## 1. Cumulative Effect of Upstream Regulation on Pass-Through

## Source details

**Canonical URL:** [1. Cumulative Effect of Upstream Regulation on Pass-Through](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021061-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021061-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021061-print-pdf.pdf.json)

---

### Main findings
- Value added taxes raise about a fifth of total tax revenues both worldwide and among the members of the OECD (OECD, 2018).
- More than 80 countries undertook VAT reforms in response to the Covid-19 pandemic, with examples including a temporary 3pp cut in the standard rate of VAT in Germany and a tripling of the VAT rate in Saudi Arabia.
- Using a panel of 14 Eurozone countries between 1999 and 2013 and extending Benedek, De Mooij, Keen and Wingender (2015, BDKW), the paper finds systematic heterogeneity in VAT pass-through related to market structure and product characteristics.
- A one standard deviation rise in the competition-friendliness of regulation (roughly equal to the difference between Austria and relatively uncompetitive Italy in 2013) increases pass-through by up to 55%.
- Greater scope for quality differentiation is associated with larger pass-through; results are consistent with theoretical predictions that price-quality complementarity raises pass-through.
- Little overall support for significant announcement (anticipation) effects: reforms announced well in advance do not show robustly different pass-through than surprise reforms.

### Theoretical mechanisms and scenarios examined
- Degree of pass-through defined: 훾_k = 휕ln p_k / 휕ln(1+휏_k).
- Four theoretical scenarios analyzed:
  - A. Imperfect Competition in the Downstream Sector
    - Producers n = 1..N choose price p_n or quantity q_n with cost C_n = a + c_n + b q_n^s / s, with a>0; c_n>0; b>0 corresponds to increasing marginal costs.
    - Competition proxied by number of firms N; effect of competition on pass-through depends on parameter b (increasing vs decreasing marginal costs).
  - B. Imperfect Competition in the Upstream Sector
    - Downstream perfectly competitive; upstream producers sell inputs under Cournot competition; upstream demand derived from downstream cost function C_d = p_x q_x = p_x d (1−ρ) q_i^{ρ/(ρ−1)} with 0<ρ<1 and d>0.
    - Qualitatively similar results to downstream case: when marginal costs are increasing, greater competition raises pass-through; when marginal costs are decreasing, greater competition lowers pass-through.
  - C. Product Differentiation (Quality Ladder)
    - Many varieties with horizontal (ξ_nk, costless) and vertical (λ_n, costly quality) differentiation following Khandelwal (2010).
    - Consumer indirect utility V_nk = δ_n + ξ_nk with δ_n ≡ [θ λ_n]^ψ − p_n]^1/ψ; parameter ψ controls substitutability between price and quality.
    - If ψ>0 (substitution), pass-through decreases with the quality ladder θ; if ψ<0 and |ψ| large enough (complementarity), pass-through increases with θ.
  - D. Early Announcement
    - Anticipation and smoothing effects: menu costs/Calvo pricing can smooth price response; consumer intertemporal demand responses for durables; rational inattention may increase salience and total pass-through.
    - Empirical strategy distinguishes ‘anticipation’ (pre-reform) effects and ‘total’ effects.

### Data, empirical specification, and sample
- Data sources and scope:
  - Monthly VAT rates across European countries and consumption categories constructed by BDKW from European Commission publication VAT Rates Applied in the Member States of the European Union and from the International Bureau for Fiscal Documentation.
  - Monthly prices from Eurostat’s Harmonized Index of Consumer Prices (COICOP); sample excludes categories not sufficiently market-driven (e.g., rental accommodation, electricity, healthcare).
  - All countries are in the Eurozone.
  - Market regulation in upstream non-manufacturing: ‘Regimpact’ indicator from OECD (REGNMI) combined with country-specific input-output weights w_jk.
  - Downstream competitiveness measures from UN Comtrade (BACI): Openness_k,i,t = (Imports_k,i,t + Exports_k,i,t) / Consumption_k,i,t; ImportConcentration_k,i,t = Σ_s s_{i s k t}^2, where s_{i s k t} = M_{i s k t} / Σ_s M_{i s k t}.
  - Quality ladder (scope for quality) from Khandelwal (2010) backed out from price and quantity data; available for subset of goods and does not vary across countries.
- Standardization and sample:
  - Regimpact, trade openness, import concentration and quality ladder are standardized so impacts are comparable.
  - Unbalanced panel of approximately 110k observations spanning January 1998 to December 2013; main reported regressions use N = 100,983 (preferred specification) and N = 107,550 in some primary regressions.
  - Secondary regressions with quality: N = 52,970; quality-inclusive table regressions: N = 49,598.
- Main estimating equation (compact form):
  - Δln(p_ikt) = β_0 + Σ_j β_{1j}·Δln(1+τ_{ikt})_{t+j} + Σ_j β_{2j}·Δln(1+τ_{ikt})_{t+j}·X_{ikt} + β_3·X_{ikt} + φ_{i t} + φ_{k t} + φ_{i k} + ε_{ikt}
  - β_{1j} capture average pass-through at horizons j; β_{2j} measure deviations from mean pass-through across covariates X (Regimpact, quality range, openness, import concentration).
  - Preferred specification: de-seasonalized and de-trended price indices, observations weighted by consumption share, cluster standard errors at country-product level, include country-time, product-time, and country-product interaction fixed effects.

### Main empirical results
- Average baseline pass-through of a VAT rise to prices is 31% in the interaction fixed effects specification (column (3)).
- The contemporaneous month of the reform accounts for most of the pass-through.
- Upstream product market regulation (Regimpact) is a significant driver:
  - A one standard deviation fall in Regimpact (i.e., a one standard deviation rise in competition-friendliness of upstream regulation) raises pass-through by 18 percentage points, corresponding to a 56 percent increase in pass-through.
- Downstream proxies (openness to trade and import concentration) show weaker and generally insignificant effects relative to Regimpact under interaction fixed effects.
- Selected coefficient estimates from the main specification (interaction fixed effects; N = 100,983):
  - Baseline β_1 (summed effects over windows):
    - Pre-Reform: 0.0247 (p-value 0.640)
    - Contemporaneous: 0.257*** (p-value 0.001)
    - Post-Reform: 0.0267 (p-value 0.712)
    - Total (sum over whole window): 0.309*** (p-value 0.001)
  - Openness: Total = 0.00263 (p-value 0.995)
  - Concentration: Total = -0.0351 (p-value 0.754)
  - Regimpact (interaction effects across windows):
    - Pre-Reform = 0.0639 (p-value 0.289)
    - Contemporaneous = -0.228*** (p-value 0.002)
    - Post-Reform = -0.0123 (p-value 0.783)
    - Total = -0.177* (p-value 0.052)
  - Note: Regimpact, openness and market concentration are standardized; coefficients report impact on pass-through of a one-standard-deviation rise in the regressor.
- Dynamics:
  - Most marginal impact of Regimpact on pass-through occurs in the month of the reform, with additional impact in the six months after the reform.
  - Cumulated baseline pass-through shows little pre-change pass-through and most of total effect in the first month of the reform.

### Scope for quality and heterogeneity
- A one standard deviation increase in the length of the ‘quality ladder’ of a product can raise pass-through by more than 40 percentage points (Table 2).
- In complementarity cases (price-quality complementarity), higher taxes can induce consumers to demand higher-quality, higher-priced products, amplifying pass-through.
- Quality effects are significant in the month of reform and continue to grow over the six months following the reform.
- Regulation and quality effects have comparable magnitudes; regulation effects are somewhat more robust across specifications.

### Announcement timing and anticipation effects
- AnnouncedEarly defined as lag between announcement and implementation > median implementation lag of 32 days.
- Table 5: VAT changes with observed announcement dates: N 565 (mean 0.01, s.d. 0.02, min -0.15, max 0.17).
- Main findings:
  - No significant anticipation effect in the six months before reform across specifications.
  - Some weak evidence that reforms announced earlier tend to have slightly larger pass-through, but no strong support for ‘anticipation’ or ‘total effect’ hypotheses.
  - Continuous-lag checks (Table 8): an additional month of implementation lag is weakly associated with up to 6% additional total pass-through (Implementation Lag: Total 0.0571* with p-value (0.053) in column (5)); Implementation Lag is measured in months, so e.g. a coefficient of 0.01 implies that announcing a VAT reform one additional month in advance is associated with a 1% increase in pass-through.
  - Durability split (Table 9): anticipation effects may be stronger for durables in some specifications, but overall no robust evidence of systematic anticipation effects across durables/non-durables.
  - Heterogeneity: substantial variation across individual reforms; examples include package holidays in Luxembourg (announcement 17th October 2011; implementation 1st January 2012; rate from 13% to 23%) showing a possible anticipation effect versus no effect for restaurants & cafés in Portugal (announcement 2nd May 2006; implementation 1st January 2007; rate from 12% to 15%).

### Robustness checks and heterogeneity analyses
- Regimpact discretized (RegimpactHML): Total effect for RegimpactHML: -0.402* (No FEs), -0.330* (Individual FEs), -0.348** (Interaction FEs) with p-values (0.083), (0.091), (0.043) respectively.
- Directional asymmetry (increases vs decreases): little evidence of consistent asymmetric pass-through between VAT increases (N 707) and decreases (N 151); quality-range driven greater pass-through appears essentially driven by reforms with VAT increases (N 373 increases vs N 80 decreases).
- Business cycle: some evidence pass-through effects stronger in expansions (300 VAT changes in expansions, 558 in contractions), but equality of pass-through coefficients across expansionary/contractionary periods cannot be rejected.
- Clustering and alternative specifications: results similar with product-level clustering; country-level clustering renders contemporaneous Regimpact significant while total effect marginally insignificant.
- Additional robustness: alternative upstream PMR measures (narrow/wide) and sectoral decomposition point to professional services (Retail, Accounting, Legal, Engineering, Architecture) as main channel.

### Key data and sample statistics
- Country and reform coverage (Table 1 summary):
  - Total reforms: 33; Products affected: 401; Product-Months affected: 846.
  - Example country entries: Finland (First year in data 1998; Number of reforms 2; Products affected 48; Product-Months affected 59), Portugal (1998; 7; 49; 193).
- VAT and price summary (Table 2):
  - VAT levels:
    - Reduced rate N 31,147 mean 0.075 s.d. 0.033 min 0.021 max 0.17
    - Standard rate N 74,010 mean 0.194 s.d. 0.02 min 0.15 max 0.23
  - VAT changes All N 846 mean 0.01 s.d. 0.02 min -0.15 max 0.17
  - VAT increase N 703 mean 0.02 s.d. 0.01 min 0.01 max 0.17
  - VAT decrease N 143 mean -0.02 s.d. 0.03 min -0.15 max -0.01
  - Price levels N 108,000 mean 102.5 s.d. 19.9 min 18.8 max 527.6
- Competitiveness variable correlations (Table 3 / Table 6 highlights):
  - Pairwise correlations: Openness vs Regimpact: -0.122*; Quality range correlations positive with Openness 0.050* and with Regimpact 0.029*.

### Policy-relevant conclusions and recommendations
- Market characteristics matter for VAT incidence:
  - Pass-through to consumer prices is greater the more competitive the upstream sector (more competition‑friendly regulation) or the wider the quality range of the taxed product.
  - Regulation and quality effects are large in absolute terms and relative to other market characteristics.
- Implications for VAT-based policy objectives:
  - Policy-makers should account for market structure when designing VAT reforms to achieve targeted price or demand effects.
  - For revenue-raising VAT increases (example: Saudi Arabia in May 2020), a greater share of the burden of higher taxes will fall on consumers relative to firms for products with higher upstream competition or wider quality range.
  - For VAT cuts aimed to stimulate consumption or support firm profits (example: Germany in June 2020), firms will retain more of the VAT cut in higher markups, and consumers will experience smaller price reductions the less competitive the upstream sector or the narrower the quality range.
  - Sector-targeted VAT changes should account for upstream regulation and quality scope (e.g., pharmaceuticals draw heavily on highly regulated industries, so pass-through is lower).
  - Where pass-through is small, alternative instruments to VAT changes may be more cost-effective to achieve targeted support.
- Fiscal forward guidance:
  - Early announcement of VAT reforms (median lag 32 days) is unlikely to generate large aggregate anticipation or total effects on prices; announcement effects are heterogeneous and not systematically large in the dataset.

*Source: IMF working paper section "1. Cumulative Effect of Upstream Regulation on Pass-Through" (wpiea2021061-print-pdf).*

### 1. Cumulative Effect of Upstream Regulation on Pass-Through __________________________________ 15

### 1. Cumulative Effect of Upstream Regulation on Pass-Through

### Main findings
- Value added taxes raise about a fifth of total tax revenues both worldwide and among the members of the OECD (OECD, 2018).
- More than 80 countries undertook VAT reforms in response to the Covid-19 pandemic, with examples including a temporary 3pp cut in the standard rate of VAT in Germany and a tripling of the VAT rate in Saudi Arabia.
- Using a panel of 14 Eurozone countries between 1999 and 2013, and extending the empirical methodology of Benedek, De Mooij, Keen and Wingender (2015, BDKW), the paper finds systematic heterogeneity in VAT pass-through related to market structure and product characteristics.
- A one standard deviation rise in the competition-friendliness of regulation (roughly equal to the difference between Austria and relatively uncompetitive Italy in 2013) increases pass-through by up to 55%.
- Greater scope for quality differentiation is associated with larger pass-through; empirical results are consistent with theoretical predictions that price-quality complementarity raises pass-through.
- There is little overall support for significant announcement (anticipation) effects: reforms announced well in advance do not show robustly different pass-through than surprise reforms.

### Theoretical mechanisms identified
- Four partial equilibrium models are developed to identify how supply and demand features influence VAT pass-through:
  - Monopolistic competition with equally productive firms competing in prices.
  - Two models building on Dierickx et al. (1988) with heterogeneous firms and Cournot competition in downstream and upstream sectors.
  - A generalized “quality ladder” model in Khandelwal (2010) allowing substitution or complementarity between consumer valuation of affordability and quality.
- Key theoretical insights:
  - The effect of competition intensity on pass-through depends on whether producers have increasing or decreasing marginal costs. In the intuitive case of increasing marginal costs, pass-through increases with competition because greater competition prevents producers from realizing and passing on savings from scaling down in response to a tax hike.
  - In the quality-ladder model, when there is sufficient price-quality complementarity, higher taxes can induce consumers to demand higher-quality (and higher-priced) products, amplifying pass-through. With less complementarity, consumers shift to lower quality and pass-through is smaller.

### Empirical approach and scope
- Empirical strategy closely follows BDKW to quantify VAT reform effects across product and country levels, enriched by interactions between VAT reforms and measures of competition and scope for quality.
- The study constructs a cross-sector database of VAT reforms including announcement dates by matching VAT change data to the Tax Policy Reform Database (Amaglobeli et al., 2018) to assess announcement effects across many product categories.
- Historical context: liberalizing reforms over the last thirty years have substantially increased competition-friendliness of regulation in European product markets, implying that VAT cuts today will be passed on to consumers more than in the past.

### Policy implications and examples
- Market structure should be a core consideration for VAT reform design:
  - For a government seeking to mobilize revenue through raising VAT (e.g. Saudi Arabia in May 2020), a greater share of the burden of higher taxes will fall on consumers relative to firms for products with higher upstream competition or for products characterized by a wider quality range.
  - For a government using a VAT cut to stimulate consumption (e.g. Germany in June 2020), or to support firm profits, the effects are opposite: firms will retain more of the VAT cut in higher markups, and consumers will experience smaller price reductions, the less competitive the upstream sector or the narrower the range of product quality.
- Sector-targeted VAT changes should account for upstream regulation and quality scope:
  - Pharmaceuticals draw heavily on highly regulated industries, so pass-through is lower; waiving VAT on coronavirus testing kits (as in the EU in April 2020) is likely to provide relatively greater support to the supply side through higher markups than to the demand side through lower prices.
- Advance announcement (fiscal forward guidance) of VAT reforms: despite theoretical importance and prevalence of advanced announcements, the paper finds little overall empirical support for anticipation or total effect hypotheses across many product categories.

*Source: IMF working paper section "1. Cumulative Effect of Upstream Regulation on Pass-Through" (chapter content provided).*

### Section III describes the data and outlines the empirical strategy. Section IV presents the results,

### wpiea2021061-print-pdf - Section III describes the data and outlines the empirical strategy. Section IV presents the results,

### Theoretical framework and scenarios examined
- Degree of pass-through defined: 훾_k = 휕ln p_k / 휕ln(1+휏_k).
- Four theoretical scenarios analyzed:
  - A. Imperfect Competition in the Downstream Sector
    - Producers n=1..N choose price p_n or quantity q_n with cost C_n = a + c_n + b q_n^s / s, with a>0; c_n>0; b>0 corresponds to increasing marginal costs.
    - Competition proxied by number of firms N; effect of competition on pass-through depends on parameter b (increasing vs decreasing marginal costs).
  - B. Imperfect Competition in the Upstream Sector
    - Downstream perfectly competitive; upstream producers sell inputs under Cournot competition; upstream demand derived from downstream cost function C_d = p_x q_x = p_x d (1−ρ) q_i^{ρ/(ρ−1)} with 0<ρ<1 and d>0.
    - Same qualitative results as downstream case: when marginal costs are increasing, greater competition raises pass-through; when marginal costs are decreasing, greater competition lowers pass-through.
  - C. Product Differentiation
    - Many varieties with horizontal (ξ_nk, costless) and vertical (λ_n, costly quality) differentiation following Khandelwal (2010).
    - Consumer indirect utility V_nk = δ_n + ξ_nk with δ_n ≡ [θ λ_n]^ψ − p_n]^1/ψ; parameter ψ controls substitutability between price and quality.
    - If ψ>0 (substitution), pass-through decreases with the quality ladder θ; if ψ<0 and |ψ| large enough (complementarity), pass-through increases with θ.
  - D. Early Announcement
    - Anticipation and smoothing effects: menu costs/Calvo pricing can smooth price response; consumer intertemporal demand responses for durables; rational inattention may increase salience and total pass-through.
    - Empirical strategy distinguishes ‘anticipation’ (pre-reform) effects and ‘total’ effects.

### Data and empirical specification (Section III)
- Data sources and scope:
  - Monthly VAT rates across European countries and consumption categories constructed by BDKW from European Commission publication VAT Rates Applied in the Member States of the European Union and from the International Bureau for Fiscal Documentation.
  - Monthly prices from Eurostat’s Harmonized Index of Consumer Prices (COICOP); sample excludes categories not sufficiently market-driven (e.g., rental accommodation, electricity, healthcare).
  - All countries are in the Eurozone.
  - Market regulation in upstream non-manufacturing: ‘Regimpact’ indicator from OECD (REGNMI) combined with country-specific input-output weights w_jk.
  - Two downstream competitiveness measures from UN Comtrade (BACI): Openness_k,i,t = (Imports_k,i,t + Exports_k,i,t) / Consumption_k,i,t; ImportConcentration_k,i,t = Σ_s s_{i s k t}^2, where s_{i s k t} = M_{i s k t} / Σ_s M_{i s k t}.
  - Quality ladder (scope for quality) from Khandelwal (2010) backed out from price and quantity data; available for subset of goods and does not vary across countries.
- Standardization and sample:
  - Regimpact, trade openness, import concentration and quality ladder are standardized so impacts are comparable.
  - Unbalanced panel of approximately 110k observations spanning January 1998 to December 2013; elsewhere N = 100,983 for the main reported regressions.
- Main estimating equation (compact form):
  - Δln(p_ikt) = β_0 + Σ_j β_{1j}·Δln(1+τ_{ikt})_{t+j} + Σ_j β_{2j}·Δln(1+τ_{ikt})_{t+j}·X_{ikt} + β_3·X_{ikt} + φ_{i t} + φ_{k t} + φ_{i k} + ε_{ikt}
  - β_{1j} capture average pass-through at horizons j; β_{2j} measure deviations from mean pass-through across covariates X (Regimpact, quality range, openness, import concentration).
  - Preferred specification: de-seasonalized and de-trended price indices, observations weighted by consumption share, cluster standard errors at country-product level, include country-time, product-time, and country-product interaction fixed effects.
- Early announcement analysis:
  - Interaction of VAT change with dummy for whether announcement-to-implementation lag is above/below median.
  - Interaction of dummy with sum of pre-reform β_2j over j∈{−6,...,−1} tests anticipation; interaction with cumulation over j∈{−6,...,6} tests total effect.

### Main results (Section IV): Product market regulation, quality scope, and early announcement
- High-level findings:
  - Average baseline pass-through of a VAT rise to prices is 31% in the interaction fixed effects specification (column (3)).
  - The contemporaneous month of the reform accounts for most of the pass-through.
  - Upstream product market regulation (Regimpact) is a significant driver: a one standard deviation fall in Regimpact (i.e., a one standard deviation rise in competition-friendliness of upstream regulation) raises pass-through by 18 percentage points, corresponding to a 56 percent increase in pass-through.
  - Openness to trade and import concentration (downstream competitiveness proxies) show weaker and generally insignificant effects relative to Regimpact, suggesting the upstream mechanism (section II.B) is stronger than the downstream mechanism (section II.A).
- Selected coefficient estimates from the main specification (interaction fixed effects; N = 100,983):
  - Baseline 훽_1 (summed effects over windows):
    - Pre-Reform: 0.0247 (p-value 0.640)
    - Contemporaneous: 0.257*** (p-value 0.001)
    - Post-Reform: 0.0267 (p-value 0.712)
    - Total (sum over whole window): 0.309*** (p-value 0.001)
  - Openness: Total = 0.00263 (p-value 0.995) — not significant under interaction FEs.
  - Concentration: Total = -0.0351 (p-value 0.754) — not significant.
  - Regimpact (interaction effects across windows):
    - Pre-Reform = 0.0639 (p-value 0.289)
    - Contemporaneous = -0.228*** (p-value 0.002)
    - Post-Reform = -0.0123 (p-value 0.783)
    - Total = -0.177* (p-value 0.052)
  - Notes on interpretation: Regimpact, openness and market concentration are standardized; coefficients reported as the impact on pass-through of a one-standard-deviation rise in the regressor.
- Dynamics:
  - Figure (described) plotting cumulated β_{1j} over the 12 months around a VAT change (interaction FEs):
    - Dashed line: pass-through over time for a consumption category with average upstream regulation, openness, and concentration: little pass-through prior to the change; most of total effect in the first month of the reform.
    - Black line: marginal impact on pass-through of having upstream regulation one standard deviation more competition-friendly than average: most marginal impact occurs in the month of the reform, with additional impact in the six months after the reform.
- Additional notes:
  - Pre-Reform, Contemporaneous, Post-Reform and Total effects were also estimated for Openness and Concentration, but under the tighter fixed effects these were not significant and are omitted for concision.
  - Results align with literature on upstream reforms amplifying downstream effects (references cited in source).

### Empirical implications and investigable predictions
- If upstream regulation becomes more competition-friendly (one standard deviation improvement), expect pass-through to increase substantially (example: 18 percentage-point rise observed in interaction-FE specification).
- Most VAT pass-through occurs contemporaneously (within the month of reform), with smaller additional adjustments over the following six months.
- Downstream measures of openness and import concentration are weaker predictors of pass-through heterogeneity relative to upstream regulatory measures in this dataset and specification.
- The sign and magnitude of competition effects on pass-through should be interpreted in light of underlying marginal cost curvature (parameter b): increasing marginal costs imply greater competition → higher pass-through; decreasing marginal costs can reverse this relationship.

*Source: wpiea2021061-print-pdf - Section III describes the data and outlines the empirical strategy. Section IV presents the results.*

### section II adjusting to the change reasonably quickly. The extent to which forewarning of the

### wpiea2021061-print-pdf - section II adjusting to the change reasonably quickly. The extent to which forewarning of the

### Impact of Upstream Regulation on VAT Pass-Through
- Reforms over the last thirty years increased the competition-friendliness of regulation in European product markets (Égert and Wanner, 2016).
- Applying the coefficient on the VAT‑PMR interaction from Table 1 to observed Regimpact changes (1999–2013) yields:
  - Median estimated increase in VAT pass-through: approximately 21 percentage points.
  - Vast majority of the distribution: increase in pass-through of more than 10 percentage points.
- Figure 1 / Figure 3 findings:
  - Cumulative baseline pass-through shifts materially when Regimpact is one standard deviation more (less) competition-friendly.
- Decomposition by upstream sector (narrow vs wide Regimpact):
  - Impact of PMR on pass-through falls substantially when using the ‘narrow’ Regimpact indicator, suggesting the effect is driven by Retail, Accounting, Legal, Engineering, and Architecture (i.e., professional services).

### Scope for Quality and Pass-Through Heterogeneity
- Table 2 main empirical result:
  - A one standard deviation increase in the length of the ‘quality ladder’ of a product can raise pass-through by more than 40 percentage points.
- Interpretation:
  - In the ‘complementarity’ case (demand for quality relatively more important when prices are higher), firms pass on more of a VAT rise rather than reduce quality—larger scope for quality differentiation → higher pass-through.
- Dynamics:
  - Significant effect in the month of reform and continued growth over the six months following the reform (Figure 9).
- Comparability:
  - Regulation and quality effects have comparable magnitudes; regulation effect is somewhat more robust across specifications.

### Announcement Timing (Early Announcement) and Anticipation Effects
- Data and construction:
  - Matched VAT changes to announcement dates in the TPRD; created AnnouncedEarly = 1 if lag between announcement and implementation > median implementation lag of 32 days.
  - Table 5: VAT changes with observed announcement dates: N 565 (mean 0.01, s.d. 0.02, min -0.15, max 0.17).
- Main findings (Table 3 and Figure 4):
  - No significant anticipation effect in the six months before reform across specifications.
  - Some weak evidence that reforms announced earlier tend to have slightly larger pass-through, but no strong support for either the ‘anticipation’ or ‘total effect’ hypotheses.
- Continuous-lag checks (Table 8):
  - An additional month of implementation lag is weakly associated with up to 6% additional total pass-through (specification results: Implementation Lag: Total 0.0571* with p-value (0.053) in column (5) — see Table 8 notes: “Implementation Lag is measured in months, so e.g. a coefficient of 0.01 implies that announcing a VAT reform one additional month in advance is associated with a 1% increase in pass-through”).
- Durability split (Table 9):
  - Anticipation effects may be stronger for durables in some specifications, but overall no robust evidence of systematic anticipation effects across durables/non-durables.
- Heterogeneity:
  - Substantial heterogeneity across individual reforms; examples include a possible anticipation effect for package holidays in Luxembourg (announcement 17th October 2011; implementation 1st January 2012 – rate from 13% to 23%) versus no effect for restaurants & cafés in Portugal (announcement 2nd May 2006; implementation 1st January 2007 – rate from 12% to 15%).

### Robustness Checks and Heterogeneity Analyses
- Regimpact discretized (RegimpactHML):
  - Table 11: RegimpactHML (top/bottom quartile indicator) retains a strong negative relationship with pass-through; Total effect for RegimpactHML: -0.402* (No FEs), -0.330* (Individual FEs), -0.348** (Interaction FEs) with p-values (0.083), (0.091), (0.043) respectively.
- Directional asymmetry (increases vs decreases):
  - Table 12: Little evidence of consistent asymmetric pass-through between VAT increases (N 707) and decreases (N 151); few significant differences.
  - Table 13 (including quality): Quality-range driven greater pass-through appears essentially driven by reforms with VAT increases (N 373 increases vs N 80 decreases).
- Business cycle:
  - Table 14: Some evidence pass-through effects stronger in expansions (300 VAT changes in expansions, 558 in contractions), but cannot reject equality of pass-through coefficients across expansionary/contractionary periods.
- Clustering and alternative specifications:
  - Results similar with product-level clustering; country-level clustering renders contemporaneous Regimpact significant while total effect marginally insignificant.
- Additional robustness:
  - Alternative upstream PMR measures (narrow/wide) and sectoral decomposition point to professional services as main channel.
  - Re-standardization applied in quality subsample so coefficients represent impact of a one-standard-deviation rise in the variable.

### Key Data and Sample Statistics
- Country and reform coverage (Table 1 summary):
  - Total reforms: 33; Products affected: 401; Product-Months affected: 846.
  - Example country entries: Finland (First year in data 1998; Number of reforms 2; Products affected 48; Product-Months affected 59), Portugal (1998; 7; 49; 193).
- VAT and price summary (Table 2):
  - VAT levels: Reduced rate N 31,147 mean 0.075 s.d. 0.033 min 0.021 max 0.17; Standard rate N 74,010 mean 0.194 s.d. 0.02 min 0.15 max 0.23.
  - VAT changes All N 846 mean 0.01 s.d. 0.02 min -0.15 max 0.17; VAT increase N 703 mean 0.02 s.d. 0.01 min 0.01 max 0.17; VAT decrease N 143 mean -0.02 s.d. 0.03 min -0.15 max -0.01.
  - Price levels N 108,000 mean 102.5 s.d. 19.9 min 18.8 max 527.6.
- Main estimation sample sizes:
  - Primary regressions: N 107,550 (price observations) and secondary regressions with quality: N 52,970 (quality range standardized).
  - Table 2 regressions: N 49,598 (quality-inclusive sample).
- Competitiveness variable correlations (Table 3 / Table 6 highlights):
  - Pairwise correlations: Openness vs Regimpact: -0.122*; Quality range correlations positive with Openness 0.050* and with Regimpact 0.029* (significance at 0.05 level noted).

### Theoretical Mechanisms (summary of Theory Appendix)
- Four market settings analyzed to show how competition and product differentiation affect pass-through:
  1. Monopolistic competition with horizontal differentiation: pass-through increases with competition if b>0 (increasing marginal costs); opposite possible when marginal costs decrease (b<0).
  2. Cournot downstream competition: similar results—pass-through increases with competition when b>0 and may decrease when b sufficiently negative.
  3. Cournot upstream / perfect competition downstream: upstream competition increases pass-through when upstream marginal costs are increasing (b>0); reverse when b<0.
  4. Quality ladder (vertical differentiation): effect of scope for quality on pass-through depends on substitution vs complementarity between price and quality:
     - Substitution-dominated case: pass-through decreases with quality ladder.
     - Complementarity-dominated case: pass-through increases with quality ladder.
- Intuition across models:
  - VAT hikes lower demand/scale; the sign and magnitude of resulting cost adjustments (and therefore pass-through) depend on marginal cost curvature (b) and market competitiveness. Greater competition generally dampens firms’ ability to adjust producer prices via scale-related cost changes.

### Policy-Relevant Conclusions and Recommendations
- Market characteristics matter for VAT incidence:
  - Pass-through to consumer prices is greater the more competitive the upstream sector (more competition‑friendly regulation) or the wider the quality range of the taxed product.
  - Regulation and quality effects are large in absolute terms and relative to other market characteristics.
- Implications for VAT-based policy objectives:
  - Policy-makers should account for market structure when designing VAT reforms to achieve targeted price or demand effects.
  - A greater or smaller VAT rate change may be required to achieve a certain consumer price variation depending on market competitiveness and quality scope.
  - Where pass-through is small (producer or consumer prices unresponsive to VAT change), alternative, more cost-effective instruments than VAT changes may be preferable to achieve targeted support.
- Fiscal forward guidance:
  - Early announcement of VAT reforms (median lag 32 days) is unlikely to generate large aggregate anticipation or total effects on prices; announcement effects are heterogeneous and not systematically large in the dataset.

*Source: wpiea2021061-print-pdf - section II adjusting to the change reasonably quickly. The extent to which forewarning of the*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021061-print-pdf.pdf_