## Dining and Wining During the Pandemic? A Quasi-Experiment on Tax Cuts and Consumer Spending in Lithuania

## Source details

**Canonical URL:** [Dining and Wining During the Pandemic? A Quasi-Experiment on Tax Cuts and Consumer Spending in Lithuania](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023188-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023188-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023188-print-pdf.pdf.json)

---

### Abstract and Main Finding
- Research question: Could temporary tax cuts stimulate consumer spending on contact-intensive services during the COVID-19 pandemic?
- Policy examined: Reduction in the standard value-added tax (VAT) rate from 21 percent to 9 percent on restaurants and catering services.
- Key empirical finding: The VAT reduction has had no statistically significant impact on consumer spending on restaurants and catering services.
- Other factors with more pronounced effects: mobility restrictions and vaccination.

### Introduction — Context and Motivation
- Global context: "the number of COVID-19 cases has reached 769.8 million, resulting more than 6.9 million deaths".
- Rationale: Targeted VAT reductions were widely used to support sectors most affected by the pandemic (e.g., accommodation, restaurants, cultural activities).
- Quasi-experiment: Lithuania’s sector-specific VAT cut provides variation to identify causal effects on consumption using a difference-in-differences framework.

### Data Overview
- Transaction data:
  - Source: Daily debit and credit card transactions from Swedbank.
  - Coverage: thirty-three spending categories.
  - Sample period: January 1, 2019 to October 2, 2022.
  - POS data exclude cash withdrawals; include both in-person and online transactions.
- Payment context in Lithuania:
  - "about 94 percent of Lithuanians with a payment account used internet banking and 69 percent used mobile payment applications."
  - Cash payment share estimates: "88.3 percent in 2006 to 68.5 percent in 2016 and 20.2 percent in 2021."
  - Informal economy estimate: "about 20-30 percent of GDP."
- COVID-19 and policy variables:
  - COVID-19 deaths and infections from OxCGRT; indices used: stringency index and economic support index (each 0 to 100).
  - Vaccination rate from Our World in Data.
- Descriptive statistics (selected, exact values preserved):
  - Mean daily debit and credit card transactions on restaurants and catering services: €870,221.
  - Minimum: €18,743; Maximum: €3.6 million.
  - Daily new COVID-19 deaths: mean 7, minimum 0, maximum 79.
  - Daily new COVID-19 infections: mean 1,245, minimum 0, maximum 15,412.
  - Stringency index: mean 37, minimum 0, maximum 87.
  - Economic support index: mean 72, minimum 0, maximum 100.

### Empirical Strategy
- Identification: Difference-in-differences comparing "treatment" (restaurants and catering) to "control" (other consumption categories taxed at the standard VAT rate).
- Model structure:
  - spend_c,t = α + β treat_c + γ vat_t + δ (treat_c * vat_t) + ϑ X_t + η_c + μ_t + ε_c,t
  - treat_c = 1 for restaurants and catering services, 0 otherwise.
  - vat_t = 1 for periods after the VAT reduction on July 1, 2021, 0 otherwise.
  - Interaction coefficient δ is the estimator of interest (treatment effect).
  - Controls X_t include COVID-19 deaths (or infections) per population, stringency index, economic support index, and vaccination rate.
  - Category fixed effects η_c and time effects μ_t included; standard errors clustered at the consumption category level.
- Causal interpretation relies on parallel trends prior to the VAT cut (graphical checks discussed; Figure 1 shows no discernible post-cut divergence).

### Results and Robustness
- Baseline result:
  - Estimated difference-in-differences coefficient δ: positive but statistically insignificant; magnitude about 0.3 percent across specifications.
  - Conclusion: Consumers did not allocate more spending to restaurants and catering after the VAT cut became effective on July 1, 2021.
- Regression estimates (selected coefficients preserved exactly as reported):
  - Baseline (Table 1):
    - Treat*VAT: 0.284 (standard error [0.060]) in columns [1]–[4]; 0.290 (standard error [0.060]) in column [5].
    - COVID-19 deaths per population: −0.024** ([0.003]) in column [1]; −0.022** ([0.003]) in columns [2]–[3]; −0.016** ([0.003]) in column [4].
    - Stringency index: −0.010*** ([0.001]) in columns [3]–[5].
    - Economic support index: 0.000*** ([0.001]) in column [4]; 0.001*** ([0.001]) in column [5].
    - COVID-19 vaccinations per population: 0.000*** ([0.000]) in column [5].
    - Number of observations: 3,018 in all specifications reported.
    - Country FE: Yes; Time FE: Yes across Table 1 specifications.
  - Robustness (Table 2, selected):
    - Treat*VAT: 0.289 ([0.060]) in columns [2]–[4]; 0.290 ([0.060]) in column [5]; 0.259 ([0.107]) in column [6].
    - COVID-19 infections per population: −0.002*** ([0.000]) in one specification.
    - COVID-19 infections: −0.001*** ([0.000]) in one specification.
    - COVID-19 deaths: −0.006*** ([0.001]) in one specification.
    - 7-day moving average of COVID-19 deaths per population: −0.018** ([0.004]) in one specification.
    - Stringency index: −0.007*** ([0.001]) to −0.010*** ([0.001]) across specifications.
    - Economic support index: 0.006*** ([0.001]) to 0.00*** ([0.001]) across specifications.
    - COVID-19 vaccinations per population: 0.000*** ([0.000]) across multiple specifications.
    - Number of observations: 3,018 in all Table 2 specifications.
    - Country FE: Yes in all Table 2 specifications.
    - Time FE: Yes in columns [2]–[5]; No in column [6].
- Additional robustness and identification checks:
  - Using COVID-19 infections instead of deaths and other alternative pandemic measures yields broadly similar results.
  - Alternative specification omitting time fixed effects but including time-varying policy variables yields similar findings.
  - Placebo regressions and graphical diagnostics for parallel trends (Appendix Figure A1) show no evidence for violation of the common pre-policy change trend assumption.
- Price pass-through and aggregate price movements:
  - The dataset does not contain transaction-level price data to calculate the rate of tax pass-through to prices.
  - Aggregate consumer price statistics from Statistics Lithuania:
    - In July 2021 (after the VAT reduction): price increase of 2.2 percent in restaurants and hotels and 1.4 percent in catering services versus 0.5 percent on average.
    - Three months after the VAT reduction: 4.3 percent in restaurants and hotels and 3 percent in catering services vs. 2.6 percent in the headline index.
  - These aggregate figures are consistent with estimations showing the VAT reduction had no significant effect on consumer spending, though other factors could contribute to relative price movements across sectors.

### Interpretation and Policy Implications
- Explanation for null effect:
  - Impact of COVID-19 and mobility restrictions on consumer behavior is far stronger than price incentives from a sector-specific VAT cut during a pandemic with high health and economic uncertainty.
  - Findings are consistent with other quasi-experimental studies on VAT cuts for labor-intensive services.
- Policy implications:
  - Expected stimulative effect of sector-specific VAT reductions during periods of constrained mobility may be limited.
  - Effective fiscal policy to counter pandemic effects likely requires measures that address mobility constraints, public health risks, and direct economic support—not only price-based tax cuts.
  - Design of interventions should consider salience, uncertainty, and the stronger role of non-tax policy levers (e.g., containment measures, vaccination campaigns).

### Key statistics and Cost Estimate
- Key sample moments (preserved exactly from Appendix Table A1):
  - Observations for debit and credit card transactions by category: 4,113 for each listed category.
  - Restaurants and catering services: Mean 870,221; Std. dev. 468,112; Minimum 18,743; Maximum 3,588,394.
  - Food and drinks: Mean 3,223,457; Std. dev. 1,653,542; Minimum 66,070; Maximum 8,855,150.
  - Fuel: Mean 1,174,323; Std. dev. 647,626; Minimum 94,725; Maximum 4,232,528.
  - Hotels and accommodation: Mean 246,620; Std. dev. 168,227; Minimum 2,618; Maximum 47,844.
  - Financial services: Mean 640,015; Std. dev. 308,175; Minimum 2,240; Maximum 2,405, (as reported).
  - Miscellaneous: Mean 1,995,054; Std. dev. 1,924,960; Minimum 132,247; Maximum 11,800,000.
- Pandemic and policy indices (observations = 3,018):
  - COVID-19 deaths: Mean 79; Std. dev. 0; Minimum 0; Maximum 79 (as reported).
  - COVID-19 infections: Mean 987; Std. dev. 1,874; Minimum 0; Maximum 15,412.
  - Stringency index: Mean 37; Std. dev. 21; Minimum 0; Maximum 87.
  - Economic support index: Mean 53; Std. dev. 36; Minimum 0; Maximum 100.
- Cost estimate of the VAT reduction:
  - Reducing the VAT rate from 21 percent to 9 percent on restaurants and catering services is estimated to lower VAT collection by about €133 million (or 0.2 percent of GDP) in 2022.

*Source: IMF Working Paper WP/23/188, Serhan Cevik, September 2023.*

### Section 1

### Dining and Wining During the Pandemic? A Quasi-Experiment on Tax Cuts and Consumer Spending in Lithuania

### Abstract and Main Finding
- Research question: Could temporary tax cuts stimulate consumer spending on contact-intensive services during the COVID-19 pandemic?
- Policy examined: Reduction in the standard value-added tax (VAT) rate from 21 percent to 9 percent on restaurants and catering services.
- Key empirical finding: The VAT reduction has had no statistically significant impact on consumer spending on restaurants and catering services.
- Other factors with more pronounced effects: mobility restrictions and vaccination.

### Introduction — Context and Motivation
- Global context: "the number of COVID-19 cases has reached 769.8 million, resulting more than 6.9 million deaths" (global figures cited).
- Rationale: Targeted VAT reductions were widely used to support sectors most affected by the pandemic (e.g., accommodation, restaurants, cultural activities).
- Quasi-experiment: Lithuania’s sector-specific VAT cut provides variation to identify causal effects on consumption using a difference-in-differences framework.

### Data Overview
- Transaction data:
  - Source: Daily debit and credit card transactions from Swedbank.
  - Coverage: thirty-three spending categories.
  - Sample period: January 1, 2019 to October 2, 2022.
  - POS data exclude cash withdrawals; include both in-person and online transactions.
- Payment context in Lithuania:
  - "about 94 percent of Lithuanians with a payment account used internet banking and 69 percent used mobile payment applications."
  - Cash payment share estimates: "88.3 percent in 2006 to 68.5 percent in 2016 and 20.2 percent in 2021."
  - Informal economy estimate: "about 20-30 percent of GDP."
- COVID-19 and policy variables:
  - COVID-19 deaths and infections from OxCGRT; indices used: stringency index and economic support index (each 0 to 100).
  - Vaccination rate from Our World in Data.
- Descriptive statistics (selected, exact values preserved):
  - Mean daily debit and credit card transactions on restaurants and catering services: €870,221.
  - Minimum: €18,743; Maximum: €3.6 million.
  - Daily new COVID-19 deaths: mean 7, minimum 0, maximum 79.
  - Daily new COVID-19 infections: mean 1,245, minimum 0, maximum 15,412.
  - Stringency index: mean 37, minimum 0, maximum 87.
  - Economic support index: mean 72, minimum 0, maximum 100.

### Empirical Strategy
- Identification: Difference-in-differences comparing "treatment" (restaurants and catering) to "control" (other consumption categories taxed at the standard VAT rate).
- Model (structure preserved conceptually): spend_c,t = α + β treat_c + γ vat_t + δ (treat_c * vat_t) + ϑ X_t + η_c + μ_t + ε_c,t
  - treat_c = 1 for restaurants and catering services, 0 otherwise.
  - vat_t = 1 for periods after the VAT reduction on July 1, 2021, 0 otherwise.
  - Interaction coefficient δ is the estimator of interest (treatment effect).
  - Controls X_t include COVID-19 deaths (or infections) per population, stringency index, economic support index, and vaccination rate.
  - Category fixed effects η_c and time effects μ_t included; standard errors clustered at the consumption category level.
- Causal interpretation relies on parallel trends prior to the VAT cut (graphical checks discussed; Figure 1 shows no discernible post-cut divergence).

### Results and Robustness
- Baseline result:
  - Estimated difference-in-differences coefficient δ: positive but statistically insignificant; magnitude about 0.3 percent across specifications.
  - Conclusion: Consumers did not allocate more spending to restaurants and catering after the VAT cut became effective on July 1, 2021.
- Controls and robustness:
  - Introducing COVID-19 deaths per population: pandemic has a significant negative effect on consumer spending; δ remains statistically insignificant.
  - Adding policy indices: stringency index has a significant negative effect; economic support index has a positive effect on restaurant/catering spending.
  - Adding vaccination rate: vaccination rate has a small positive effect on consumer spending; δ remains statistically insignificant and broadly unchanged in magnitude.
  - Alternative control group restricted to contact-intensive services yields similar insignificant results.
- Timing and policy details preserved:
  - VAT reduction effective July 1, 2021 to last until December 31, 2022; Parliament decision on November 22, 2022 extended the temporary measure until the end of 2023 (as liquidity and support measure).

### Interpretation and Policy Implications
- Explanation for null effect:
  - Impact of COVID-19 and mobility restrictions on consumer behavior is far stronger than price incentives from a sector-specific VAT cut during a pandemic with high health and economic uncertainty.
  - Findings are consistent with other quasi-experimental studies on VAT cuts for labor-intensive services.
- Policy implications:
  - Expected stimulative effect of sector-specific VAT reductions during periods of constrained mobility may be limited.
  - Effective fiscal policy to counter pandemic effects likely requires measures that address mobility constraints, public health risks, and direct economic support—not only price-based tax cuts.
  - Design of interventions should consider salience, uncertainty, and the stronger role of non-tax policy levers (e.g., containment measures, vaccination campaigns).

*Source: IMF Working Paper WP/23/188, Serhan Cevik, September 2023.*

### Section 2

### Section 2

### Main empirical finding
- The difference-in-differences method indicates the VAT reduction has had no discernible positive impact on consumer demand for restaurant and catering services in Lithuania.
- The empirical analysis finds that the reduction in the VAT rate on restaurants and catering services from 21 percent to 9 percent has had no statistically significant impact on consumer spending on restaurants and catering services.
- Other policy interventions such as mobility restrictions and vaccination are found to have more pronounced influence over consumption behavior.

### Price pass-through and aggregate price movements
- The dataset does not contain transaction-level price data to calculate the rate of tax pass-through to prices.
- Aggregate consumer price statistics from Statistics Lithuania:
  - In July 2021 (after the VAT reduction): price increase of 2.2 percent in restaurants and hotels and 1.4 percent in catering services versus 0.5 percent on average.
  - Three months after the VAT reduction: 4.3 percent in restaurants and hotels and 3 percent in catering services vs. 2.6 percent in the headline index.
- These aggregate figures are consistent with estimations showing the VAT reduction had no significant effect on consumer spending, though other factors could contribute to relative price movements across sectors.

### Regression estimates (selected coefficients preserved exactly as reported)
- Baseline columns (Table 1):
  - Treat*VAT: 0.284 (standard error [0.060]) in columns [1]–[4]; 0.290 (standard error [0.060]) in column [5].
  - COVID-19 deaths per population: −0.024** ([0.003]) in column [1]; −0.022** ([0.003]) in columns [2]–[3]; −0.016** ([0.003]) in column [4].
  - Stringency index: −0.010*** ([0.001]) in columns [3]–[5].
  - Economic support index: 0.000*** ([0.001]) in column [4]; 0.001*** ([0.001]) in column [5].
  - COVID-19 vaccinations per population: 0.000*** ([0.000]) in column [5].
  - Number of observations: 3,018 in all specifications reported.
  - Country FE: Yes; Time FE: Yes across Table 1 specifications.

- Robustness checks (Table 2, selected columns):
  - Treat*VAT: 0.289 ([0.060]) in columns [2]–[4]; 0.290 ([0.060]) in column [5]; 0.259 ([0.107]) in column [6].
  - COVID-19 infections per population: −0.002*** ([0.000]) in one specification.
  - COVID-19 infections: −0.001*** ([0.000]) in one specification.
  - COVID-19 deaths: −0.006*** ([0.001]) in one specification.
  - 7-day moving average of COVID-19 deaths per population: −0.018** ([0.004]) in one specification.
  - Stringency index: −0.007*** ([0.001]) to −0.010*** ([0.001]) across specifications.
  - Economic support index: 0.006*** ([0.001]) to 0.00*** ([0.001]) across specifications.
  - COVID-19 vaccinations per population: 0.000*** ([0.000]) across multiple specifications.
  - Number of observations: 3,018 in all Table 2 specifications.
  - Country FE: Yes in all Table 2 specifications.
  - Time FE: Yes in columns [2]–[5]; No in column [6].

- Notes about estimation:
  - Dependent variable is consumer spending as measured by daily debit and credit card transactions.
  - A constant is included in all specifications, and robust standard errors are reported in brackets.
  - *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

### Robustness and identification checks
- Robustness checks include using the number of COVID-19 infections instead of deaths and other alternative measures of the pandemic.
- An alternative specification omits time fixed effects but includes time-varying policy variables for the pandemic response; these estimations yield broadly similar results.
- Placebo regressions and graphical diagnostics for parallel trends (Appendix Figure A1) show no evidence for a violation of the common pre-policy change trend assumption in the baseline setting used.

### Policy implications and interpretation
- The lack of a significant treatment effect is consistent with other quasi-experimental studies of VAT cuts for labor-intensive services.
- During the pandemic, the impact of COVID-19 and mobility restrictions on consumer behavior is far more important than the VAT reduction given significant health and economic uncertainty.
- The results have important policy implications for the expected stimulative effect of sector-specific VAT reductions and the effective design of fiscal policy interventions to counter the impact of crises.

### Key statistics and descriptive sample moments (preserved exactly from Appendix Table A1)
- Observations for debit and credit card transactions by category: 4,113 for each listed category (e.g., Restaurants and catering services, Motor vehicles and services, Travelling and transportation, etc.).
- Selected category means and ranges:
  - Restaurants and catering services: Mean 870,221; Std. dev. 468,112; Minimum 18,743; Maximum 3,588,394.
  - Food and drinks: Mean 3,223,457; Std. dev. 1,653,542; Minimum 66,070; Maximum 8,855,150.
  - Fuel: Mean 1,174,323; Std. dev. 647,626; Minimum 94,725; Maximum 4,232,528.
  - Hotels and accommodation: Mean 246,620; Std. dev. 168,227; Minimum 2,618; Maximum 47,844.
  - Financial services: Mean 640,015; Std. dev. 308,175; Minimum 2,240; Maximum 2,405, (as reported).
  - Miscellaneous: Mean 1,995,054; Std. dev. 1,924,960; Minimum 132,247; Maximum 11,800,000.
- Pandemic and policy indices (observations = 3,018):
  - COVID-19 deaths: Mean 79; Std. dev. 0; Minimum 0; Maximum 79 (as reported).
  - COVID-19 infections: Mean 987; Std. dev. 1,874; Minimum 0; Maximum 15,412.
  - Stringency index: Mean 37; Std. dev. 21; Minimum 0; Maximum 87.
  - Economic support index: Mean 53; Std. dev. 36; Minimum 0; Maximum 100.

### Cost estimate of the VAT reduction
- Reducing the VAT rate from 21 percent to 9 percent on restaurants and catering services is estimated to lower VAT collection by about €133 million (or 0.2 percent of GDP) in 2022.

*Source: wpiea2023188-print-pdf - Section 2; author's estimations; Statistics Lithuania; Swedbank; OxCGRT.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023188-print-pdf.pdf_
