## wp18240

## Source details

**Canonical URL:** [wp18240](https://www.imf.org/-/media/files/publications/wp/2018/wp18240.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp18240.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp18240.pdf.json)

---

### Introduction and key conclusions
- VAT prevalence and Mexican context:
  - About 136 countries have some type of VAT; in Latin America the VAT is close to 30 percent of the tax revenue.
  - Mexico’s VAT law differentiates regions and goods, most importantly separating the general VAT rate and the border-city rate.
- Identification strategy:
  - March of 1995: general rate increased from 10 to 15 percent (border cities excluded).
  - January of 2014: border-city VAT increased from 11 to 16 percent (general rate unchanged).
- Main empirical findings:
  - Full pass-through of VAT changes to consumer prices is rare; pass-through is limited and heterogeneous across goods and regions.
  - For the 1995 reform, estimated monthly inflation effects in April lie between 1.2 and 2.0 (percentage points), implying an elasticity of the VAT between 0.2 and 0.4 percentage points; accumulating April to June yields an elasticity up to 0.6 percentage points.
  - For the 2014 reform, only January is significant, with estimated elasticity between 0.14 and 0.19 percentage points.
  - Including a cross-section inflation control (π_CS) lowers estimated coefficients relative to regressions without it.
  - Net reduction of pass-through from 1995 to 2014 is estimated at about one percentage point (alternative to 1.3 percentage points from the full sample); northern/border-city competition and lower inflation persistence contributed to the reduction.

### Data and measurement
- Price and inflation series:
  - CPI data for 46 cities and 283 categories (subclass-product).
  - Base period: second half of December 2010.
  - Monthly series from January of 1990 to April 2017.
  - Source: INEGI (www.inegi.org.mx).
  - Seasonality adjustment: ARIMAX-12 under the additive option applied to the first difference of the logarithm of the price index; seasonally adjusted log-differences summed to recover levels.
  - Inflation rate used: month-to-month percent change of the seasonally adjusted price index and non-seasonally adjusted series.
- Cross-sectional dependence proxy:
  - π_CS_it = ω′_i π_t, where π_t is a 46×1 vector of city inflation rates and ω_i is a normalized 46×1 weight vector using the inverse Euclidean distance to other cities; variable calculated for each city using the other 45 cities.
- Household expenditure data:
  - 2010 Household Expenditure Survey (INEGI (2011)); household files merged to aggregate expenditure by CPI component.
  - Survey limitations: some expenses understated (e.g., liquor, cigarettes, beer); some goods reported as zero — addressed with a censored regression model.
  - Final household CPI-expenditure database merged with pass-through estimates for 1995 and 2014 to estimate demand system and welfare effects.

### Econometric identification and DID methodology
- Quasi-experimental setup: border vs. non-border comparison enables DID design.
- General regression specification:
  - π_it = μ_0 + μ_C DC_i + μ_T_j DT_{t+j} + φ_j DT_{t+j}·DC_i + θ π_CS_it + ε_it,
    - π_it is monthly inflation, DC_i is city treatment dummy, DT_{t+j} are time dummies with m leads and q lags (j = −q,...,m), π_CS_it is cross-section inflation proxy, and φ captures the VAT net effect.
- Timing and robustness:
  - Leads and lags: six months before and after the VAT change included.
  - Placebo tests on VAT reforms of 1991 and 2010: placebo tests largely show insignificance of φ, with exception for 2010 (significant in January) due to coincident fuel, transportation, and tariff price surges.
- Role of cross-section inflation:
  - Including π_CS controls for geographically correlated inflation and reduces estimated pass-through coefficients.

### Pass-through heterogeneity, dynamics, and interpretation
- Temporal/regional heterogeneity:
  - 1995 effects larger and more persistent than 2014 effects, consistent with higher inflation and greater persistence in 1995.
  - Improved inflation targeting and reduced inflation persistence since 2001 dilute pass-through in later years.
  - Border-region competition (northern border cities) contributes to reduced pass-through.
- Key DID point estimates (selected exact values preserved):
  - Extended DID (mean-group, seasonally adjusted, N=282 subclasses):
    - Overall Inflation: VAT 1995 (Apr) = 1.50*** (Simple) (0.202); VAT 1995 (Apr) = 1.36*** (CS-Inflation) (0.197).
    - Overall Inflation: VAT 2014 (Jan) = 0.73*** (Simple) (0.137); VAT 2014 (Jan) = 0.69*** (CS-Inflation) (0.126).
  - By VAT regime (Table 7):
    - Taxed (Count = 149): VAT 1995 (Apr) = 2.55*** (0.222), 2.47*** (0.229); VAT 2014 (Jan) = 1.26*** (0.156), 1.21*** (0.151).
    - Exempt (Count = 23): VAT 1995 (Apr) = 1.17** (0.486), 0.84* (0.437); VAT 2014 (Jan) = 0.61* (0.340), 0.59* (0.327).
  - By component (Table 6):
    - Core (Count = 229): VAT 1995 (Apr) = 1.85*** (0.182), 1.71*** (0.188); VAT 2014 (Jan) = 0.87*** (0.124), 0.86*** (0.119).
    - Merchandise (Count = 183): VAT 1995 (Apr) = 2.07*** (0.210), 1.93*** (0.217); VAT 2014 (Jan) = 0.68*** (0.131), 0.69*** (0.129).
    - Services (Count = 46): VAT 1995 (Apr) = 0.96*** (0.325), 0.87** (0.338); VAT 2014 (Jan) = 1.65*** (0.302), 1.57*** (0.277).
- Composition and interpretation:
  - Tradable merchandise pass-through decreased over time (greater exposure to competition, more elastic supply).
  - Core-services pass-through increased, consistent with maintained market power and increased demand for services.
  - Taxed goods show the largest pass-through: a five-percentage-point VAT increase in 1995 raised inflation of VAT-taxed goods by more than two and a half percentage points (50 percent pass-through); in 2014 effect closer to one percentage point (20 percent pass-through).
  - Zero-rate goods show no significant effect; exempt goods show significant effects because sellers of exempt goods cannot claim input VAT credits.

### Detailed month-by-month and diagnostic evidence
- Selected Extended DID table entries (exact values preserved):
  - VAT 1995 (Apr): 1.436*** (Inflation (NSA), Simple) (0.199); 1.218*** (CS-Inflation) (0.207); other columns 1.888*** (0.369), 1.987*** (0.413).
  - VAT 1995 (May): 0.637*** (0.236), 0.537** (0.239), 0.636*** (0.240), 0.688*** (0.263).
  - VAT 1995 (Jun): 0.427** (0.172), 0.376** (0.189), 0.682** (0.302), 0.734** (0.354).
  - VAT 2014 (Jan): 0.952*** (0.154), 0.787*** (0.171), 0.722** (0.281), 0.744** (0.318).
- Diagnostics (Table 3 examples):
  - CD Test: 72.27***, 14.81***, 64.15***, 10.32*** across specifications.
  - Avg.Rho CSD: 0.62, 0.13, 0.55, 0.09.
  - BGW Test: 340.59***, 238.75***, 327.67***, 266.03***.
- Simple DID (Table 4 selected entries):
  - VAT 1995 (Apr): 1.534*** and 1.768*** (two columns).
  - VAT 2014 (Jan): 1.005*** and 0.696**.

### Demand estimation and welfare methodology
- Demand system:
  - Censored-augmented QUAIDS estimated using 2010 Household Survey; two-step Heckman correction: first-stage independent probits to estimate inverse Mills ratios; second-stage NLSUR with inverse Mills ratio included.
  - QUAIDS specification (equation IV.1):
    - w_ih = α_i + Σ_j γ_ij ln p_j + β_i ln(y_h / a(p)) + λ_i b(p) [ln(y_h / a(p))]^2 + δ_i R̂_ih,
    - lna(p) = α_0 + Σ_i α_i ln p_i + 1/2 Σ_i Σ_j γ_ij ln p_i ln p_j,
    - b(p) = Π_i p_i^{β_i}.
  - Theoretical restrictions imposed: adding-up, homogeneity, Slutsky symmetry (as listed in source).
- Elasticities (selected exact entries from Table 9):
  - Food and non-alcoholic beverages: Percent of Expenditure 29.30 | Income Elasticity 0.14 | Own-Price Elasticity (Compensated) -0.86 | Own-Price Elasticity (Uncompensated) -0.95.
  - Schooling: Percent of Expenditure 4.57 | Income Elasticity 2.24 | Own-Price Elasticity (Compensated) -1.10 | Own-Price Elasticity (Uncompensated) -1.23.
  - Health care: Percent of Expenditure 1.88 | Income Elasticity 1.96 | Own-Price Elasticity (Compensated) -74.77 | Own-Price Elasticity (Uncompensated) -74.80.
  - Diverse goods and services: Percent of Expenditure 8.04 | Income Elasticity 1.18 | Own-Price Elasticity (Compensated) -159.37 | Own-Price Elasticity (Uncompensated) -159.43.
- Expenditure mapping (12 CPI categories, selected exact series for Food and non-alcoholic beverages):
  - Food and non-alcoholic beverages: 36.61 37.05 35.86 33.41 31.05 29.72 26.78 23.98 20.12 14.43 29.30 (series as reported).

### Welfare loss measurement and results
- Welfare measure: equivalent variation (equivalent expenditure); formulas preserved:
  - ∆Welfare_h = e(p,V(p,y_h)) − e( p̃ ,V(p,y_h)) = y_h − ỹ_h .(IV.3)
  - ỹ_h = lna( p̃ ) + b( p̃ )[lny_h − lna(p)] / b(p) + [lny_h − lna(p)]·[λ(p) − λ( p̃ )].(IV.4)
- Aggregate and normalized losses:
  - Average household lost 1.4 percent of its monthly expenditure from the 1995 reform.
  - Average household lost 0.7 percent of its monthly expenditure from the 2014 reform.
  - Normalized to a one percent increase in the VAT rate:
    - 1995: 0.27 percent loss of monthly expenditure for a one percent increase in the VAT rate.
    - 2014: 0.15 percent loss of monthly expenditure for a one percent increase in the VAT rate.
- Distribution by decile (Table 10 exact entries):
  - Decile 1: Original Expenditure 3,071.1 | Welfare Loss VAT 1995 -39.21 (MX$) | Welfare Loss VAT 2014 -19.49 (MX$) | -1.28 (Percent) | -0.63 (Percent).
  - Decile 5: Original Expenditure 6,365.5 | Welfare Loss VAT 1995 -83.73 | Welfare Loss VAT 2014 -43.66 | -1.32 | -0.69.
  - Decile 10: Original Expenditure 24,323.2 | Welfare Loss VAT 1995 -340.01 | Welfare Loss VAT 2014 -196.04 | -1.40 | -0.81.
  - Overall: Original Expenditure 8,733.1 | Welfare Loss VAT 1995 -117.48 | Welfare Loss VAT 2014 -63.69 | -1.35 | -0.73.
- Normalized-to-1%-VAT findings:
  - 10th decile lose on average 0.28 percent of total expenditure in 1995 and 0.16 percent in 2014 (normalized to a 1 percent VAT increase).
  - Lowest decile lose 0.26 percent and 0.13 percent respectively (normalized to a 1 percent VAT increase).

### Drivers of distributional outcomes and policy-relevant findings
- Reasons for progressive incidence:
  - Low-income households spend a higher share on untaxed goods (food and non-alcoholic beverages).
  - Most affected price categories (communications, healthcare, schooling) are more intensively consumed by high-income households.
  - Diverse goods and services (many VAT-taxed) are an important source of regressivity because low-income households spend a sizable share on them.
- Policy-relevant methodological conclusions:
  - Engel curves are non-linear and price-elasticities are not perfectly inelastic.
  - Implication: assuming uniform VAT rates and perfectly inelastic demand can overestimate revenue gains from removing exemptions.
  - Applying VAT to currently exempt or zero-rated goods and ignoring quantity responses likely overstates revenue projections.
  - Informality matters: literature suggests "approximately 47 percent of the expenditure on food, clothing, and footwear occur in informal establishments in Mexico" (textual note from source).

### Aggregate pass-through and concluding quantitative summaries
- Pass-through and welfare elasticities (conclusions; exact preserved ratios):
  - A one percentage point increase in the VAT rate translates into a 0.4 percentage point growth in prices (40 percent pass-through) and a 0.27 percent loss of monthly expenditure for the average household (1995).
  - In 2014, a one percentage point VAT increase translated into a 0.14 percentage point growth in prices and a 0.15 percent loss of monthly expenditure for the average household.
- Sectoral contrasts:
  - Taxed goods: 1995 forty-to-fifty percent pass-through (example: five-percentage-point VAT → >2.5 percentage-point inflation in taxed goods); 2014 about 20 percent pass-through.
  - Zero-rate goods: no significant pass-through.
  - Exempt goods: significant effects (up to 12 percent in 2014, 17 percent in 1995 in some measures), reflecting inability of sellers to claim input VAT credits.
- Incidence between producers and consumers:
  - Limited price effects in 2014 imply the burden of the VAT tax was shared between producers and consumers.

*Source: wp18240 - References (IMF working paper content provided).*

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

### wp18240 - References

### Introduction and key conclusions
- The value-added tax (VAT) is widely used: about 136 countries have some type of VAT; in Latin America the VAT is close to 30 percent of the tax revenue.
- Mexico’s VAT law differentiates regions and goods, most importantly separating the general VAT rate and the border-city rate.
- Two policy changes form the identification strategy:
  - March of 1995: general rate increased from 10 to 15 percent (border cities excluded).
  - January of 2014: border-city VAT increased from 11 to 16 percent (general rate unchanged).
- Main empirical findings:
  - Full pass-through of VAT changes to consumer prices is rare; pass-through is limited and heterogeneous across goods and regions.
  - For the 1995 reform, estimated monthly inflation effects in April lie between 1.2 and 2.0 (percentage points), implying an elasticity of the VAT between 0.2 and 0.4 percentage points; accumulating April to June yields an elasticity up to 0.6 percentage points.
  - For the 2014 reform, only January is significant, with estimated elasticity between 0.14 and 0.19 percentage points.
  - Including a cross-section inflation control (captures geographically correlated inflation) lowers estimated coefficients relative to regressions without it.
  - The net reduction of the pass-through from 1995 to 2014 is estimated at about one percentage point (rather than 1.3 percentage points from the full sample), with northern/border-city competition and lower inflation persistence contributing to the reduction.
- Welfare implications:
  - Using a system of demand equations estimated on the 2010 Household Survey, demand for most goods and services responds negatively to price increases.
  - Food, beverages, and clothing behave like necessities, with lower demand elasticity among high-income households.
  - Demand elasticity for schooling, furniture, and cultural activities is positively related to household income.
  - The welfare effect of the two reforms (April 1995 and January 2014) was negative and progressive: higher-income households tend to pay more in absolute and relative terms. This outcome reflects uneven pass-through across CPI components, consumption patterns, and a significant share of the lowest deciles’ consumption basket being zero-rated.

### Data and measurement
- Price and inflation series:
  - CPI data for 46 cities and 283 categories (subclass-product).
  - Base period: second half of December 2010.
  - Monthly series from January of 1990 to April 2017.
  - Source: INEGI (www.inegi.org.mx).
  - Seasonality adjustment: ARIMAX-12 under the additive option applied to the first difference of the logarithm of the price index; seasonally adjusted log-differences summed to recover levels.
  - Inflation rate used: month-to-month percent change of the seasonally adjusted price index and non-seasonally adjusted series.
- Cross-sectional dependence proxy:
  - Defined as π_CS_it = ω′_i π_t, where π_t is a 46×1 vector of city inflation rates and ω_i is a normalized 46×1 weight vector using the inverse Euclidean distance to other cities.
  - Variable calculated for each of the 46 cities using the other 45 cities.
- Household expenditure data:
  - 2010 Household Expenditure Survey (INEGI (2011)); household files merged to aggregate expenditure by CPI component.
  - Notes on survey limitations: some expenses are understated (e.g., liquor, cigarettes, beer); some goods reported as zero — addressed with a censored regression model.
  - Expenditure mapping: each household survey expenditure code is mapped to CPI components and aggregated to obtain household total expenditure per CPI component.
  - Final household CPI-expenditure database merged with pass-through estimates for 1995 and 2014 to estimate demand system and welfare effects.

### Econometric identification and DID methodology
- Quasi-experimental setup:
  - Two VAT reforms affected one group of cities but not the other (border vs. non-border), enabling a difference-in-differences (DID) design.
- Regression specification (general form):
  - π_it = μ_0 + μ_C DC_i + μ_T_j DT_{t+j} + φ_j DT_{t+j}·DC_i + θ π_CS_it + ε_it,
    - where π_it is monthly inflation, DC_i is city treatment dummy, DT_{t+j} are time dummies with m leads and q lags (j = −q,...,m), π_CS_it is cross-section inflation proxy, and φ captures the VAT net effect.
- Timing choices and robustness:
  - Leads and lags: six months before and after the VAT change are included to capture anticipatory and second-round effects.
  - Cross-section inflation included to control for correlated shocks across cities derived from trade, proximity, and spillovers.
  - Placebo tests conducted on VAT reforms of 1991 and 2010 (when VAT changed for both groups) to validate the counterfactual assumption; placebo tests largely show insignificance of φ, with some exceptions (2010 has a significant effect in January).
  - Simple DID (without π_CS or pre/post effects) provides similar magnitudes but may overestimate the effect on the switching date and underestimate effects outside that date.

### Pass-through heterogeneity, dynamics, and interpretation
- Temporal and regional heterogeneity:
  - 1995 effects are larger and more persistent than 2014 effects, consistent with higher inflation and greater persistence in 1995.
  - Improved inflation targeting and reduced inflation persistence since 2001 help dilute pass-through in later years.
  - Border-region competition (especially northern border cities) contributes to reduced pass-through.
- Cross-sectional dependence:
  - Non-seasonally adjusted regressions show high cross-sectional dependence (Pesaran CD Test and residual cross-correlation).
  - Including π_CS reduces estimated coefficients, indicating transmission of inflation shocks through geographic proximity.
- Robustness checks:
  - Re-estimations excluding center cities yield lower coefficients than full sample.
  - Re-estimations excluding northern cities (for 2014) yield higher coefficients than full sample, but small number of border cities in this subsample reduces precision.
  - Net estimated reduction in pass-through from 1995 to 2014 is about one percentage point (alternative to 1.3 percentage points from full sample).

### Welfare analysis approach
- Demand estimation:
  - System of demand equations estimated using the 2010 Household Survey.
  - Key patterns: negative price responses across goods; necessities show low elasticity among high-income households; luxury/education/furnishing categories have elasticities positively related to income.
- Welfare measurement:
  - Approach follows King (1983) using a system of demand equations and value function to measure welfare changes.
  - Welfare calculations combine estimated pass-through by CPI component with household expenditure patterns by decile.
- Distributional result:
  - Reforms produced a negative welfare effect overall and a progressive incidence: higher-income deciles bear larger absolute and relative burdens.
  - Drivers: uneven VAT pass-through across CPI components, consumption pattern differences by income decile, and extensive zero-rated items in the consumption baskets of lowest deciles.

*Source: wp18240 - References (IMF working paper content provided).*

### section inflation, the January effect remains significant (Appendix Table 6). In the next section,

### wp18240 - section inflation, the January effect remains significant (Appendix Table 6). In the next section,

### Extended DID and Simple DID Findings on Inflation
- Table 3 (Extended DID): Significant positive responses to VAT 1995 (Apr) and VAT 1995 (May):
  - VAT 1995 (Apr): 1.436*** (Inflation (NSA), Simple), 1.218*** (CS-Inflation), 1.888***, 1.987*** (other columns). Standard errors reported in parentheses (e.g., (0.199), (0.207), (0.369), (0.413)).
  - VAT 1995 (May): 0.637***, 0.537**, 0.636***, 0.688*** (standard errors (0.236), (0.239), (0.240), (0.263)).
  - VAT 1995 (Jun): 0.427**, 0.376**, 0.682**, 0.734** (standard errors (0.172), (0.189), (0.302), (0.354)).
- Diagnostic tests reported (Table 3):
  - CD Test: 72.27***, 14.81***, 64.15***, 10.32*** in different specifications.
  - Avg.Rho CSD: 0.62, 0.13, 0.55, 0.09.
  - BGW Test: 340.59***, 238.75***, 327.67***, 266.03***.
- VAT 2014 results (Table 3):
  - VAT 2014 (Jan): 0.952***, 0.787***, 0.722**, 0.744** (standard errors (0.154), (0.171), (0.281), (0.318)).
  - Other months around 2013–2014 show smaller or insignificant coefficients (e.g., VAT 2013 (Dec): -0.053, 0.022, 0.026, 0.027).
  - CD Test and BGW Test remain significant in many specifications (e.g., CD Test 33.58***, BGW Test 191.48***).
- Table 4 (Simple DID): Key point estimates:
  - VAT 1995 (Mar): 0.519* and 0.982*** (two columns).
  - VAT 1995 (Apr): 1.534*** and 1.768***.
  - VAT 1995 (May): 0.734* and 0.516.
  - VAT 2014 (Jan): 1.005*** and 0.696**.
  - VAT 2013 (Dec): -1.348 and 0.096 (not significant in all specs).
- Figure 4 (DID estimates): Visual evidence of a marked increase in monthly inflation in March–April 1995 for border cities relative to counterfactual non-border cities; and a December–January 2014 spike consistent with the estimated coefficients.

### Bottom-Up (283 CPI Subclasses) Mean-Group Results — Composition of Pass-Through
- Aggregation by expense type (Table 5): mean-group DID estimates (seasonally adjusted)
  - Overall Inflation (N=282 subclasses): VAT 1995 (Apr) = 1.50*** (Simple), 1.36*** (CS-Inflation); VAT 2014 (Jan) = 0.73*** (Simple), 0.69*** (CS-Inflation). Standard errors: (0.202), (0.197), (0.137), (0.126).
  - Significant increases in VAT 1995 (Apr) concentrated in:
    - Alcoholic beverages and tobacco: 5.48*** (Simple), 5.18*** (CS-Inflation) (standard errors (0.630), (0.801)).
    - Clothing and footwear: 2.93***, 3.03*** (std. err. (0.443), (0.469)).
    - Furniture and house supplies: 2.13***, 2.03***.
    - Transportation: 2.87***, 2.20**.
    - Recreation and cultural activities: 2.68***, 2.61***.
    - Diverse goods and services: 3.03***, 2.99***.
  - For VAT 2014 (Jan), notable increases concentrated in:
    - Communications: 2.46** (Simple), 2.29** (CS-Inflation) (std. err. (0.844), (0.778)).
    - Health care: 1.34***, 1.35***.
    - Diverse goods and services: 2.28***, 2.25***.
    - Several categories show small or insignificant estimates (e.g., Food and non-alcoholic beverages: -0.14, -0.18).
  - Specific small-N categories: Schooling (Count = 8) shows VAT 2014 (Jan) = 0.20***, 0.12**.
- Aggregation by inflation component (Table 6):
  - Core (Count = 229): VAT 1995 (Apr) = 1.85*** (Simple), 1.71*** (CS-Inflation); VAT 2014 (Jan) = 0.87***, 0.86***.
  - Merchandise (Count = 183): VAT 1995 (Apr) = 2.07***, 1.93***; VAT 2014 (Jan) = 0.68***, 0.69***.
  - Services (Count = 46): VAT 1995 (Apr) = 0.96***, 0.87**; VAT 2014 (Jan) = 1.65***, 1.57***.
  - Non-core (Count = 53): VAT 1995 (Apr) = 0.00, -0.22; VAT 2014 (Jan) = 0.13, -0.10.
  - Energy and Govt. Tariffs (Count = 13): VAT 1995 (Apr) = 1.43, 0.66; VAT 2014 (Jan) = 1.03, 0.68 (not uniformly significant).
- Aggregation by VAT regime (Table 7):
  - Zero-Rate (Count = 110): VAT 1995 (Apr) = 0.15, -0.03; VAT 2014 (Jan) = 0.05, 0.01 (not significant).
  - Taxed (Count = 149): VAT 1995 (Apr) = 2.55***, 2.47***; VAT 2014 (Jan) = 1.26***, 1.21***.
  - Exempt (Count = 23): VAT 1995 (Apr) = 1.17**, 0.84*; VAT 2014 (Jan) = 0.61*, 0.59*.
- Interpretation highlighted in text:
  - Pass-through from VAT rate changes to inflation decreased between 1995 and 2014 overall, but the composition of the decrease is mixed across subclasses.
  - Tradable merchandise pass-through decreased over time (interpreted as greater exposure to competition and more elastic supply).
  - Core-services pass-through increased, consistent with maintained monopolistic power and increased demand for services.
  - Taxed goods show the largest and most significant pass-through: e.g., a five-percentage-point increase in the VAT rate in 1995 raised inflation of VAT-taxed goods by more than two and a half percentage points (50 percent pass-through); in 2014 the effect is closer to one percentage point (20 percent pass-through).
  - Zero-rate goods show no significant effect; exempt goods show significant effects because businesses selling exempt goods cannot claim input VAT credits, making VAT non-refundable and more likely to be passed to final prices.
- Placebo/robustness notes:
  - Border vs. non-border city counterfactuals passed the placebo test for the 1991 reform but failed for the VAT reform of 2010, particularly in January when fuel, transportation, and government tariff price surges coincided with the VAT change. The disaggregated analysis attributes this failure mainly to non-core components (fuels, transportation, government tariffs) affecting all VAT regimes (appendix references).

### Distributional and Welfare Analysis: Demand System and Pass-Through Imputation
- Approach (Section IV):
  - Use Mexico’s Household Survey of 2010 and estimate a censored-augmented Quadratic Almost Ideal Demand System (QUAIDS) to evaluate welfare effects by imputing pass-through estimates from the 1995 and 2014 VAT reforms into 2010 prices.
  - Estimation accounts for zero expenditures via a two-step Heckman (1979)-style correction: first-stage independent probits to estimate inverse Mills ratios; second-stage Non-linear Seemingly Unrelated Regressions (NLSUR) with the inverse Mills ratio included.
- QUAIDS specification (equation IV.1):
  - w_ih = α_i + Σ_j γ_ij ln p_j + β_i ln(y_h / a(p)) + λ_i b(p) [ln(y_h / a(p))]^2 + δ_i R̂_ih,
    - where w_ih is expenditure share of good i for household h, ln p_j is log price index for good j, y_h is total household expenditure, R̂_ih is inverse Mills ratio.
  - Price index and aggregator:
    - lna(p) = α_0 + Σ_i α_i ln p_i + 1/2 Σ_i Σ_j γ_ij ln p_i ln p_j.
    - b(p) = Π_i p_i^{β_i}.
  - Theoretical restrictions (equation IV.2) imposed: adding-up, homogeneity, and Slutsky symmetry (listed exactly as in the source).
- Elasticities and behavioral implications (summary of Table 9 narrative and text):
  - All aggregates estimated as normal goods (negative own-price elasticities).
  - Food, beverages, tobacco, clothing, and footwear: necessary goods (income elasticity < 1 and λ_i negative → concave Engel curves).
  - Healthcare, transportation, schooling: highest price elasticities (close substitutes or public provision).
  - Schooling, furniture, house supplies, recreation and cultural activities: luxury goods (large income elasticity, convex Engel curves).
  - Diverse goods: large variance across components; price-elasticity result reflects heterogeneity within the category.
- Policy-discussion observations from demand estimates:
  - Mixed VAT rates and exemptions:
    - Zero-rate and exempt regimes matter because of differential ability to claim input VAT credits; exempt goods more likely to pass VAT to final prices.
    - Price-elasticities are non-zero and relatively elastic for many goods; applying a positive VAT rate to currently exempt or zero-rated goods would reduce quantities traded and likely yield lower revenues than static calculations that ignore quantity responses.
    - Literature cited (in text) suggests implicit subsidies from zero-rates could be large, and top deciles may capture a disproportionate share; the authors note the need to consider elasticity heterogeneity and informality (approximate figure from literature: "approximately 47 percent of the expenditure on food, clothing, and footwear occur in informal establishments in Mexico").
- Next step announced in the text:
  - Using the two sets of imputed prices (1995- and 2014-style pass-through applied to 2010 prices) and the estimated QUAIDS, the authors proceed to estimate welfare losses and tax incidence across households (details beyond this section).

### Key Exact Numerical Highlights (preserved)
- Overall Inflation (mean-group, seasonally adjusted, N=282 subclasses):
  - VAT 1995 (Apr): 1.50*** (Simple) (0.202).
  - VAT 1995 (Apr): 1.36*** (CS-Inflation) (0.197).
  - VAT 2014 (Jan): 0.73*** (Simple) (0.137).
  - VAT 2014 (Jan): 0.69*** (CS-Inflation) (0.126).
- By VAT regime (Table 7 exact entries):
  - Taxed (Count = 149): VAT 1995 (Apr) = 2.55*** (0.222), 2.47*** (0.229); VAT 2014 (Jan) = 1.26*** (0.156), 1.21*** (0.151).
  - Exempt (Count = 23): VAT 1995 (Apr) = 1.17** (0.486), 0.84* (0.437); VAT 2014 (Jan) = 0.61* (0.340), 0.59* (0.327).
- Bottom-up by component (Table 6 exact entries):
  - Core (Count = 229): VAT 1995 (Apr) = 1.85*** (0.182), 1.71*** (0.188); VAT 2014 (Jan) = 0.87*** (0.124), 0.86*** (0.119).
  - Merchandise (Count = 183): VAT 1995 (Apr) = 2.07*** (0.210), 1.93*** (0.217); VAT 2014 (Jan) = 0.68*** (0.131), 0.69*** (0.129).
  - Services (Count = 46): VAT 1995 (Apr) = 0.96*** (0.325), 0.87** (0.338); VAT 2014 (Jan) = 1.65*** (0.302), 1.57*** (0.277).

*Sources: INEGI and authors’ calculations.*

### 1. Food and non-alcoholic beverages36.6137.0535.8633.4131.0529.7226.7823.9820.1214.4329.30

### wp18240 - 1. Food and non-alcoholic beverages36.6137.0535.8633.4131.0529.7226.7823.9820.1214.4329.30

### Expenditure mapping (12 CPI categories)
- Expenditure in the mapping bundle of the 283 CPI goods and services for urban and rural households (INEGI, 2010 Household Survey (ENIGH) and authors’ calculations):
  - 1. Food and non-alcoholic beverages: 36.61 37.05 35.86 33.41 31.05 29.72 26.78 23.98 20.12 14.43 29.30
  - 2. Alcoholic beverages and tobacco: 0.41 0.48 0.43 0.44 0.51 0.39 0.47 0.42 0.45 0.41 0.44
  - 3. Clothing and footwear: 3.84 4.35 4.83 5.24 5.37 5.57 5.81 6.00 6.38 6.33 5.32
  - 4. Housing and utilities: 34.18 28.47 26.12 25.90 25.54 25.68 25.81 25.63 26.24 26.29 27.11
  - 5. Furniture and house supplies: 2.23 1.99 2.10 2.12 2.26 2.23 2.37 2.50 2.94 3.44 2.40
  - 6. Health care: 1.92 1.81 1.68 1.65 1.69 1.67 1.74 1.99 2.23 2.44 1.88
  - 7. Transportation: 4.70 5.96 6.81 7.39 7.82 8.01 8.31 8.61 8.52 8.67 7.40
  - 8. Communications: 1.54 2.16 2.75 3.19 3.72 4.14 4.73 5.32 6.03 7.87 4.04
  - 9. Recreation and cultural activities: 1.78 2.40 2.73 3.30 3.56 4.09 4.53 5.36 5.89 7.01 3.97
  - 10. Schooling: 1.99 2.87 3.49 3.83 4.63 4.86 5.32 5.77 6.40 7.55 4.57
  - 11. Restaurants and hotels: 2.50 3.66 4.31 4.87 5.35 5.60 6.23 6.88 7.54 9.45 5.53
  - 12. Diverse goods and services: 8.29 8.80 8.89 8.66 8.51 8.04 7.89 7.56 7.26 6.10 8.04

### Estimated income and own-price elasticities (Table 9)
- Notes: Estimates from a Quadratic Almost Ideal Demand System (QUAIDS) using a Non-linear system of Seemingly Unrelated Regressions (NLSUR).
- Key entries (Percent of Expenditure | Income Elasticities | Own-Price Elasticities (Compensated) | Own-Price Elasticities (Uncompensated)):
  - 1. Food and non-alcoholic beverages: 29.30 | 0.14 | -0.86 | -0.95
  - 2. Alcoholic beverages and tobacco: 0.44 | 0.73 | -0.91 | -0.94
  - 3. Clothing and footwear: 5.32 | 0.94 | -0.93 | -1.00
  - 4. Housing and utilities: 27.11 | 1.17 | -0.84 | -1.05
  - 5. Furniture and house supplies: 2.40 | 2.06 | -1.00 | -1.04
  - 6. Health care: 1.88 | 1.96 | -74.77 | -74.80
  - 7. Transportation: 7.40 | 1.31 | -1.39 | -1.48
  - 8. Communications: 4.04 | 1.31 | -0.97 | -1.03
  - 9. Recreation and cultural activities: 3.97 | 1.66 | -0.43 | -0.51
  - 10. Schooling: 4.57 | 2.24 | -1.10 | -1.23
  - 11. Restaurants and hotels: 5.53 | 1.32 | -0.94 | -1.04
  - 12. Diverse goods and services: 8.04 | 1.18 | -159.37 | -159.43

### Measurement of welfare loss — methodology and formulae
- Welfare measure: equivalent variation (equivalent expenditure) defined as the amount of money that allows a household to get the same utility level as before the VAT reforms.
- Change in household welfare (∆Welfare_h):
  - ∆Welfare_h = e(p,V(p,y_h)) − e( p̃ ,V(p,y_h)) = y_h − ỹ_h .(IV.3)
  - Where e(·) is the expenditure function and V(·) is the value function; p is the vector of actual prices and p̃ is the imputed price vector from the VAT reforms.
- Equivalent expenditure expression:
  - ỹ_h = lna( p̃ ) + b( p̃ )[lny_h − lna(p)] / b(p) + [lny_h − lna(p)]·[λ(p) − λ( p̃ )].(IV.4)
- Implementation note: coefficients φ statistically significant at 10 percent used; zero effect otherwise.

### Welfare loss estimates — aggregate and per-1%-VAT-normalized effects
- Average household lost:
  - 1.4 percent of its monthly expenditure from the 1995 reform.
  - 0.7 percent of its monthly expenditure from the 2014 reform.
- Normalized to a one percent increase in the VAT rate:
  - 1995: 0.27 percent loss of monthly expenditure for a one percent increase in the VAT rate.
  - 2014: 0.15 percent loss of monthly expenditure for a one percent increase in the VAT rate.
- Interpretation: The welfare loss from the 1995 VAT reform is higher than the 2014 reform because inflation was a higher and more persistent process in 1995.

### Distribution of welfare loss by deciles (Table 10)
- Table 10: Monthly average per household (MX$ 2010) and Percent of original expenditure.
- Decile-level results (Original Expenditure | Welfare Loss VAT 1995 (MX$) | Welfare Loss VAT 2014 (MX$) | Welfare Loss VAT 1995 (Percent of original expenditure) | Welfare Loss VAT 2014 (Percent of original expenditure)):
  - Decile 1: 3,071.1 | -39.21 | -19.49 | -1.28 | -0.63
  - Decile 2: 3,966.4 | -51.17 | -25.87 | -1.29 | -0.65
  - Decile 3: 4,587.0 | -59.22 | -29.92 | -1.29 | -0.65
  - Decile 4: 5,454.9 | -71.20 | -36.65 | -1.31 | -0.67
  - Decile 5: 6,365.5 | -83.73 | -43.66 | -1.32 | -0.69
  - Decile 6: 7,246.0 | -95.69 | -50.21 | -1.32 | -0.69
  - Decile 7: 8,531.9 | -113.33 | -60.04 | -1.33 | -0.70
  - Decile 8: 10,284.3 | -137.54 | -73.67 | -1.34 | -0.72
  - Decile 9: 13,522.8 | -184.03 | -101.51 | -1.36 | -0.75
  - Decile 10: 24,323.2 | -340.01 | -196.04 | -1.40 | -0.81
  - Overall: 8,733.1 | -117.48 | -63.69 | -1.35 | -0.73
- Normalized-to-1%-VAT findings highlighted in text:
  - Households in the 10th decile lose on average 0.28 percent of total expenditure in 1995 and 0.16 percent in 2014 (normalized to a 1 percent VAT increase).
  - Households in the lowest decile lose 0.26 percent and 0.13 percent respectively (normalized to a 1 percent VAT increase).

### Drivers of distributional outcomes and incidence
- Two main reasons for progressive effect (absolute and relative):
  - Low-income households spend a higher share of income on untaxed goods, like food and non-alcoholic beverages.
  - Tax incidence on prices is heterogeneous; the most affected prices from the reforms were communications, healthcare, and schooling, which are more intensively consumed by high-income households.
- Regressivity source identified:
  - Diverse goods and services show a source of regressivity because most components are taxed with the VAT and low-income households spend an important share on them (includes housekeeping products, personal care, funerals, and government fees).
- Consistency: expenditure patterns align with the income-elasticities reported in Table 9.

### Pass-through, inflation regime, and composition effects (Conclusions)
- Aggregate pass-through and welfare elasticity:
  - A one percentage point increase in the VAT rate translates into a 0.4 percentage point growth in prices (40 percent pass-through) and a 0.27 percent loss of monthly expenditure for the average household (1995).
  - In 2014, a one percentage point VAT increase translated into a 0.14 percentage point growth in prices and a 0.15 percent loss of monthly expenditure for the average household.
- Taxed vs. zero-rate vs. exempt goods:
  - Taxed goods: highest burden with a fifty percent pass-through from the VAT hike in 1995, and less than half of that pass-through for the VAT reform of 2014.
  - Zero-rate goods: do not show any significance.
  - Exempt goods: show significant effects up to 12 percent in 2014, coming from a 17 percent pass-through in 1995.
- Role of inflation regime:
  - A low inflation regime dampens the overall effect of VAT rate changes on inflation and reduces welfare loss; reduction of VAT effect on inflation from 1995 to 2014 was driven mainly by merchandise (beverages, clothing, furniture), while effects on some services increased (communications, education, healthcare).
- Incidence between producers and consumers:
  - Since the effect on prices was limited in 2014, the burden of the VAT tax was shared between producers and consumers.
- Policy-relevant methodological findings:
  - Engel curves are non-linear, and price-elasticities are not perfectly inelastic.
  - Implication: assuming uniform VAT rates and perfectly inelastic demand can overestimate potential revenue from removing exemptions.

*Sources: INEGI, 2010 Household Survey (ENIGH) and authors’ calculations.*

### Appendix Table 5. DID on Inflation for the VAT Reform of 1995 without Center Cities

### Appendix Table 5. DID on Inflation for the VAT Reform of 1995 without Center Cities

### Main regression results (dependent variable: Monthly Inflation (NSA) / Monthly Inflation)
- Non-Border city coefficients (standard errors in parentheses):
  - −0.853 ∗∗ (0.337)
  - −0.483 (0.372)
  - −0.483 (0.372)
  - −0.483 (0.372)
  - −0.207 (0.185)
  - −0.189 (0.207)
  - −1.207 ∗∗∗ (0.337)
  - −0.424 (0.372)
  - −0.424 (0.372)
  - −0.424 (0.372)
  - −0.322 (0.299)
  - −0.371 (0.292)

- VAT 1995 (Mar) coefficients (standard errors in parentheses):
  - 0.840 ∗∗∗ (0.295)
  - 3.329 ∗∗∗ (0.289)
  - 0.244 (0.432)
  - 0.472 (0.295)
  - 3.131 ∗∗∗ (0.371)
  - 0.191 (0.412)

- Non-Border city * VAT 1995 (Mar) coefficients (standard errors in parentheses):
  - 0.371 (0.345)
  - −0.276 (0.354)
  - −0.322 (0.361)
  - 0.782 ∗∗ (0.345)
  - −0.102 (0.453)
  - −0.142 (0.425)

- VAT 1995 (Apr) coefficients (standard errors in parentheses):
  - 0.867 ∗∗ (0.360)
  - 4.196 ∗∗∗ (0.131)
  - −0.664 (0.643)
  - 0.672 (0.360)
  - 3.803 ∗∗∗ (0.359)
  - −0.840 (0.628)

- Non-Border city * VAT 1995 (Apr) coefficients (standard errors in parentheses):
  - 1.412 ∗∗∗ (0.470)
  - 1.136 ∗∗∗ (0.288)
  - 1.114 ∗∗∗ (0.323)
  - 1.638 ∗∗∗ (0.470)
  - 1.536 ∗∗∗ (0.444)
  - 1.785 ∗∗∗ (0.466)

- VAT 1995 (May) coefficients (standard errors in parentheses):
  - −2.136 ∗∗∗ (0.425)
  - 1.192 ∗∗∗ (0.208)
  - −0.334 (0.281)
  - −1.706 ∗∗∗ (0.425)
  - 1.425 ∗∗∗ (0.210)
  - −0.347 (0.347)

- Non-Border city * VAT 1995 (May) coefficients (standard errors in parentheses):
  - 0.854 ∗ (0.454)
  - 0.578 ∗∗ (0.232)
  - 0.535 ∗∗ (0.244)
  - 0.750 (0.454)
  - 0.648 (0.254)
  - 0.732 ∗∗ (0.286)

- VAT 1995 (Jun) coefficients (standard errors in parentheses):
  - −2.848 ∗∗∗ (0.381)
  - 0.481 ∗∗∗ (0.154)
  - −0.219 (0.172)
  - −2.671 ∗∗∗ (0.381)
  - 0.460 (0.301)
  - −0.453 (0.320)

- Non-Border city * VAT 1995 (Jun) coefficients (standard errors in parentheses):
  - 0.642 (0.427)
  - 0.366 ∗ (0.209)
  - 0.361 ∗ (0.217)
  - 0.711 (0.427)
  - 0.609 ∗ (0.349)
  - 0.734 ∗∗ (0.367)

- Cross-section Inflation coefficients:
  - Cross-section Inflation (NSA): 0.979 ∗∗∗ (0.126)
  - Cross-section Inflation: 0.951 ∗∗∗ (0.114)

- Constant terms (standard errors in parentheses):
  - 4.875 ∗∗∗ (0.299)
  - 5.715 ∗∗∗ (0.330)
  - 5.715 ∗∗∗ (0.330)
  - 5.715 ∗∗∗ (0.330)
  - 2.386 ∗∗∗ (0.149)
  - 0.176 (0.374)
  - 5.185 ∗∗∗ (0.299)
  - 5.657 ∗∗∗ (0.330)
  - 5.657 ∗∗∗ (0.330)
  - 5.657 ∗∗∗ (0.330)
  - 2.526 ∗∗∗ (0.266)
  - 0.359 (0.390)

### Key empirical patterns and statistical significance
- Several VAT month indicators show large positive and statistically significant point estimates (e.g., VAT 1995 (Mar): 0.840 ∗∗∗; VAT 1995 (Apr): 0.867 ∗∗; VAT 1995 (May): −2.136 ∗∗∗; VAT 1995 (Jun): −2.848 ∗∗∗), indicating heterogeneous month-by-month effects around the April 1995 reform.
- Interaction terms Non-Border city * VAT 1995 (Apr) are consistently positive and highly significant across specifications (e.g., 1.412 ∗∗∗; 1.136 ∗∗∗; 1.114 ∗∗∗), suggesting differential post-reform inflation responses for non-border cities in April 1995.
- Non-Border city main effect is negative in multiple specifications and reaches statistical significance in some columns (e.g., −0.853 ∗∗; −1.207 ∗∗∗), indicating lower baseline monthly inflation in non-border cities relative to the omitted category in certain specifications.
- Cross-section inflation (weighted average cross-section inflation rate of near cities) is positive and highly significant (0.979 ∗∗∗ and 0.951 ∗∗∗), implying spillovers or common movements captured by the cross-section measure.

### Sample composition (cities included in the sample)
- Campeche, Camp.; Chetumal, Q.R.; Mérida, Yuc.; Tapachula, Chis.; Tehuantepec, Oax.; Villahermosa, Tab.; Cd. Acuña, Coah.; Cd. Jimenez, Chih.; Cd. Juárez, Chih.; Culiacan, Sin.; Chihuahua, Chih.; Durango, Dgo.; Hermosillo, Son.; Huatabampo, Son.; La Paz, B.C.S.; Matamoros, Tamps.; Mexicali, B.C.; Monclova, Coah.; Monterrey, N.L.; Tijuana, B.C.; Torreón, Coah.

*Sources: INEGI and authors’ calculations.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp18240.pdf_