and enforcement heterogeneity (Appendix Table B10)
- Triple-difference model includes interactions: LTO × POST × ENF (ENF = indicator for experiencing enforcement: underpayment letter, collection letter, audit).
- Baseline conditional LTO × POST effects on final tax assessments (conditional on enforcement type):
  - Under-payment letter (columns 1–4): LTO × POST = 2.02*** (Total), 0.30*** (DPT), 1.12*** (VAT), 0.64*** (WHT).
  - Tax collection letter (columns 5–8): LTO × POST = 1.50*** (Total), 0.19*** (DPT), 0.72*** (VAT), 0.60*** (WHT).
  - Being audited (columns 9–12): LTO × POST = 1.25*** (Total), 0.18*** (DPT), 0.44*** (VAT), 0.64*** (WHT).
- Triple-interaction (LTO × POST × ENF) coefficients:
  - Under-payment letter: -1.44*** (Total), -0.26*** (DPT), -1.03*** (VAT), -0.19** (WHT).
  - Tax collection letter: -0.75** (Total), -0.11 (DPT), -0.40 (VAT), -0.25** (WHT).
  - Audited: 1.18*** (Total), -0.003 (DPT), 1.52*** (VAT), -0.34*** (WHT).
- Interpretation:
  - LTO taxpayers generally report higher final tax assessments across major tax types.
  - Negative triple-interaction coefficients indicate taxpayers who attract enforcement had underreporting already detected/corrected via letters/audits, leaving smaller incremental assessment increases post-LTO compared with other LTO taxpayers.
  - Positive triple-interaction for some audited specifications (e.g., VAT) indicates heterogeneity across enforcement channels.

### Sectoral heterogeneity (triple-difference interacting LTO × POST × NACE)
- Notable sectoral magnitudes (Appendix Tables B11–B12; exact table values preserved in source):
  - Information and Communication:
    - WHT rises by 11.7 million GEL per taxpayer per year.
    - DPT rises by 0.9 million GEL per taxpayer per year.
  - Transportation and Storage:
    - VAT increases by 14.6 million GEL per taxpayer per year.
    - DPT increases by 0.6 million GEL per taxpayer per year.
  - Water Supply, Sewerage, and Waste Management: tax assessments rise across categories, particularly VAT and WHT.
  - Arts, Entertainment, and Recreation: strong growth in tax assessments.
  - Electricity, Gas, Steam, and Air Conditioning: moderate increases in DPT.
  - Sectors with weaker or negative LTO effects:
    - Wholesale and Retail Trade: DPT declines by –2.4 million GEL per taxpayer per year.
    - Financial and Insurance Activities: lower WHT assessments.
    - Construction: significant reductions in DPT and WHT.
    - Human Health and Social Work Activities: smaller tax assessments, notably in DPT and WHT.
- Interpretation: LTO oversight most effective in formalized, capital-intensive sectors with transparent revenue flows and third-party reporting; industries with greater scope for income shifting or deductible expenses display muted or negative responses.

### Robustness, dynamics, and interpretation
- Pre-treatment balance: weighted pre-treatment differences shrink and lose statistical significance.
- Time pattern:
  - For first cohort, total final assessment effect grows from 1.13 million GEL in 2022 to 1.93 million GEL in 2024.
  - For combined cohorts, effects surge in 2024 to roughly 2.4 million GEL per taxpayer per year.
- Auxiliary exercises: results robust to alternative sample definitions, weighting schemes, placebo tests, and restricting to the top 2 percent of taxpayers.
- Anticipation and avoidance: discussed as possible behaviors; dynamic analyses around 2024 thresholds help disentangle these effects.
- Aggregate conclusion:
  - Enhanced oversight of large taxpayers generated sizeable and durable fiscal gains, equivalent to about 0.4–0.7 percent of GDP.
  - Effects strengthen as the institution matures; persistence in the second cohort validates recurring compliance improvements when oversight extends to new firms.
  - Mechanisms combine improved compliance, more selective enforcement, and earlier voluntary reporting.

### Policy implications and recommendations
- Establishing an LTO aligns incentives via credible enforcement and tailored taxpayer services, constituting an investment in state fiscal capacity.
- Targeted institutional reforms—risk-based compliance management, segmentation of taxpayer populations, and better use of data—can yield revenue gains comparable to large policy changes at lower economic cost.
- Well-designed LTOs are presented as a scalable, cost-effective policy to deliver durable improvements in compliance and state capacity, particularly relevant for many emerging economies seeking to raise revenues.

*Italic source: Who Pays When Tax Administration Improves? Working Paper No. WP/2026/031*

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

### wpiea2026031-source-pdf - References .............................................................................................................

### Introduction and key findings
- Fiscal context: "the average tax-to-GDP ratio in emerging economies hovers around 15 percent" relative to advanced economies, constraining public investment and fiscal space.
- Role of LTOs: "less than 5 percent of economic entities—predominantly, though not exclusively, corporations—account for more than 40–50 percent of total tax revenues."
- Georgia case: An LTO was re-established in September 2021, initially covering 203 of the country’s largest firms. In January 2024, the GRS revised thresholds and expanded coverage to a total of 264 firms.
- Main empirical finding (from paper summary): "We find significant and persistent revenue gains—about 0.4–0.7 percent of GDP annually, despite no change in statutory tax rates, indicating stronger compliance."
- Distribution of gains: "The gains are largest in value-added and withholding taxes. They are also the largest in sectors with low cash usage and more paper trail."
- Spillovers: The paper reports "suggestive evidence that shifting scrutiny toward monitoring large firms has a modest impact on reducing compliance among smaller taxpayers."
- Identification strategy: uses administrative records for all corporate taxpayers from 2017 to 2024 and a weighted difference-in-differences framework exploiting staggered implementation and revised eligibility thresholds (2021 rollout and 2024 threshold revision).

### Table 1 — LTO share of net revenue (reported)
- Advanced Economies: Number of Obs. 25; Mean 42.2; Median 44.3.
- Emerging Markets: Number of Obs. 72; Mean 53.6; Median 53.0.
- Low-Income and Developing Countries: Number of Obs. 25; Mean 58.1; Median 64.2.
- Source: International Survey on Revenue Administration (ISORA) and authors’ calculations.
- Notes: "Country groups are as per the International Monetary Fund World Economic Outlook (October 2025 vintage). The data presented reflect the most recent figures available for each country; as a result, the reference years may differ across countries."

### Table 2 — Threshold criteria for Large Taxpayer designation (2021 vs. 2024)
- VAT Turnover: 2021 Thresholds > 80 million ; 2024 Thresholds >100 million
- Exports: 2021 Thresholds > 40 million ; 2024 Thresholds > 50 million
- Declared property value: 2021 Thresholds > 150 million ; 2024 Thresholds > 200 million
- Tax assessments/employment: 2021 Thresholds Tax assessments >20 million or ≥ 1000 employees ; 2024 Thresholds Tax assessments >25 million or ≥ 1300 employees
- Banks: 2021 Thresholds Automatically included ; 2024 Thresholds Automatically included
- Gambling: 2021 Thresholds Paid taxes including gambling duty > 1.5 million GEL ; 2024 Thresholds Paid taxes including gambling duty > 2 million GEL
- Source: Georgia Revenue Service.
- Notes: "All numbers except employment are denominated in GEL."

### Institutional context and LTO design in Georgia
- Tax system parameters reported:
  - "distributed profit tax (DPT) ... taxed only when distributed as dividends, at a flat rate of 15 percent (20 percent for financial institutions, and also for the gambling sector since 2025)."
  - "The PIT is a flat rate of 20 percent, the VAT is 18 percent"
- GRS coverage: "The GRS managed about 1.7 million taxpayers as of 2024, including around 150,000 firms subject to the DPT and CIT."
- History:
  - LTO dismantled in 2010; re-established in September 2021 after commitments in 2017 and delays from COVID-19.
  - Between September 15, 2021, and January 1, 2024, designation used several thresholds (Table 2); "over 75 percent of firms qualified on [the VAT] basis."
- 2024 threshold revision outcomes:
  - "expanding coverage by more than 25 percent (a total of 264 firms)."
  - "Specifically, 161 firms remained in the LTO, 103 firms entered, and 42 firms exited."
  - "Firms were given no intimation about the revision of the criteria before its announcement."
- LTO functions:
  - Dual mission: "to assist large taxpayers in meeting their obligations through enhanced communication, consultation, and tailored services; and second, to strengthen enforcement by identifying high-risk taxpayers using data-driven risk analysis."
  - Risk framework: classifies large taxpayers as "low, medium, or high risk—or as key taxpayers—based on more than 50 risk indicators."
  - 2023 LTO activity: "identified 67 high-risk companies and flagged potential tax losses exceeding 259 million GEL, primarily due to underreporting."
  - Survey evidence: "A 2023 survey found that 98 percent of large taxpayers were satisfied with LTO services, while 80 percent reported that the assigned officer improved their understanding of tax obligations."

### Data, sample, and empirical strategy
- Data universe: "anonymized microdata provided by the GRS, covering the universe of 1.45 million active taxpayers registered in Georgia from 2017 to 2024."
- Focus: "restrict our analysis to firm taxpayers (henceforth “taxpayers”)."
- Cohort definitions: "firms assigned to the LTO in 2021 as the 'first cohort' and those assigned under the 2024 revision of threshold criteria as the 'second cohort'."
- Variables included: (i) LTO selection criteria (VAT turnover, exports, net declared property value, recognized tax assessments, number of employees) and sector flags (banking, gambling); (ii) tax assessment data—initial, additional, and final tax assessments (for DPT, VAT, and withholding (WHT)); (iii) taxpayer characteristics—sector (NACE), number of establishments, number of legal founders; (iv) enforcement records—audits, required corrections, underpayment letters, tax collection letters.
- Deflators and Appendix: "We use the Georgian CPI index to deflate nominal real variables to 2021... The list of all variables used in this analysis is presented in Appendix Table B1."
- Sample restrictions and final analysis sample:
  - Retained only taxpayers with positive VAT turnover; removed non-VAT-eligible or inactive taxpayers (zero or missing VAT turnover).
  - Restricted analysis to LTO and non-LTO taxpayers in the top 5 percent of the VAT turnover distribution (above the 95th percentile).
  - Resulting dataset: "25,292 taxpayer-year observations and 4,215 unique taxpayers, of which 190 and 235 are in the LTO cohorts of 2021 and 2024, respectively."
  - Rationale: "selection into the Georgia LTO depends on criteria other than VAT turnover... assignment to the LTO is likely to be marginal yet maximizes comparability of firms in scale and potential risk profile."
- Empirical approach: weighted difference-in-differences framework accounting for non-random assignment, exploiting staggered implementation (2021 rollout) and revised eligibility thresholds (2024) to identify revenue impacts and dynamic validation of effects.

*Source: wpiea2026031-source-pdf - References (content extracted from the supplied PDF chapter/section).*

### Appendix Table B1 reports descriptive statistics of the selected sample taxpayers, separately for those

### Appendix Table B1 reports descriptive statistics of the selected sample taxpayers, separately for those 

### Descriptive statistics and sample characteristics
- LTO taxpayers are substantially larger than non-LTO taxpayers across all LTO selection criteria.
- VAT turnover for LTO taxpayers is, on average, 16 times larger than for non-LTO taxpayers.
- Large taxpayers account for 44 percent of total VAT turnover, despite representing only 4.5 percent of taxpayers in the selected sample.
- LTO taxpayers’ final total tax assessment is, on average, 10 times higher than for non-LTO taxpayers and accounts for 26 percent of total tax assessments of the selected sample.
- Indicators of compliance:
  - LTO taxpayers show lower rates of late declarations.
  - LTO taxpayers show lower rates of correction for the initial VAT assessments.
  - When tax assessment corrections are requested, the average amount requested by the GRS is either negative for LTO taxpayers—indicating that LTO taxpayers overreported tax liabilities—or lower for LTO taxpayers than for non-LTO taxpayers.

### LTO coverage by sector and selection criteria distribution
- Sectoral concentration of LTO taxpayers (Appendix Figure A1):
  - Wholesale and retail trade: 39 percent
  - Manufacturing: 18 percent
  - Financial and insurance: 8.9 percent
  - Transportation and storage: 7.4 percent
  - Construction: 6.4 percent
  - Electricity and gas: 5.9 percent
- Firms in manufacturing, financial and insurance, and utilities are more likely to be assigned to the LTO; banking firms are required to be part of the LTO regardless of size.
- Distribution of selection criteria (Appendix Figure A2):
  - VAT turnover is the most distinguishing criterion (smallest overlapping region between LTO and non-LTO taxpayers).
  - 77 percent of LTO taxpayers are classified based on VAT turnover alone.
  - Post-LTO period shows a notable rightward shift in the distribution of VAT turnover and the number of employees among LTO firms, more pronounced than for non-LTO firms.
  - A significant rightward shift is also observed in the distribution of tax assessments and accrual tax paid (Appendix Figure A3), suggesting the LTO contributed to raising tax revenues among LTO taxpayers.

### Empirical strategy
- Objective: estimate the additional tax assessment associated with the introduction of the LTO.
- Identification challenges: LTO and non-LTO taxpayers differ systematically (Appendix Tables B1 and B2).
- Estimation approach:
  - Use the Covariate Balancing Propensity Score (CBPS) estimator of Imai and Ratkovic (2014, 2015) to estimate the probability of treatment and to balance covariates.
  - Use a weighted difference-in-difference (WDID), weighting taxpayer-level observations with propensity weights from the CBPS, to estimate causal effects of LTO on tax assessments.
- Treatment and control definitions:
  - Treatment group: taxpayers assigned to the 2021 LTO cohort.
  - Control group: taxpayers never assigned in both the first and second LTO cohorts.
  - When analyzing the second cohort assigned in 2024, treatment is those assigned in 2024 and control is those never assigned.
- Covariates used in propensity score: VAT turnover, exports, net property values, and the number of employees.
  - Excluded criteria: banks and gambling sector (banks automatically assigned to LTO; negligible gambling assignments).
  - Recognized tax assessment criterion excluded to avoid conflation with the objective to analyze LTO impact on tax revenue.
- CBPS properties:
  - Trades off fit accuracy to ensure better covariate balance.
  - Generates weights directly related to the treatment assignment mechanism.
  - After applying CBPS weights, Appendix Table B2 shows perfect balance in observable characteristics of treated and control groups per Rubin’s (2001) criteria.
- WDID regression specification:
  - y_i,t^k = alpha_i + delta_t + beta^k × (LTO_i × POST_t) + theta X_i,t + epsilon_i,t
  - y_i,t^k: k-th outcome (tax, enforcement, or compliance measure) for taxpayer i in year t.
  - LTO_i: indicator equal to 1 if taxpayer is assigned to the LTO.
  - POST_t: indicator equal to 1 in the year a taxpayer is assigned to the LTO and every subsequent year, 0 otherwise.
  - X: firm-level control variables.
  - alpha_i and delta_t: taxpayer and year fixed effects.
  - Beta^k captures the causal average treatment effect of the LTO on the k-th outcome.
- Common support restriction:
  - Drop taxpayers below the 2.5th and above the 99.9th percentiles of the VAT turnover distribution in either treated or control groups.
  - Excludes very small non-LTO taxpayers and very large LTO taxpayers in VAT turnover.
  - Different threshold values and less restrictive criteria applied in robustness checks.

### Empirical findings — Impact on tax revenues of the first LTO cohort
- Outcomes examined: final and initial total tax assessments, and assessments of components (DPT, VAT, WHT).
- WDID regression results (Table 3) — main observations:
  - Positive and statistically significant average treatment effect of the LTO on final and initial tax assessments across total and component taxes.
- Benchmark impact (final tax assessments, column (1)):
  - LTO raised tax assessments by an average of 1.39 million GEL per LTO taxpayer per year.
- Component contributions to the average increase in final tax assessments:
  - VAT: about 0.67 million GEL
  - WHT: about 0.56 million GEL
  - DPT: about 0.17 million GEL
- Supporting administrative evidence:
  - Number of collection letters sent to LTO taxpayers decreased from 113 in 2020 to 25 in 2021.
  - Higher tax assessments with a marked reduction in collection letters indicate vastly improved compliance.
- Aggregate extrapolation to all LTO taxpayers (conservative lower bound assuming excluded larger LTO taxpayers experience same GEL increase):
  - Total effect on final tax assessments: 264 million GEL, or 0.4 percent of GDP.
  - By tax type:
    - VAT: 126 million GEL or 0.21 percent of GDP
    - WHT: 106 million GEL or 0.17 percent of GDP
    - DPT: 33 million GEL or 0.05 percent of GDP
  - Note: these extrapolated “total effect” estimates are partial-equilibrium and subject to the “missing intercept” problem (Wolf, 2023); they do not capture general-equilibrium effects.
- LTO administrative costs and net fiscal return:
  - 47 staff in 2021 with average monthly salary of 4,572 GEL.
  - Total personnel costs: about 2.6 million GEL.
  - Assuming other administrative and operating expenses twice personnel costs, total annual costs reach 7.8 million GEL.
  - Total annual costs are equivalent to about 3 percent of the total revenue gain, indicating strong cost-effectiveness.
- Initial versus final assessments:
  - Average impact on total initial tax assessments (column (5)): 1.31 million GEL per LTO taxpayer per year.
  - Comparison: initial assessment impact of 1.31 million GEL vs final assessment impact of 1.39 million GEL—most revenue gain is captured at the initial assessment stage, with a small subsequent adjustment before finalization.
  - The difference between final and initial assessments is most significant for VAT (column (6)).

*IMF Working Papers — Who Pays When Tax Administration Improves?*

### 0.58 million in the initial assessment versus 0.67 million GEL for the final assessment.

### LTO assignment effects on tax assessments and revenues

### Baseline estimated impacts (first cohort)
- Average effect per taxpayer per year (퐿푇푂푖 × 푃푂푆푇푡):
  - Final total tax assessment: 1.39*** million GEL (standard error (0.10))
  - Final DPT: 0.18*** million GEL (0.03)
  - Final VAT: 0.67*** million GEL (0.09)
  - Final WHT: 0.56*** million GEL (0.03)
  - Initial total tax assessment: 1.31*** million GEL (0.16)
  - Initial DPT: 0.17*** million GEL (0.03)
  - Initial VAT: 0.59*** million GEL (0.09)
  - Initial WHT: 0.58*** million GEL (0.13)
- Aggregate magnitude (multiplying per-taxpayer effect by LTO taxpayers in initial database):
  - Total Effect (million GEL), final assessments: 263.79 (Total), 33.24 (DPT), 126.48 (VAT), 106.02 (WHT)
  - Total Effect (million GEL), initial assessments: 249.45 (Total), 32.42 (DPT), 111.92 (VAT), 109.48 (WHT)
  - Total Effect (Percent of GDP), final assessments: 0.43 (Total), 0.05 (DPT), 0.21 (VAT), 0.17 (WHT)
  - Total Effect (Percent of GDP), initial assessments: 0.41 (Total), 0.05 (DPT), 0.18 (VAT), 0.18 (WHT)
- Pre-treatment balance diagnostics (2017–21):
  - Unweighted pre-treatment difference (Total final assessments): 3.71 (p-value 0.00)
  - Weighted pre-treatment difference (Total final assessments): 0.30 (p-value 0.72)
- Notes on interpretation:
  - Differences between initial and final assessments smaller for WHT (0.58 vs. 0.55 million GEL) and DPT (0.171 vs. 0.175 million GEL).
  - Results indicate LTO firms increasingly comply earlier in the assessment process; tailored taxpayer services likely also contributed.

### Role of weighting and selection correction
- Unweighted estimates overstate effects relative to CBPS-weighted results:
  - Unweighted total final assessment effect: 1.73 million GEL per taxpayer per year (about 25 percent higher than weighted 1.39 million GEL).
  - Unweighted initial total assessment effect: 1.82 million GEL (nearly 40 percent higher than weighted 1.31 million GEL).
  - In percent of GDP, unweighted total effects are higher by 0.1–0.15 percentage point of GDP than weighted estimates.
- Conclusion: taxpayer-level weighting via CBPS corrects selection bias and yields more accurate attribution of revenue gains to LTO assignment.

### Combined first and second cohort impacts (revised thresholds, 2024)
- Average effect per taxpayer per year (퐿푇푂푖 × 푃푂푆푇푡) when treating all firms ever assigned to LTO as treated:
  - Final total tax assessment: 1.51*** million GEL (0.09)
  - Final DPT: 0.26*** million GEL (0.02)
  - Final VAT: 0.63*** million GEL (0.08)
  - Final WHT: 0.63*** million GEL (0.03)
  - Initial total tax assessment: 1.54*** million GEL (0.13)
  - Initial DPT: 0.26*** million GEL (0.03)
  - Initial VAT: 0.65*** million GEL (0.08)
  - Initial WHT: 0.65*** million GEL (0.11)
- Aggregate magnitudes:
  - Total Effect (million GEL), final assessments: 413.83 (Total), 70.98 (DPT), 172.992 (VAT), 173.626 (WHT)
  - Total Effect (million GEL), initial assessments: 424.05 (Total), 70.34 (DPT), 178.80 (VAT), 178.92 (WHT)
  - Total Effect (Percent of GDP), final assessments: 0.67 (Total), 0.12 (DPT), 0.28 (VAT), 0.28 (WHT)
  - Total Effect (Percent of GDP), initial assessments: 0.69 (Total), 0.11 (DPT), 0.29 (VAT), 0.29 (WHT)
- Key pattern:
  - Impact strengthened as LTO matured; the difference between final and initial assessments narrowed (from 0.08 million GEL in first cohort to -0.03 million GEL in the combined cohorts), indicating revenue gains shifted toward filing-stage compliance rather than post-assessment corrections.

### Dynamic and subgroup responses to threshold revisions (year-by-year and group effects)
- Firms that remained in LTO (Panel A):
  - Year=2022: Final total 1.52*** million GEL; Initial total 1.54*** million GEL
  - Year=2023: Final total 1.73*** million GEL; Initial total 1.73*** million GEL
  - Year=2024: Final total 3.48*** million GEL; Initial total 3.371*** million GEL
  - Pattern: large and persistent effects with a surge in 2024, consistent with sustained compliance and growing enforcement capability.
- Drop-out firms (Panel B):
  - Year=2022: Final total 0.86*** million GEL; Initial total 0.76* million GEL
  - Year=2023: Final total 0.65** million GEL; Initial total 0.53 million GEL
  - Year=2024: Final total 0.53* million GEL; Initial total 0.40 million GEL
  - Interpretation: VAT and total final assessments decline but remain somewhat elevated in 2024; exit from LTO associated with partial erosion of reporting, with an estimated revenue loss of about 0.1 percent of GDP relative to continued LTO behavior.
- Drop-in firms (Panel C; newly added in 2024):
  - Year=2022: Final total 0.37 million GEL; Initial total 0.58 million GEL (not significant in many cells)
  - Year=2023: Final total 0.981*** million GEL; Initial total 1.25*** million GEL
  - Year=2024: Final total 3.01*** million GEL; Initial total 3.29*** million GEL
  - Interpretation: entry triggers large, immediate adjustments in declared liabilities (total final assessments rising to 3.01 million GEL per taxpayer per year), especially in DPT and WHT.

### Mechanisms: enforcement, voluntary compliance, and administrative efficiency
- Conceptual channels:
  - Stronger enforcement increases perceived audit/detection probability and penalties (deterrence), reducing evasion.
  - Lower compliance costs via enhanced advisory services and dedicated officers (facilitation) improves voluntary, upfront accurate reporting.
- Empirical indicators (summary of Appendix Table B9):
  - Accrual tax paid rises significantly for LTO taxpayers.
  - Probability of receiving collection letters increases in first years after assignment.
  - Probability of filing additional (corrective) assessments falls significantly, indicating more accurate upfront reporting.
  - Probability of audit falls significantly after LTO assignment, even as reported liabilities and payments rise.
  - Operational outcome: number of collection orders fell from 113 in 2020 to 25 in 2021, signaling improved payment discipline and reduced arrears.
- Net effect: higher revenue collected with more targeted enforcement and lower marginal enforcement costs—compliance improved while traditional audits became less frequent.

### Robustness and validity checks
- Pre-treatment balance: weighted pre-treatment differences shrink and lose statistical significance, supporting the conditional balancing approach.
- Time pattern:
  - For first cohort, total final assessment effect grows from 1.13 million GEL in 2022 to 1.93 million GEL in 2024 (Appendix Table B4, Panel A).
  - For combined cohorts, effects surge in 2024 to roughly 2.4 million GEL per taxpayer per year (Appendix Table B4, Panel B).
- Auxiliary exercises:
  - Results robust to alternative sample definitions, weighting schemes, and placebo tests (Appendix Tables B5–B8 and Figures A4–A5).
  - Restricting to the top 2 percent of taxpayers yields similar or stronger effects.
- Anticipation and avoidance:
  - Possible anticipatory compliance and strategic avoidance are discussed; such behaviors do not threaten WDID consistency but can bias estimated treatment effects downward or upward depending on direction. Dynamic analyses around 2024 thresholds help disentangle these effects.

*Italic source: IMF Working Paper — selected excerpts from the provided PDF content.*

### Appendix Table B10 reports the results of three triple-difference estimator following the methodology of Olden

### wpiea2026031-source-pdf - Appendix Table B10 reports the results of three triple-difference estimator following the methodology of Olden

### Methodology: triple-difference estimator and interpretation
- Model specification (equation (3)):
  - 푦푖.푡푘,푗 = 훼푖 + 훿푡 + 휎푘,푗 ENF푖 + 훽푘,푗 (LTO푖 × POST푡) + 훾푘,푗 (ENF푖 × POST푡) + 휃푘,푗 (LTO푖 × ENF푖) + 휅푘,푗 (LTO푖 × ENF푖 × POST푡) + 휀푖,푡푘,푗
  - 푦푖.푡푘,푗: k-th outcome (one of four tax assessments: total, VAT, DPT, WHT) in the j-th enforcement action (underpayment letter, collection letter, audit) for taxpayer i in year t.
  - ENF푖: indicator for LTO firms identified as high-risk because they witness any enforcement action.
  - LTO푖 and POST푡 are omitted due to taxpayer-level and year fixed effects.
- Interpretation:
  - Coefficient 휅푘,푗 on the triple interaction captures whether the impact of LTO assignment on tax assessment differs for high-risk taxpayers (those experiencing enforcement) relative to other LTO taxpayers.
  - A negative 휅푘,푗 implies taxpayers who received letters or audits experience smaller incremental increases in assessed taxes after LTO assignment relative to other LTO taxpayers — consistent with the idea that enforcement already detected and corrected noncompliance, leaving less room for further increases.

### Key findings from Appendix Table B10 (LTO assignment effects conditional on enforcement measures)
- Baseline LTO × POST effects on final tax assessments (conditional specifications reported across columns):
  - Conditional on receiving an under-payment letter (columns 1–4): LTO × POST = 2.02*** (Total), 0.30*** (DPT), 1.12*** (VAT), 0.64*** (WHT). (Standard errors shown in table.)
  - Conditional on receiving a tax collection letter (columns 5–8): LTO × POST = 1.50*** (Total), 0.19*** (DPT), 0.72*** (VAT), 0.60*** (WHT).
  - Conditional on being audited (columns 9–12): LTO × POST = 1.25*** (Total), 0.18*** (DPT), 0.44*** (VAT), 0.64*** (WHT).
- Triple-interaction (LTO × POST × ENF) coefficients (showing how effect differs for taxpayers who received enforcement):
  - Conditional on under-payment letter (columns 1–4): LTO × POST × ENF = -1.44*** (Total), -0.26*** (DPT), -1.03*** (VAT), -0.19** (WHT).
  - Conditional on tax collection letter (columns 5–8): LTO × POST × ENF = -0.75** (Total), -0.11 (DPT), -0.40 (VAT), -0.25** (WHT).
  - Conditional on being audited (columns 9–12): LTO × POST × ENF = 1.18*** (Total), -0.003 (DPT), 1.52*** (VAT), -0.34*** (WHT).
- Additional conditional and interaction coefficients reported in Table B10 (exact values preserved in source table), including:
  - ENF푖 main effects, ENF푖 × POST푡, and LTO푖 × ENF푖 coefficients for each conditional specification (values appear in Table B10).
- Empirical interpretation emphasized in text:
  - All LTO taxpayers report higher final tax assessments across major tax types, confirming improved compliance under intensive monitoring.
  - Among LTO taxpayers who receive enforcement actions, estimated LTO effects on final assessments are generally smaller (negative and significant interaction coefficients in several specifications).
  - Negative interaction coefficients do not imply enforcement reduces compliance; rather, they indicate that taxpayers attracting enforcement had underreporting already detected/corrected via letters/audits, leaving smaller incremental assessment increases post-LTO compared with other LTO taxpayers.

### Sectoral impacts (triple-difference interacting LTO × POST × NACE sector; equation (4) and Appendix Tables B11–B12)
- Model (equation (4)):
  - 푦푖.푡푘 = 훼푖 + 훿푡 + 훽푘 (LTO푖 × POST푡) + 휅푘 (LTO푖 × NACE푖푗 × POST푡) + 휀푖,푡푘
  - 휅푘 captures whether LTO impact on tax assessment differs by NACE sector.
- Reported sectoral results (Appendix Tables B11 and B12) — pronounced variation across sectors:
  - Information and Communication:
    - WHT rises by 11.7 million GEL per taxpayer per year (table text).
    - DPT rises by 0.9 million GEL per taxpayer per year (table text).
  - Transportation and Storage:
    - VAT increases by 14.6 million GEL per taxpayer per year.
    - DPT increases by 0.6 million GEL per taxpayer per year.
  - Water Supply, Sewerage, and Waste Management:
    - Tax assessments rise across all categories, particularly VAT and WHT (exact magnitudes in Tables B11–B12).
  - Arts, Entertainment, and Recreation:
    - Strong growth in tax assessments (magnitudes in Tables B11–B12).
  - Electricity, Gas, Steam, and Air Conditioning:
    - Moderate increases in DPT.
  - Sectors with weaker or negative LTO effects:
    - Wholesale and Retail Trade: DPT declines by –2.4 million GEL per taxpayer per year (text).
    - Financial and Insurance Activities: lower WHT assessments (text).
    - Construction: significant reductions in DPT and WHT (text).
    - Human Health and Social Work Activities: smaller tax assessments, notably in DPT and WHT (text).
- Interpretation:
  - LTO oversight is most effective in formalized, capital-intensive sectors with transparent revenue flows and third-party reporting.
  - Industries with greater scope for income shifting or deductible expenses (e.g., wholesale/retail, construction) display muted or negative responses.
  - Policy implication: sector-specific enforcement strategies are important to maximize revenue gains while minimizing distortions.

### Aggregate fiscal impact, mechanisms, and dynamics (conclusions)
- Aggregate fiscal gains:
  - Enhanced oversight of large taxpayers generated sizeable and durable fiscal gains, equivalent to about 0.4–0.7 percent of GDP (text).
  - Effects strengthen as the institution matures; persistence in second cohort validates that compliance improvements recur when oversight is extended to new firms.
- Mechanisms behind revenue gains (evidence across channels and robustness tables):
  - Combination of improved compliance, more selective enforcement, and earlier voluntary reporting of tax liabilities.
  - LTO improved compliance not by expanding the number of audits or letters, but by applying enforcement more selectively and focusing on taxpayers already identified as high-risk.
  - Low-risk taxpayers also improved compliance via deterrence and use of GRS-provided services.
- Policy takeaway:
  - Establishing an LTO aligns incentives via credible enforcement and tailored taxpayer services, constituting an investment in state fiscal capacity.
  - Targeted institutional reforms—risk-based compliance management, segmentation of taxpayer populations, and better use of data—can yield revenue gains comparable to large policy changes at lower economic cost.
  - Well-designed LTOs are presented as a scalable, cost-effective policy to deliver durable improvements in compliance and state capacity, particularly relevant for many emerging economies seeking to raise revenues.

*IMF WORKING PAPERS Who Pays When Tax Administration Improves? INTERNATIONAL MONETARY FUND*

### References

### References

### Theoretical foundations and classic models
- Allingham, M. G., & Sandmo, A. (1972). Income tax evasion: A theoretical analysis. Journal of Public Economics, 1, 323–338.
- Becker, G. S. (1968). Crime and punishment: An economic approach. Journal of Political Economy, 76(2), 169–217.
- Festinger, L. (1962). Cognitive dissonance. Scientific American, 207(4), 93–106.
- Morvan, C., & O’Connor, A. (2017). An analysis of Leon Festinger’s A theory of cognitive dissonance. Macat Library.
- Spicer, M. W. (1986). Civilization at a discount: The problem of tax evasion. National Tax Journal, 39(1), 13–20.

### Tax evasion, avoidance, and compliance literature
- Chetty, R. (2009). Is the taxable income elasticity sufficient to calculate deadweight loss? The implications of evasion and avoidance. American Economic Journal: Economic Policy, 1(2), 31–52.
- Feldstein, M. (1999). Tax avoidance and the deadweight loss of the income tax. Review of Economics and Statistics, 81(4), 674–680.
- Pomeranz, D. (2015). No taxation without information: Deterrence and self-enforcement in the value added tax. American Economic Review, 105(8), 2539–2569. https://doi.org/10.1257/aer.20130393
- Slemrod, J., & Yitzhaki, S. (2002). Tax avoidance, evasion, and administration. In A. J. Auerbach & M. Feldstein (Eds.), Handbook of public economics (Vol. 3, pp. 1423–1470). Elsevier.
- Saez, E. (2010). Do taxpayers bunch at kink points? American Economic Journal: Economic Policy, 2(3), 180–212. https://doi.org/10.1257/pol.2.3.180
- Carrillo, P., Pomeranz, D., & Singhal, M. (2017). Dodging the taxman: Firm misreporting and limits to tax enforcement. American Economic Journal: Applied Economics, 9(2), 144–164. https://doi.org/10.1257/app.20140495
- Kleven, H. J., Knudsen, M. B., Kreiner, C. T., Pedersen, S., & Saez, E. (2011). Unwilling or unable to cheat? Evidence from a tax audit experiment in Denmark. Econometrica, 79(3), 651–692. https://doi.org/10.3982/ECTA9113
- Kleven, H. J., & Waseem, M. (2013). Using notches to uncover optimization frictions and structural elasticities: Evidence from Pakistan. The Quarterly Journal of Economics, 128(2), 669–723. https://doi.org/10.1093/qje/qjt004
- Naritomi, J. (2019). Consumers as tax auditors. American Economic Review, 109(9), 3031–3072. https://doi.org/10.1257/aer.20160658
- Besley, T., Jensen, A., & Persson, T. (2023). Norms, enforcement, and tax evasion. Review of Economics and Statistics, 105(4), 998–1007. https://doi.org/10.1162/rest_a_01123
- Chetty, R. (2009). Is the taxable income elasticity sufficient to calculate deadweight loss? The implications of evasion and avoidance. American Economic Journal: Economic Policy, 1(2), 31–52.

### Empirical analyses of tax administration and compliance
- Almunia, M., & Lopez-Rodriguez, D. (2018). Under the radar: The effects of monitoring firms on tax compliance. American Economic Journal: Economic Policy, 10(1), 1–38. https://doi.org/10.1257/pol.20160229
- Asatryan, Z., & Peichl, A. (2017). Responses of firms to tax, administrative and accounting rules: Evidence from Armenia. SSRN. https://doi.org/10.2139/ssrn.3106710
- Basri, M. C., Felix, M., Hanna, R., & Olken, B. A. (2021). Tax administration versus tax rates: Evidence from corporate taxation in Indonesia. American Economic Review, 111(12), 3827–3871. https://doi.org/10.1257/aer.20201237
- Bachas, P., Fisher-Post, M., Jensen, A., & Zucman, G. (2022). Capital taxation, development, and globalization: Evidence from a macro-historical database (NBER Working Paper No. 29819). National Bureau of Economic Research. https://doi.org/10.3386/w29819
- Brockmeyer, A., & Hernandez, M. (2023). Taxation and development: Evidence from administrative tax data (World Bank Policy Research Working Paper No. 10500).
- Basri, M. C., Felix, M., Hanna, R., & Olken, B. A. (2021). Tax administration versus tax rates: Evidence from corporate taxation in Indonesia. American Economic Review, 111(12), 3827–3871. https://doi.org/10.1257/aer.20201237
- Naritomi, J. (2019). Consumers as tax auditors. American Economic Review, 109(9), 3031–3072. https://doi.org/10.1257/aer.20160658
- Pomeranz, D. (2015). No taxation without information: Deterrence and self-enforcement in the value added tax. American Economic Review, 105(8), 2539–2569. https://doi.org/10.1257/aer.20130393

### Tax administration reforms, capacity, and performance (IMF working papers and notes)
- Adan, H., Atsebi, J.-M., Gueorguiev, N., Honda, J., & Nose, M. (2023). Quantifying the revenue yields from tax administration reforms (IMF Working Paper No. 2023/231). International Monetary Fund. https://doi.org/10.5089/9798400258923.001
- Atsebi, J.-M., Gueorguiev, N., & Nose, M. (2025). Enhancing tax capacity: Revenue gains from strengthening tax administration (IMF Working Paper No. 2025/219). International Monetary Fund. https://doi.org/10.5089/9798229029827.001
- Baer, K., Barra, P., & Benitez, J. C. (2025). Closing the gap: How tax administration performance shapes compliance (IMF Working Paper No. 2025/209). International Monetary Fund. https://doi.org/10.5089/9798229027076.001
- Bellon, M., & Warwick, R. (2025). State capacity, institutions and growth: Taxing for takeoff—Revisiting the tax tipping point (IMF Working Paper No. 2025/203). International Monetary Fund. https://doi.org/10.5089/9798229028295.001
- Das, M., De Mooij, R., Hebous, S., & Vellutini, C. (2024). Distributed profit taxes: An analysis of their impact on revenue, investment, and finance (mimeo). International Monetary Fund.
- International Monetary Fund (IMF). (2025). G20 background note on enhancing domestic revenue mobilization through strengthening revenue administration. https://www.imf.org/-/media/files/research/imf-and-g20/2025/g20-background-note-enhancing-domestic-revenue-mobilization-through-strengthening-revenue-administra.pdf

### Country reports, case studies, and administrative data sources
- Georgia Revenue Service (GRS). (2021). 2021 Annual Report. Ministry of Finance, Republic of Georgia. https://www.rs.ge/Media/Default/Docs/2021_Annual_Report.pdf
- Georgia Revenue Service (GRS). (2023). 2023 Annual Report. Ministry of Finance, Republic of Georgia. https://www.rs.ge/Media/Default/Docs/ANGARISHI_2023ENG-1.pdf
- Georgia Revenue Service (GRS). (2024). 2024 Annual Report. Ministry of Finance, Republic of Georgia. https://www.rs.ge/Media/Default/Docs/ANGARISHI_2024_ENG_KDA.pdf
- International Monetary Fund (IMF). (2007). Republic of Armenia: Fifth review under the three-year arrangement under the Poverty Reduction and Growth Facility and request for modification of performance criteria (IMF Staff Country Report No. 07/377). https://doi.org/10.5089/9781451801675.002
- International Monetary Fund (IMF). (2010). Jamaica: 2009 Article IV consultation and request for a Stand-By Arrangement (IMF Staff Country Report No. 10/267). https://doi.org/10.5089/9781455203475.002
- International Monetary Fund (IMF). (2016). Georgia: Technical assessment report—Tax Administration Diagnostic Assessment Tool (TADAT): Performance assessment report (IMF Staff Country Report No. 16/283). https://doi.org/10.5089/9781475531848.002
- International Monetary Fund (IMF). (2018). Georgia: 2018 Article IV consultation and second review under the Extended Fund Facility (IMF Staff Country Report No. 18/198). https://doi.org/10.5089/9781484364055.002

### Methods, estimation, and causal inference
- Hainmueller, J. (2012). Entropy balancing for causal effects: A multivariate reweighting method to produce balanced samples in observational studies. Political Analysis, 20(1), 25–46. https://doi.org/10.1093/pan/mpr025
- Imai, K., & Ratkovic, M. (2014). Covariate balancing propensity score. Journal of the Royal Statistical Society: Series B, 76(1), 243–263. https://doi.org/10.1111/rssb.12027
- Imai, K., & Ratkovic, M. (2015). Robust estimation of inverse probability weights for marginal structural models. Journal of the American Statistical Association, 110, 1013–1022.
- Rubin, D. B. (2001). Using propensity scores to help design observational studies: Application to the tobacco litigation. Health Services and Outcomes Research Methodology, 2(3), 169–188.
- Olden, A., & Møen, J. (2022). The triple difference estimator. The Econometrics Journal, 25(3), 531–553.

### Experimental and behavioral evidence on tax administration
- Khan, A. Q., Khwaja, A. I., & Olken, B. A. (2016). Tax farming redux: Experimental evidence on performance pay for tax collectors. The Quarterly Journal of Economics, 131(1), 219–271. https://doi.org/10.1093/qje/qjv042
- Jenkins, G. P., & Forlemu, E. N. (1993). Enhancing voluntary compliance by reducing compliance costs: A taxpayer service approach (No. 448). Harvard Institute for International Development.
- Besley, T., & Persson, T. (2014). Why do developing countries tax so little? Journal of Economic Perspectives, 28(4), 99–120. https://doi.org/10.1257/jep.28.4.99
- Small, O., & Brown, L. (2020). Taxpayer service provision and tax compliance: Evidence for large taxpayers in Jamaica. Public Finance Review, 48(2), 250–277. https://doi.org/10.1177/1091142119881532
- Naritomi, J. (2019). Consumers as tax auditors. American Economic Review, 109(9), 3031–3072. https://doi.org/10.1257/aer.20160658

### Additional econometric and theoretical contributions
- Becker, G. S. (1968). Crime and punishment: An economic approach. Journal of Political Economy, 76(2), 169–217.
- Wolf, C. K. (2023). The missing intercept: A demand equivalence approach. American Economic Review, 113(8), 2232-2269.
- Kraay, A. (2012). How large is the government spending multiplier? Evidence from World Bank lending. The Quarterly Journal of Economics, 127(2), 829–887. https://doi.org/10.1093/qje/qjs008
- Jenkins, G. P., & Forlemu, E. N. (1993). Enhancing voluntary compliance by reducing compliance costs: A taxpayer service approach (No. 448). Harvard Institute for International Development.

*Who Pays When Tax Administration Improves? Working Paper No. WP/2026/031*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026031-source-pdf.pdf_