## 1. VAT Invoice Management in Latin America

## Source details

**Canonical URL:** [1. VAT Invoice Management in Latin America](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021231-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021231-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021231-print-pdf.pdf.json)

---

### Introduction and reform context
- Invoices document economic transactions and are critical to assess VAT liabilities or verify tax deductions, credits, and refunds.
- Latin American tax administrations have used relatively tight invoice management: i) regulating invoice characteristics, ii) authorizing invoice quantities, and iii) enforcing issuance and book-keeping.
- Dominican Republic reform timeline and key design elements:
  - Norma General 2018-06 issued on February 1, 2018.
  - Non-issued invoices would become void after May 1, 2018.
  - Taxpayers could request new invoices starting on April 2 (large taxpayers), April 9 (small corporations), and April 16 (other taxpayers).
  - First phase (from April 2018): decisions depended on compliance with filing and information reporting obligations.
  - Second phase (starting January 2019): decisions additionally depended on pending payment obligations (tax arrears).
  - For new registered taxpayers, approval granted only for a small number of invoices; period of validity of invoices introduced as two years (January 2019 onward).
- Framework: paper-based invoicing framework at the time, with findings applicable to both paper-based and electronic invoicing frameworks.

### Evaluation strategy and data
- Identification strategy: compare taxpayer outcomes around their invoice request depending on whether the request was approved or rejected.
- Specifications include fixed effects to capture constant taxpayer characteristics and common time trends; main specifications at daily and monthly frequency.
- Second-phase focus: taxpayers with tax arrears; treatment = denied after January 2019; control = approved before January 2019.
- Administrative datasets used:
  - Invoice Request (date and DGII decision);
  - VAT Returns (monthly original and amended);
  - VAT Information Returns (invoices received and issued);
  - VAT Payments (date and amount);
  - Income Tax Returns (annual original and amended).
- Sample period: September 2017 to February 2020.
- Sample sizes:
  - First wave (risk-based rationing introduced): 31,366 taxpayers.
  - Second wave (tax arrears included): 2,924 taxpayers.
  - First-phase analysis random sample: 15,683 taxpayers in treatment and 15,683 taxpayers in control.
  - Second-phase analysis: 1,462 taxpayers in control and 1,462 taxpayers in treatment.

### Main findings — compliance with filing and information reporting obligations
- The reform led to improvements in compliance with filing and information reporting obligations; effects larger in the second phase.
- Key quantified impacts:
  - Probability of past-due filing of VAT returns and information returns rose sharply by almost 10 percentage points on the day of the rejection in the first phase.
  - Results were larger in the second phase (Appendix A notes an improvement in past-due VAT filing of almost 30 percentage points on the day of rejection).
  - Probability of filing income tax returns rose by 4.9 percentage points on the day of rejection (first-phase daily result), and by 5.3 percentage points on day of rejection in the second phase (Appendix A).
  - On-time VAT returns: first phase probability increased by 3.8 percent in the month of rejection and reached 9.6 percent six months later; second phase stabilized around 10 percentage points.
- Information reporting (F606 purchases, F607 sales):
  - Treated taxpayers sharply increased probabilities of reporting past-due information returns immediately upon treatment; effects more than twice as large in the second phase (Appendix B notes around 20 percentage points increases on day of rejection).
  - Medium-term Table 4 figures (percentage-point increases):
    - First Phase — On-Time Reporting of Invoices Received: Average 0.052*** (0.004).
    - First Phase — On-Time Reporting of Invoices Issued: Average 0.187*** (0.003).
    - Second Phase — Past-Due Reporting of Invoices Received: Average 0.193*** (0.011).
    - Second Phase — Past-Due Reporting of Invoices Issued: Average 0.208*** (0.011).
- Mechanisms:
  - Limited approved invoice stocks create recurrent requests, increasing incentives to regularize filings.
  - Enhanced salience: taxpayers perceive increased tax administration attention.
  - Authorities reported improvements in the accuracy of registered taxpayer information.

### Main findings — tax liabilities and behavior
- VAT liability:
  - First phase: output VAT rose more than input VAT in the first two months after rejection, leading to a small increase in net VAT liability, although not significant. Two months after rejection treated taxpayers reported higher output VAT by 12,741 Dominican pesos (USD 257) and higher input VAT by 10,042 Dominican pesos (USD 202).
  - Growth in input VAT accelerated later, suggesting taxpayers sought additional sources to offset tax liability.
  - Second phase: two months after rejection treated taxpayers reported higher output VAT for 96,900 Dominican pesos (USD 1,952), higher input VAT for 38,668 Dominican pesos (USD 779), and higher net VAT for 49,780 Dominican pesos (USD 1,003) on average; output increased faster than input and induced continuous significant improvements in VAT liabilities.
  - Significant increase in probability of amending previously-filed VAT returns in both phases.
- Income tax liability:
  - Income tax liability for SME taxpayers increased by 6.5 percentage points (reported as 7 percent elsewhere; paper notes monetary changes stem from log linear specifications and are referred to as percent for exposition).
  - No significant income tax liability change for large taxpayers.
  - Table 3 selected figures (log-based point estimates and standard errors):
    - Income Tax: Average 0.065* (0.038); Large -0.077 (0.132); SME 0.070* (0.039).

### Main findings — tax collections and payments
- VAT collections and payments:
  - First phase:
    - Probability of VAT payments rose by 2.3 percentage points on the day of rejection; split equally between timely and past-due payments.
    - SME taxpayers increased amounts paid by 2.4 percentage points on average (significant at the 10 percent level); no significant change for large taxpayers.
    - Extrapolated impact for 2018 rejections: 0.02 percent of GDP or around 0.7 percent of VAT collections (paper notes even if all exits explained by reform, net impact on VAT collections would still be positive at around 0.5 percent of total collections).
  - Second phase:
    - Probability of VAT payments increased by 4.3 percentage points among treated taxpayers on the day of rejection, driven entirely by payment of past-due taxes.
    - Probability of payment increase significant for both SME and large taxpayers, with large taxpayers responding more in probability but no significant change in amounts paid.
- Table 2 selected exact figures:
  - First Phase — Probability of Payment: Average 0.001*** (0.000); Large 0.003*** (0.001); SME 0.001*** (0.000).
  - First Phase — Amount Paid (in logs): Average 0.016 (0.013); Large -0.075 (0.051); SME 0.024* (0.013).
  - Second Phase — Probability of Payment: Average 0.005*** (0.000); Large 0.007*** (0.003); SME 0.004*** (0.000).
  - Second Phase — Amount Paid (in logs): Average 0.029 (0.050); Large -0.153 (0.199); SME 0.054 (0.050).
  - Number of Taxpayers (First Phase): Average 31,366; Large 1,466; SME 29,900.
  - Number of Taxpayers (Second Phase): Average 2,924; Large 112; SME 2,812.
  - Number of Observations (First Phase): Average 27,790,276; Large 1,298,876; SME 26,491,400.
  - Number of Observations (Second Phase): Average 2,590,664; Large 99,232; SME 2,491,432.

### Institutional background and motive for reform
- Pre-reform DGII practices:
  - Standardized invoice format, mandatory sequential numbering, differentiated treatment for invoices entitling credits/deductions/refunds (mostly B2B), extensive information reporting obligations, mandatory electronic fiscal devices for final consumer transactions, certification of authorized printing companies.
  - Digitalization of most VAT invoice management procedures and phased introduction of electronic invoicing.
- Main flaw addressed: invoice management conducted in isolation without leveraging taxpayers’ compliance history when authorizing quantities; authorized invoices were valid indefinitely, enabling disappearance from oversight and facilitating fake invoices and fraudulent VAT and income tax practices.
- Reform objective: integrate taxpayer information across core tax functions and actively use risk-based tools in invoice management.

### Distributional, administrative, and robustness considerations
- Exits and distributional impacts:
  - 7,017 taxpayers whose requests were never approved were excluded from analysis; 99 percent were SME taxpayers; more than half reported no taxable sales; between January and March 2018 they contributed less than 0.2 percent of VAT collections.
  - The 1,462 treated taxpayers (second-phase analysis) represent roughly 45 percent of the 34,582 taxpayers treated by the reform (taxpayers whose first request after April 2019 was rejected).
  - Reform appears to have accelerated exit among some taxpayers, predominantly SMEs and agricultural taxpayers — raising concerns about transaction costs and disproportionate impacts on smaller or more informal taxpayers.
- E-invoicing implications:
  - E-invoice regimes can simplify procedures and reduce compliance costs, mitigating negative impacts on small taxpayers and facilitating extension of similar rationing mechanisms.
- Robustness checks:
  - Alternative treatment definitions (two rejected requests; approval for only a small number of invoices) produce similar dynamics and magnitudes.
  - Results robust to industry-by-month time fixed effects and alternative trend specifications.
  - Standard errors clustered by taxpayer; monetary variables winsorized at the 1st and 99th percentiles; growth rates trimmed at the 99th percentile.

### Selected statistics and context (preserved)
- Reform introduced in 2018; invoices not issued before May 2018 were declared void; requests began April 2018.
- Second phase started January 2019.
- Probability of past-due filing rose by almost 10 percentage points on the day of rejection (first phase).
- Probability of VAT payments rose by 2 percentage points on the day of rejection (first phase).
- Probability of VAT payments increased by 4.3 percentage points among treated taxpayers (second phase).
- Income tax liability for SMEs increased by 7 percent (log-linear exposition).
- SMEs increased amount paid by 2.4 percentage points on average during the first phase.
- In the 2014-2017 period, the VAT compliance gap remained around 44 percent with respect to potential collection; the CIT compliance gap remained around 62 percent.
- GDP growth averaged 6.3 percent in 2014-2017.
- Tax arrears in the Dominican Republic reached 7.6 percent of GDP in 2020.

### Policy-relevant implications and conclusion
- Empirical summary:
  - Risk-based rationing of invoices—denying or partially approving invoices for high-risk taxpayers until regularization—led to significant and persistent improvements in filing and reporting compliance; higher VAT collections (quantified as 0.02 percent of GDP or around 0.7 percent of VAT collections extrapolated for 2018 rejections); modest increases in reported VAT and income tax liabilities; and strengthened VAT self-enforcement and information for tax administration enforcement.
- Policy considerations:
  - Use of taxpayers’ compliance history in invoice authorization can produce large compliance gains.
  - Inclusion of payment obligations (arrears) in the rationing algorithm amplifies payment responses, particularly late payments.
  - Need to assess compliance costs and exit risks for SMEs and agricultural taxpayers; e-invoicing can mitigate compliance costs and support equitable implementation.
  - Complement rationing with remediation pathways (resubmission after resolving causes: update registration, file past-due returns, pay or apply for payment plans).
- Broader relevance:
  - Adds to literature on tax enforcement and digitalization reforms; situates risk-based invoice rationing as a viable tool within modern tax administration toolkits, especially when integrated with digital systems.

*Source: IMF working paper — “1. VAT Invoice Management in Latin America” (chapter/section content, wpiea2021231-print-pdf).*

### 1. VAT Invoice Management in Latin America __________________________________________________________ 7

### 1. VAT Invoice Management in Latin America

### Introduction and reform context
- Invoices document economic transactions and are critical to assess VAT liabilities or verify tax deductions, credits, and refunds.
- Latin American tax administrations have used relatively tight invoice management: i) regulating invoice characteristics, ii) authorizing invoice quantities, and iii) enforcing issuance and book-keeping.
- The Dominican Republic introduced risk-based rationing of invoices in 2018.
  - Invoices approved but not issued before May 2018 were declared void; all VAT-registered taxpayers had to request approval for new invoices starting in April 2018.
  - The tax administration rejected invoice requests or approved only a fraction of requested invoices based on taxpayers’ compliance history.
  - First phase (from April 2018): decisions depended on compliance with filing and information reporting obligations.
  - Second phase (starting January 2019): decisions additionally depended on pending payment obligations (tax arrears).
- The Dominican Republic has a paper-based invoicing framework, though it is working on implementing e-invoicing. The findings apply to both paper-based and electronic invoicing frameworks.

### Evaluation strategy and data
- Identification strategy: compare taxpayer outcomes around their invoice request depending on whether the request was approved or rejected.
- Specifications include fixed effects to capture constant taxpayer characteristics and common time trends.
- Main specifications estimated at daily and monthly frequency; monthly specifications allow flexible parametric time trends to differ across treatment and control.
- For the second phase, focus on taxpayers with tax arrears: treatment = denied after January 2019; control = approved before January 2019.
- The paper builds on DGII (2019).

### Main findings — compliance with filing and information reporting obligations
- The reform led to an improvement in compliance with filing and information reporting obligations.
- Key quantified impacts:
  - The probability of past-due filing of VAT returns and information returns rose sharply by almost 10 percentage points on the day of the rejection in the first phase.
  - This improvement was persistent into the medium term.
  - Results were larger in the second phase of the reform.
  - The probability of filing income tax returns also increased sharply.
- Mechanisms discussed:
  - Limited number of approved invoices creates recurrent need for taxpayers to request additional invoices, increasing incentives to comply.
  - Enhanced salience: taxpayers may perceive that the tax administration is paying attention, increasing compliance incentives.
- Authorities reported improvements in the accuracy of registered taxpayer information.

### Main findings — tax liabilities and behavior
- VAT liability:
  - First phase: output VAT rose more than input VAT in the first two months after rejection, leading to a small increase in net VAT liability, although not significant.
  - Growth in input VAT accelerated later, suggesting taxpayers sought additional sources to offset tax liability.
  - Impact on VAT liability was significantly larger in the second phase.
- Income tax liability:
  - Income tax liability for small (SME) taxpayers increased by 7 percent.
  - Income tax liability remained unchanged for large taxpayers.
  - Monetary changes reported stem from log linear specifications and correspond to log percentage points; the paper refers to them as percent for exposition.

### Main findings — tax collections and payments
- VAT collections increased:
  - First phase: probability of VAT payments rose by 2 percentage points on the day of rejection, reflecting both timely and past-due payments.
    - Note: tax arrears were not part of the rationing algorithm in the first phase.
  - Second phase: when tax arrears were included in the rationing algorithm, the probability of VAT payments increased by 4.3 percentage points among treated taxpayers, entirely due to payment of past-due taxes.
  - Small (SME) taxpayers increased their amount paid by 2.4 percentage points on average during the first phase.
  - No significant response in amounts paid among large taxpayers, including in the second phase.

### Institutional background and motive for reform
- Prior to the reform, DGII enforced invoice authorization, issuance, and book-keeping with measures including:
  - Standardized invoice format with minimum information and mandatory sequential numbering.
  - Differentiated treatment for invoices entitling credits, deductions, refunds (mostly B2B).
  - Extensive information reporting obligations for invoices issued and received, mandatory electronic fiscal devices for final consumer transactions, and certification of authorized printing companies.
  - Digitalization of most VAT invoice management procedures and phased introduction of electronic invoicing.
- Main flaw: invoice management was conducted in isolation and did not leverage taxpayers’ compliance history when authorizing invoice quantities.
  - Newly registered taxpayers received authorizations without proper due diligence.
  - Authorized invoices were valid indefinitely, allowing taxpayers to disappear from tax administration oversight after authorization, increasing risk of fictitious credits, deductions, and refunds.
  - Resulted in fake invoices facilitating fraudulent practices in VAT and income tax.
- Reform aimed to integrate taxpayer information across core tax functions and actively use risk-based tools in invoice management.

### Contribution to literature and policy implications
- Adds to literature on tax enforcement and digitalization reforms (electronic fiscal devices, e-invoicing) showing positive impacts on compliance and tax liabilities.
- Fills gap by studying explicit use of taxpayers’ information for invoice rationing; shows that active use of information can lead to large gains in compliance.
- Situates reform within broader calls for risk-based tools as core components of modern tax administration.

### Selected statistics and context (preserved as reported)
- Reform introduced in 2018; invoices not issued before May 2018 were declared void; requests began April 2018.
- Second phase started January 2019.
- Probability of past-due filing rose by almost 10 percentage points on the day of rejection (first phase).
- Probability of VAT payments rose by 2 percentage points on the day of rejection (first phase).
- Probability of VAT payments increased by 4.3 percentage points among treated taxpayers (second phase).
- Income tax liability for SMEs increased by 7 percent.
- SMEs increased amount paid by 2.4 percentage points on average during the first phase.
- In the 2014-2017 period, the VAT compliance gap remained around 44 percent with respect to potential collection; the CIT compliance gap remained around 62 percent.
- GDP growth averaged 6.3 percent in 2014-2017.

*Source: IMF working paper — “1. VAT Invoice Management in Latin America” (chapter/section content).*

### Box 1. VAT Invoice Management  in Latin America  (cont.)

### Box 1. VAT Invoice Management in Latin America (cont.)

### Background and Legal Framework
- Countries implementing invoice authorization/registration controls include: Chile (1976), Bolivia (1987), Mexico (1989), Argentina and Peru (1992), Brazil (1995), Panama (1996), Venezuela (1999), Ecuador (2002), Colombia (2005), and the Dominican Republic (2007) (Cardoza and Aybar, 2014).9
- Tax procedure codes (or equivalent legislation) grant broad powers to tax administrations to regulate authorization, issuance, and registration of invoices.
- Legal frameworks for VAT and income tax explicitly provide for repudiation of tax credits, deductions, and refunds when supporting invoices do not comply with regulations issued by tax administrations.
- Effects noted:
  - Contributed to the fight against tax evasion.
  - Improved documentation and accounting practices across the economy.
  - Strengthened legal claims on rights and obligations derived from agreements between parties.

### Description of the Reform (Dominican Republic)
- Timeline and legal enactment:
  - The reform was formally enacted by Norma General 2018-06 issued on February 1, 2018.10
  - Non-issued invoices would become void after May 1, 2018.
  - Taxpayers could request new invoices starting on April 2 (large taxpayers), April 9 (small corporations), and April 16 (other taxpayers).
- Early 2018 actions by the DGII:
  - DGII canceled every non-issued invoice, forcing new applications for all registered taxpayers.10
  - All procedures were digitalized; applications for new invoices started to be pre-validated online.
- Pre-validation checks included verification of:
  - Entitlement to request invoices based on customer types (final consumers, businesses under free-trade zones, non-residents, the public sector, foreign markets, non-registered businesses, among others).
  - Whether type and quantity of invoices requested were commensurate with the taxpayer’s industry and tax obligations.
  - Taxpayer tags: inactive/dormant, not localized, high-risk due to factors such as:
    - i) not being reported on by withholding agents or third-party information providers,
    - ii) filing null returns recurrently,
    - iii) being identified as a shell company,
    - iv) having ignored examination/verification appointments at the DGII in the past,
    - v) having been blocked by DGII’s tax auditors because of irregular use of invoices.
- Key component: rationing invoice quantities based on tax compliance history using a new algorithm combining:
  - Stock of invoices non-issued by the taxpayer (based on information reported by others).
  - Average rate of invoice issuance of previous years.
  - Taxpayer compliance score (between 0 and 1), reflecting assessment of compliance with VAT and income tax filing obligations (tax returns and information returns of invoices issued due to sales and received due to purchases).
- January 2019 (second phase):
  - Compliance with payment obligations, including instalment plans, was incorporated into the algorithm.11
  - For new registered taxpayers, approval granted only for a small number of invoices as they were classified as high-risk.
  - Introduced a period of validity of two years for all invoices.
- Rationing based on arrears (second phase) rationale and considerations:
  - Novel strategy by the DGII aiming to increase tax collection and provide incentives to pay on time.
  - Potential concern: could unduly burden taxpayers with weak cash flows.
  - Context: tax arrears growing in the Dominican Republic, reaching 7.6 percent of GDP in 2020; other structural reforms would take time to address these challenges.11
- Remediation pathway:
  - If an invoice request is rejected, taxpayers can resubmit after resolving the rejection cause (update registration, file past-due tax returns, submit information returns, amend VAT or income tax returns, pay or apply for payment plans for tax arrears, etc.).

### Impact on Invoice Approvals and Rejections
- Pre-reform: no applications for invoices were rejected; all invoices requested were approved.
- Post-reform (April 2018 onward):
  - Number of rejections soared; DGII approved only a fraction of requested invoices.
  - On April 9, 2018 (when small corporations could request new invoices), the number of rejections peaked at 3,642.
  - Rejections declined after the peak but remained significant after May 2018.
  - Share of invoices approved (as percent of those requested) declined to about 20 percent in the first weeks of the reform and recovered afterwards, oscillating between 40 and 50 percent ever since, well below the 100 percent pre-reform level.
- Figure notes:
  - Panel A shows number of applications for invoices rejected by the DGII each day.
  - Panel B shows average across taxpayers of the number of approved invoices as a percent of requested invoices each day (only includes taxpayers whose request was approved).

### Data Used for Evaluation
- Administrative data provided by the DGII covering variables related to treatment, compliance, payments, and tax liabilities.
- Datasets used:
  - Invoice Request: each invoice request including date and DGII decision (identifies rejection timing).
  - VAT Returns: all monthly VAT returns filed (original and amended), including date of filing, fiscal period, and amounts of input, output, and net VAT.
  - VAT Information Returns: date of filing and period for reporting invoices received (purchases) and invoices issued (sales).
  - VAT Payments: date and amount of each VAT payment made by the taxpayer.
  - Income Tax Returns: all annual income tax returns filed (original and amended), including date of filing, fiscal period, and amounts of revenues, expenditures, net income, and income tax liability.
- Sample period: September 2017 to February 2020.
- Sample sizes:
  - First wave of the reform (risk-based rationing introduced): 31,366 taxpayers.
  - Second wave (tax arrears included in rationing algorithm): 2,924 taxpayers.
- Table 1 summary statistics (selected preserved figures and definitions):
  - For the first phase, random sample drawn for analysis: 15,683 taxpayers in treatment and 15,683 taxpayers in control (random set due to data processing limitations).12
  - For the second phase, 1,462 taxpayers in control and 1,462 taxpayers in treatment (as presented in Table 1).
  - Share of SME taxpayers: 95.6, 95.0, 96.4, 95.9 (columns reflect control/treatment across phases as in Table 1).
  - Sector shares (examples): Share of taxpayers in services: 84.2, 83.7, 79.7, 78.6.
  - Taxable sales (in Dominican pesos): 10th percentile: 25,923; 25,031; 27,188; 26,000. Median: 196,553; 189,929; 193,433; 180,064. Mean: 681,754; 689,114; 751,904; 705,461. 90th percentile: 1,529,129; 1,490,072; 1,481,179; 1,276,596.
  - Monthly probability of filing timely VAT returns (before treatment): 0.44; 0.43; difference -0.004; and for the second phase: 0.25; 0.18; difference -0.069***.
  - Past-due VAT returns (before treatment): 0.20; 0.20; difference -0.005**; second phase: 0.13; 0.20; difference 0.067***.
  - Monthly probability of filing on-time information returns on invoices issued (before treatment): 0.29; 0.09; difference -0.202***; second phase: 0.14; 0.07; difference -0.062***.
  - Invoices received (on-time): 0.33; 0.31; difference -0.016***; second phase: 0.17; 0.09; difference -0.086***.
  - Monthly probability of filing past-due information returns on invoices issued: 0.19; 0.32; difference 0.131***; second phase: 0.13; 0.22; difference 0.083***.
  - Invoices received (past-due): 0.26; 0.23; difference -0.026***; second phase: 0.16; 0.23; difference 0.076***.
- Notes on Table 1:
  - The first three columns reflect the first phase (April 2018). The last three columns correspond to the second phase (January 2019).
  - Stars in “Difference” columns refer to statistical significance. * p<0.1 ** p<0.05 *** p<0.01.
  - For the first phase, treatment group = taxpayers whose first request for invoices after the reform was rejected. Control group = taxpayers whose requests were not rejected, selected via propensity score matching.
  - For the second phase, treatment group = taxpayers with outstanding tax arrears whose first request after January 2019 was rejected. Control group = taxpayers with outstanding tax arrears who requested invoices before January 2019 and whose request was approved (selected via propensity score matching).
  - Large taxpayers include those classified by the DGII as “grande nacional” and “grande local.” SME taxpayers are classified as “normal” by the DGII.

### Identification Strategy
- Treatment definition:
  - First phase: taxpayer treated if their first request for invoices under the new scheme was rejected. Random set of 15,683 taxpayers drawn for analysis due to data processing limitations.12
  - Alternative treatment definitions (Appendix): approval for only a small number of invoices or more than one rejection; results robust to these definitions.
- Control group selection:
  - Taxpayers whose first request was approved used as control.
  - Control group selected via propensity score matching on size, industry, region, total sales, and taxable sales to control trends across industries and taxpayer types.
  - Table 1 first two columns show treatment and control are fairly similar on matched dimensions; they differ on pre-reform compliance with tax obligations (treatment group less likely to comply).
- Second phase focus:
  - Focus on taxpayers with outstanding tax arrears directly affected by inclusion of arrears in the rationing algorithm in January 2019.

*Source: wpiea2021231-print-pdf - Box 1. VAT Invoice Management in Latin America (cont.).*

### 2019. Taxpayers with tax arrears that requested invoices starting in  January 2019 and had their

### 2019. Taxpayers with tax arrears that requested invoices starting in January 2019 and had their requests rejected constitute the treatment group

### Methodology
- Treatment group: 1,462 taxpayers whose requests for invoices starting in January 2019 were rejected.
- Control group: sample of taxpayers with tax arrears that requested invoices before January 2019 and whose requests were approved; propensity score matching used to balance size, industry, region, total sales, and taxable sales.
- Fixed-effects specification:
  - 푖푖 identifies the taxpayer and 푡푡 the time period (day or month).
  - Indicator 푇푇푇푇푇푇푇푇푡푡푇푇푇푇푖푖,푡푡+푗푗 = 1 if taxpayer 푖푖 was treated at time 푡푡+푗푗; normalization assumes 훽훽
−1 = 0.
  - 휅휅푖푖 taxpayer fixed effect; 휏휏푇푇(푡푡) month fixed effect; 휀휀푖푖,푡푡 error term.
  - Controls X include day-of-week dummy (daily) or cubic trend interacted with treatment dummy (monthly); linear, quadratic, and no trends also explored with no significant change.
- Outcomes Y include:
  - compliance with filing and information reporting obligations;
  - whether it pays VAT and amount paid;
  - monetary variables in monthly VAT returns: input, output, and net VAT;
  - monetary variables in annual income tax returns: revenues, expenditures, net income, and tax liability.
- Data processing:
  - Standard errors clustered at the taxpayer level.
  - Variables in Dominican pesos winsorized at the 1st and 99th percentiles.
  - Growth rates trimmed at the 99th percentile.
- Sample notes:
  - 7,017 taxpayers whose requests for invoices were never approved were excluded from analysis; 99 percent were SME taxpayers; more than half reported no taxable sales; between January and March 2018 they contributed less than 0.2 percent of VAT collections.
  - Total number of active taxpayers in 2018 in the Dominican Republic was 184,360.
  - The 1,462 treated taxpayers represent roughly 45 percent of the 34,582 taxpayers treated by the reform (the taxpayers whose first request for invoices after April 2019 was rejected).

### Results — Overview
- Two phases of the rationing reform:
  - First phase: implementation of risk-based rationing in April 2018.
  - Second phase: inclusion of tax arrears in the rationing mechanism in January 2019.
- Outcomes evaluated: filing, payment, reporting of tax liabilities, and information reporting (invoices issued and received).

### A. Filing Obligations — Key Findings
- Past-due VAT returns:
  - Probability of filing at least one past-due VAT return increased by 9.7 percentage points the day the taxpayer’s request was rejected in the first phase.
  - This represents almost 50 percent of the average monthly probability of filing past-due returns by the control group.
  - Small but statistically significant pre-trend in the two weeks before rejection; sharp increase on day of treatment; effect becomes non-significant three days after rejection.
  - Second phase: similar dynamics but larger magnitudes.
- Income tax returns:
  - Probability of filing an original income tax return rose by 4.9 percentage points on the day of rejection and remained around 1 percentage point higher and statistically significant for the two weeks after rejection.
  - Improvement on impact slightly larger in second phase but became insignificant soon afterwards.
- On-time VAT returns:
  - First phase: probability of filing on-time VAT returns increased by 3.8 percent in the month of rejection; reached 9.6 percent six months later.
  - Second phase: sharper increase on impact, stabilized around 10 percentage points.
- Medium-term past-due VAT filing:
  - Probability rose by 10.1 percentage points during the month of rejection, fell by 1.9 percentage points the following month and 1.4 percentage points the next month; became not significant afterwards.
  - Second phase: larger impact on impact and remained positive and significant over the medium term at smaller magnitudes.

### B. Payment Obligations — Key Findings and Magnitudes
- First phase:
  - On the day of rejection, treated taxpayers became 2.3 percentage points more likely to make a VAT payment than control taxpayers.
  - Improvement split equally between timely and late payments.
  - Dynamics:
    - Timely payments rose by 1 percentage point on day of rejection, peaked at 2.4 percentage points 11 days after rejection, then turned negative and significant by end of two weeks (likely due to bunching).
    - Late payments rose by 1 percentage point on day of rejection and remained positive for a few days at declining magnitudes.
  - Heterogeneity by size:
    - Probability of payment increase significant for both large and SME taxpayers.
    - Amount paid increased only among SME taxpayers by 2.4 percentage points (significant at the 10 percent level); no significant change in amount paid for large taxpayers.
  - Extrapolated impact:
    - Extrapolating the increase in probability of payments and amounts paid to all taxpayers treated and all rejections in 2018 yields 0.02 percent of GDP or around 0.7 percent of VAT collections.
    - Footnote: Even if all exits were explained by the reform, net impact on VAT collections would still be positive at around 0.5 percent of total collections.
- Second phase:
  - On the day of rejection, treated taxpayers became 4.3 percentage points more likely to make a VAT payment than control taxpayers, more than twice the first phase response.
  - Improvement driven entirely by late payments.
  - Probability of payment increased and was significant for both SME and large taxpayers, with large taxpayers responding more.
  - No significant impact on amount paid for any group.

- Table 2 summary (selected exact figures preserved):
  - First Phase — Probability of Payment: Average 0.001*** (0.000); Large 0.003*** (0.001); SME 0.001*** (0.000).
  - First Phase — Amount Paid (in logs): Average 0.016 (0.013); Large -0.075 (0.051); SME 0.024* (0.013).
  - Second Phase — Probability of Payment: Average 0.005*** (0.000); Large 0.007*** (0.003); SME 0.004*** (0.000).
  - Second Phase — Amount Paid (in logs): Average 0.029 (0.050); Large -0.153 (0.199); SME 0.054 (0.050).
  - Number of Taxpayers (First Phase): Average 31,366; Large 1,466; SME 29,900.
  - Number of Taxpayers (Second Phase): Average 2,924; Large 112; SME 2,812.
  - Number of Observations (First Phase): Average 27,790,276; Large 1,298,876; SME 26,491,400.
  - Number of Observations (Second Phase): Average 2,590,664; Large 99,232; SME 2,491,432.

### C. Tax Liabilities — Key Findings and Magnitudes
- First phase:
  - Two months after rejection, treated taxpayers reported:
    - Higher output VAT by 12,741 Dominican pesos (USD 257) on average.
    - Higher input VAT by 10,042 Dominican pesos (USD 202) on average.
  - Net impact on tax liability: positive up to the third month and negative up to the fourth month, but not statistically significant.
  - Interpretation: similar increases in input and output VAT may reflect efforts to reduce declared tax liability by collecting further input VAT or over-declaration of input VAT.
- Second phase:
  - Two months after rejection, treated taxpayers reported:
    - Higher output VAT for 96,900 Dominican pesos (USD 1,952) on average.
    - Higher input VAT for 38,668 Dominican pesos (USD 779) on average.
    - Higher net VAT for 49,780 Dominican pesos (USD 1,003) on average per taxpayer; highly statistically and economically significant.
  - Output and input VAT continued to grow in following months, with output increasing faster and inducing continuous and significant improvements in VAT liabilities.
  - Note: In both phases there was a significant increase in the probability of amending previously-filed VAT returns.

- Income tax liability:
  - Mild positive impact for SME taxpayers in the year following rejection:
    - No significant difference in revenues, expenditures, or net income declared on income tax returns overall, but point estimates positive for revenues and net income.
    - Income tax liability rose by 6.5 percentage points more for treated taxpayers than for control taxpayers (significant at the 10 percent level).
    - Effect driven entirely by SME taxpayers; no significant change for large taxpayers.
  - Table 3 selected exact figures:
    - Revenues: Average 0.022 (0.042); Large 0.148 (0.197); SME 0.019 (0.042).
    - Expenditures: Average -0.004 (0.024); Large 0.123 (0.084); SME -0.009 (0.024).
    - Net Income: Average 0.054 (0.042); Large 0.043 (0.125); SME 0.053 (0.043).
    - Income Tax: Average 0.065* (0.038); Large -0.077 (0.132); SME 0.070* (0.039).
    - Number of Taxpayers: Average 30,043; Large 1,407; SME 28,636.
    - Number of Observations: Average 42,481; Large 1,713; SME 40,768.

### D. Information Reporting Obligations — Key Findings
- Universe required to submit informative returns F606 (purchases) and F607 (sales): 234,856 VAT taxpayers.
- Baseline: on-time filing rates of these informative returns are low, particularly for treated taxpayers before the reform.
- Short-term dynamics:
  - Treated taxpayers sharply increased probabilities of reporting past-due information returns on invoices received (purchases) and invoices issued (sales) immediately upon treatment.
  - Effects remained positive and significant for more than 10 days after treatment, then became not statistically significant or weakly negative.
  - Small negative pretrend suggests possible anticipation effects.
  - Second phase produced similar dynamics but more than twice as large.
- Medium-term (Table 4 selected exact figures — increases in probability in percentage points):
  - First Phase — On-Time Reporting of Invoices Received: Average 0.052*** (0.004); Large 0.083*** (0.020); SME 0.050*** (0.004).
  - First Phase — On-Time Reporting of Invoices Issued: Average 0.187*** (0.003); Large 0.284*** (0.019); SME 0.183*** (0.003).
  - First Phase — Past-Due Reporting of Invoices Received: Average 0.028*** (0.004); Large 0.037* (0.020); SME 0.029*** (0.004).
  - First Phase — Past-Due Reporting of Invoices Issued: Average -0.080*** (0.004); Large -0.138*** (0.021); SME -0.077*** (0.004).
  - Second Phase — On-Time Reporting of Invoices Received: Average 0.050*** (0.007); Large 0.026*** (0.044); SME 0.027*** (0.007).
  - Second Phase — On-Time Reporting of Invoices Issued: Average 0.183*** (0.007); Large 0.041*** (0.042); SME 0.039*** (0.007).
  - Second Phase — Past-Due Reporting of Invoices Received: Average 0.193*** (0.011); Large 0.203*** (0.057); SME 0.194*** (0.011).
  - Second Phase — Past-Due Reporting of Invoices Issued: Average 0.208*** (0.011); Large 0.168*** (0.056); SME 0.211*** (0.011).
  - Sample sizes: First Phase — Number of Taxpayers Average 31,366; Large 1,466; SME 29,900. Observations: Average 940,980; Large 43,980; SME 897,000. Second Phase — Number of Taxpayers Average 2,924; Large 112; SME 2,812. Observations: Average 87,720; Large 3,360; SME 84,360.
- Net effect: the reform narrowed the noncompliance gap of treated taxpayers, notably almost fully reversing an initial 20.2 percent gap in timely reporting of invoices issued.

### Policy-relevant Implications and Interpretation
- The rationing reform produced significant and persistent improvements in:
  - compliance with filing and reporting obligations for VAT and income tax;
  - VAT collections (notably via increased probability of payments and amounts paid among SMEs in the first phase and strong increases in late payments in the second phase);
  - reported VAT and income tax liabilities (modest increases, particularly large net VAT increases in the second phase).
- Mechanisms:
  - Strengthened VAT self-enforcing chain by incentivizing taxpayers to collect invoices from purchases.
  - Increased information reporting enhances the tax administration’s ability to crosscheck data, perform inspections, audits, and notifications.
  - Some behavior may reflect over-declaration of input VAT or strategies to reduce declared net liability; increases in amendments to previously filed VAT returns were observed.
- Distributional and administrative considerations:
  - The reform appears to have accelerated exit among some taxpayers (7,017 with never-approved requests), predominantly SMEs and agricultural taxpayers; this points to transaction costs and potential disproportionate impacts on smaller or more informal taxpayers.
  - E-invoice regimes can simplify procedures and reduce compliance costs, mitigating negative impacts on small taxpayers and facilitating extension of similar rationing mechanisms.

### Conclusion
- In the Dominican Republic, risk-based rationing of invoices—denying or partially approving invoices for high-risk taxpayers until regularization—led to:
  - significant and persistent improvements in filing and reporting compliance;
  - higher VAT collections (quantified as 0.02 percent of GDP or around 0.7 percent of VAT collections extrapolated for 2018 rejections);
  - modest increases in reported VAT and income tax liabilities;
  - strengthened VAT self-enforcement and better information for tax administration enforcement.
- The reform’s costs and effects on SME and agricultural taxpayers highlight the importance of assessing compliance costs, especially in high-informality environments.
- The approach is compatible with and can be extended to e-invoice systems, which would reduce compliance costs and administrative burdens.

*Source: IMF working paper content from wpiea2021231-print-pdf.*

### Introduction  of Electronic Tax Filing.”  World Bank Policy Research Working Paper No. 8452.

### Introduction  of Electronic Tax Filing.”  World Bank Policy Research Working Paper No. 8452.

### Appendix A — Filing Obligations in the Second Phase of the Reform
- The second phase led to an immediate improvement in compliance with VAT filing obligations (Figure A.1, Panel A).
- The probability of filing past-due VAT returns shows a dynamic around the time of the rejection similar to the first phase, but magnitudes are larger:
  - First phase improvement: around 10 percentage points.
  - Second phase improvement: almost 30 percentage points.
- Income tax returns:
  - The probability of filing an original income tax return rose by 5.3 percentage points on the day of the rejection, but became not significant from zero afterwards (Figure A.1, Panel B).
- Medium-term persistence:
  - Treated taxpayers increased their timely filing of VAT returns by 9.7 percentage points in the month of the rejection; this improvement remained stable and significant for the following six months (Figure A.1, Panel C).
  - Filing of past-due VAT returns also increased sharply in the month of the rejection; while it declined afterwards, it remained positive and significant (Figure A.1, Panel D).
- Estimation details (short-term, daily frequency):
  - Specification shown as: 푌푌푖푖,푡푡 = ∑ 훽훽푗푗 퐼퐼(푇푇푇푇푇푇푇푇 푡푡푇푇푇푇푖푖,푡푡+푗푗 =1) 17 푗푗=−16 + 휅휅푖푖 + 휏휏𝑇(𝑡) + 𝑋𝑋푖푖,푡푡 + 휀휀푖푖,푡푡, where 휏휏𝑇(𝑡) is a month fixed effect and 𝑋𝑋푖푖,푡푡 includes a day-of-the-week fixed effect. Cumulative effects 훽훽−16 and 훽훽17 not shown.
- Estimation details (medium-term, monthly frequency):
  - Specification shown as: 푌푌푖푖,푡푡 = ∑ 훽훽푗푗 퐼퐼(푇푇푇푇푇푇푇푇 푡푡푇푇푇푇푖푖,푡푡+푗푗 =1) 7 푗푗=−4 + 휅휅푖푖 + 휏휏𝑇(𝑡) + 𝑋𝑋푖푖,푡푡 + 휀휀푖푖,푡푡, where 휏휏𝑇(𝑡) is a month fixed effect and 𝑋𝑋푖푖,푡푡 includes a cubic trend interacted by a dummy identifying treatment/control. Cumulative effects 훽훽−4 and 훽훽7 not shown.
- Note: Standard errors clustered by taxpayer. Dashed lines correspond to 95%-level confidence intervals.

### Appendix B — Information Reporting Obligations in the Second Phase of the Reform
- Compliance with information returns rose by more than in the first phase of the reform (Figure A.2).
- The probability of filing information returns rose sharply on the day of the rejection for treated taxpayers:
  - Increase on the day of the rejection: around 20 percentage points for both returns on invoices received and issued.
  - The response declined soon afterwards.
- Estimation details (short-term, daily frequency):
  - Same daily-frequency specification as in Appendix A: 푌푌푖푖,푡푡 = ∑ 훽훽푗푗 퐼퐼(푇푇푇푇푇푇푇푇 푡푡푇푇푇푇푖푖,푡푡+푗푗 =1) 17 푗푗=−16 + 휅휅푖푖 + 휏휏𝑇(𝑡) + 𝑋𝑋푖푖,푡푡 + 휀휀푖푖,푡푡, where 휏휏𝑇(𝑡) is a month fixed effect and 𝑋𝑋푖푖,푡푡 includes a day-of-the-week fixed effect. Cumulative effects 훽훽−16 and 훽훽17 not shown.
- Note: Standard errors clustered by taxpayer. Dashed lines correspond to 95%-level confidence intervals.

### Appendix C — Robustness
- Coarser definition of treatment:
  - Treated taxpayers defined as those with two requests rejected (taxpayers with only one request rejected are dropped).
  - Dynamics around time of rejection are quite similar to baseline, with point estimates for compliance with filing, payment, and reporting obligations being slightly higher (Figure A.3).
- Broader definition of treatment:
  - Treated taxpayers include those who got fewer than five invoices approved (including taxpayers whose first request was not rejected but who likely returned for additional invoices).
  - Dynamics and magnitudes around time of treatment remain very similar to baseline estimates (Figure A.4).
- Heterogeneous time trends by industry:
  - Specification estimated with industry-by-month time fixed effects; results are broadly unchanged (Figure A.5).
  - The study already selected controls to resemble treatment group by industry, region, and size so common time trend should capture industry-specific dynamics; industry-by-month fixed effects provide a more flexible check.
- Estimation and inference notes repeated across robustness checks:
  - Standard errors clustered by taxpayer. Dashed lines correspond to 95%-level confidence intervals.
  - Daily-frequency specification: 푌푌푖푖,푡푡 = ∑ 훽훽푗푗 퐼퐼(푇푇푇푇푇푇푇푇 푡푡푇푇푇푇푖푖,푡푡+푗푗 =1) 17 푗푗=−16 + 휅휅푖푖 + 휏휏𝑇(𝑡) + 𝑋𝑋푖푖,푡푡 + 휀휀푖푖,푡푡, where 휏휏𝑇(𝑡) is a month fixed effect and 𝑋𝑋푖푖,푡푡 includes a day-of-the-week fixed effect. Cumulative effects 훽훽−16 and 훽훽17 not shown.

*Source: Introduction  of Electronic Tax Filing.”  World Bank Policy Research Working Paper No. 8452. World Bank Group, Washington, DC.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021231-print-pdf.pdf_