## _wp16179

## Source details

**Canonical URL:** [_wp16179](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16179.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16179.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16179.pdf.json)

---

### I. INTRODUCTION

- Background and objective
  - Tax policy and administration affect retail prices and consumer behavior; effective taxation of tobacco raises revenue and discourages cigarette consumption.
  - Empirical literature reviewed indicates "a 10 percent increase in cigarette prices lowers consumption by about 4-6 percent."
  - Pakistan context:
    - Tobacco sector accounts for about 4 percent of sales tax and excises collected at the federal level, and 56 percent of total excise taxes (up from 36 percent in 2010).
    - Smoking prevalence: about 40 percent of the male population and 9 percent of the female population smoke regularly.
    - Annual consumption: over 82 billion cigarettes.
    - Minimum price: PRs 44 per pack of 20 cigarettes.
    - Market structure: dominated by two multinational-owned producers at manufacturer level; retail-level concentration diluted by smuggling and illicit domestic production.

- Research question
  - Estimate and compare the speed and magnitude of tax pass-through to retail cigarette prices across brands at different price points in Pakistan.
  - No prior study on tax pass-through to retail cigarette prices in Pakistan; this paper combines two panel datasets to conduct the empirical analysis.

- Key empirical finding (summary)
  - Pass-through is fast but incomplete:
    - Over 90 percent of pass-through occurs contemporaneously.
    - Adjustment takes place within two months.
    - A one-rupee increase in cigarette taxes leads to an increase of only PRs 0.8 in retail prices on average (undershifting).

### II. DATA

- Datasets used
  - Monthly product- and city-level price data on four major factory-made cigarette brands in 50 cities: July 2004–December 2015.
    - Balanced panel drawn from Pakistan Bureau of Statistics (PBS) CPI price surveys.
    - PBS CPI city coverage changed in July 2008 from 35 cities to 40 cities; effective number of 40 cities in a given month; dataset covers 50 cities with additions/deletions.
  - Recommended retail price figures obtained from manufacturers:
    - Philip Morris International: K-2 and Morven Gold.
    - Pakistan Tobacco Company (subsidiary of British American Tobacco): Capstan and Gold Leaf.
  - Constructed rupee-equivalent cigarette tax series using Federal Board of Revenue (FBR) tax policy changes (ad valorem + specific excises + GST).

- Descriptive price statistics (sample)
  - Budget brand (K-2): sample mean PRs 21 per pack of 20 cigarettes; minimum PRs 6; maximum PRs 60.
  - Premium brand (Gold Leaf): sample mean PRs 66; minimum PRs 37; maximum PRs 135.
  - Mid-range brands (Capstan, Morven Gold): prices ranged from minimum PRs 14 to maximum PRs 70 per pack.
  - GST: standard rate currently set at 17 percent.

- Tax mechanism and administration notes
  - Two types of excises: specific (quantity-based) and ad valorem (value-based).
  - Tobacco manufacturers pay taxes based on the "recommended retail price"; tax liability created at time of sale to distributors.
  - Unit root testing: price variables non-stationary in levels; regression models estimated in first differences (Im-Pesaran-Shin (2003) panel unit root test applied; results available upon request).

### III. ESTIMATION STRATEGY AND RESULTS

- Econometric model
  - Estimated in log differences following Benedek, De Mooij, Keen, and Wingender (2015).
  - Dependent variable: ∆ln(price_ct) where price is retail price of a pack of 20 cigarettes including all taxes.
  - Key regressor: ∆ln(tax_ct) where tax is combined cigarette taxes (ad valorem + specific excises + GST) measured in rupees.
  - Control: ∆ln(cpi_ct) (consumer price index).
  - Model estimated separately for each brand to capture brand-specific pass-through dynamics; lags/leads j ∈ (-12,12) tested; baseline includes contemporaneous and t+1 coefficients.
  - Fixed effects estimation with robust standard errors clustered at the city level; city fixed effects (η_c) and time effects (μ_t) included.

- Interpretation of coefficients
  - β_j measures fraction of tax change passed through to consumer prices at month t+j.
  - Cumulative β = 1 implies full pass-through; >1 overshifting; <1 undershifting.

- Main results (Table 1 summary)
  - Average pass-through: a one-rupee increase in taxes → PRs 0.8 increase in retail prices (undershifting).
  - Speed: over 90 percent of pass-through contemporaneous; whole adjustment within two months.
  - Brand-specific contemporaneous tax pass-through coefficients:
    - Budget (K-2): 0.70*** (standard error 0.034)
    - Mid-range (Capstan): 0.69*** (standard error 0.022)
    - Mid-range (Morven Gold): 0.78*** (standard error 0.023)
    - Premium (Gold Leaf): 0.82*** (standard error 0.060)
  - t+1 coefficients:
    - K-2: 0.03 (0.08)
    - Capstan: 0.00 (0.093)
    - Morven Gold: 0.06 (0.057)
    - Gold Leaf: 0.16* (0.098)
  - CPI coefficients:
    - K-2: 0.01 (0.042)
    - Capstan: 0.11*** (0.033)
    - Morven Gold: 0.00 (0.031)
    - Gold Leaf: 0.06** (0.027)
  - Regression fit and sample details:
    - Number of observations per regression: 4680
    - Number of cities: 40
    - R^2: K-2 0.56; Capstan 0.75; Morven Gold 0.77; Gold Leaf 0.67
    - F-statistics significant with p-values 0.00
  - Alternative robustness checks excluding outliers and cities with incomplete data (balanced panel 2004–2015 with 25 cities) deliver a similar pattern with higher coefficients (Appendix Table 2).

- Interpretation and contributing factors
  - Tax pass-through almost complete for the premium brand (estimated ~0.99 in other summary passages) and significantly higher than for budget and mid-range brands, indicating different demand elasticities across product tiers.
  - Possible drivers of undershifting at the budget end:
    - Competitive market pressures at budget end.
    - Changing consumption patterns and greater awareness of health risks.
    - Smuggling and illicit domestic production increasing competitive pressure.
      - Annual cigarette consumption fell from 84.1 billion sticks in 2009 to 82.4 billion in 2015.
      - Illicit cigarettes reportedly priced at average PRs 27 per pack (below minimum price PRs 44 and below cheapest brand price PRs 60).
      - Nielsen (2016): share of illicit cigarettes doubled from 14 percent in 2010 to 28 percent in 2015.

### IV. CONCLUSION AND POLICY IMPLICATIONS

- Conclusions
  - Tax pass-through to retail cigarette prices in Pakistan:
    - Fast: over 90 percent contemporaneous; adjustment completes within two months.
    - Mostly incomplete in magnitude: average pass-through ~0.8 (PRs 0.8 increase in retail price per PRs 1 tax increase).
    - Variation across price spectrum: near-complete pass-through for premium brands; lower pass-through for budget and mid-range brands.

- Policy recommendations
  - Tax design:
    - Adopt a simpler structure of taxation without tiers to better influence relative prices across the price bands and reduce incentives to switch down to cheaper brands.
  - Administration and enforcement:
    - Strengthen administrative efforts against smuggling and illicit domestic production of cigarettes to reduce downward pressure on legal market prices.
  - Consider market heterogeneity:
    - Tax policy should account for cigarette market conditions across the price spectrum and for consumers’ price sensitivity, brand switching behavior, and manufacturers’ pricing strategies (brand repositioning, differential tax shifting, cross-brand price subsidy), which affect revenue collection and public health objectives.

### Appendix Table 1. List of Cities (selected regression summary rows preserved)

- Table header (preserved exact text)
  - BudgetPremium
  - BrandK-2CapstanMorven GoldGold Lea
  - f

- Tax pass-through (preserve exact layout and numeric values)
  - Contemporaneous
    - 0.77*** (0.037)
    - 0.73*** (0.022)
    - 0.81*** (0.023)
    - 0.89*** (0.027)
  - t+10.
    - 0.03 (0.088)
    - 0.00 (0.096)
    - 0.11* (0.056)
    - 0.18** (0.093)
  - CPI
    - 0.0 (0.054)
    - 0.1 0** (0.040)
    - 0.04 (0.036)
    - 0.11*** (0.031)

- Regression controls and sample
  - Fixed effects: Yes, Yes, Yes, Yes
  - Number of observations: 3400, 3400, 3400, 3400
  - Number of cities: 25, 25, 25, 25

- Goodness-of-fit and significance
  - R2: 0.57, 0.77, 0.81, 0.74
  - F-statistic: 31.2, 78.8, 100.2, 68.5
  - [p-value]: 0.00, 0.00, 0.00, 0.00

- Notes and sources (preserved exact wording)
  - Note: Robust standard errors are reported in parantheses. ***, **, and * denote significance at the 1, 5, and 10 percent levels. All regressions include a constant term, which is not displayed in the table. (Balanced panel excluding cities with incomplete data)
  - Source: Pakistan Bureau of Statistics.
  - Sources: Author's calculations.

*Source: IMF working paper content (I. INTRODUCTION through IV. CONCLUSION, dataset description, estimation results, and policy discussion) as provided in the supplied content unit.*

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

### I.   INTRODUCTION

### Background and objective
- Tax policy and administration affect retail prices and consumer behavior; effective taxation of tobacco raises revenue and discourages cigarette consumption.
- Empirical literature reviewed indicates "a 10 percent increase in cigarette prices lowers consumption by about 4-6 percent."
- Pakistan context:
  - Tobacco sector accounts for about 4 percent of sales tax and excises collected at the federal level, and 56 percent of total excise taxes (up from 36 percent in 2010).
  - Smoking prevalence: about 40 percent of the male population and 9 percent of the female population smoke regularly.
  - Annual consumption: over 82 billion cigarettes.
  - Minimum price: PRs 44 per pack of 20 cigarettes.
  - Market structure: dominated by two multinational-owned producers at manufacturer level; retail-level concentration diluted by smuggling and illicit domestic production.

### Research question
- Estimate and compare the speed and magnitude of tax pass-through to retail cigarette prices across brands at different price points in Pakistan.
- No prior study on tax pass-through to retail cigarette prices in Pakistan; this paper combines two panel datasets to conduct the empirical analysis.

### Key empirical finding (summary)
- Pass-through is fast but incomplete:
  - Over 90 percent of pass-through occurs contemporaneously.
  - Adjustment takes place within two months.
  - A one-rupee increase in cigarette taxes leads to an increase of only PRs 0.8 in retail prices on average (undershifting).

---

### II.   DATA

### Datasets used
- Monthly product- and city-level price data on four major factory-made cigarette brands in 50 cities: July 2004–December 2015.
  - Balanced panel drawn from Pakistan Bureau of Statistics (PBS) CPI price surveys.
  - PBS CPI city coverage changed in July 2008 from 35 cities to 40 cities; effective number of 40 cities in a given month; dataset covers 50 cities with additions/deletions.
- Recommended retail price figures obtained from manufacturers:
  - Philip Morris International: K-2 and Morven Gold.
  - Pakistan Tobacco Company (subsidiary of British American Tobacco): Capstan and Gold Leaf.
- Constructed rupee-equivalent cigarette tax series using Federal Board of Revenue (FBR) tax policy changes (ad valorem + specific excises + GST).

### Descriptive price statistics (sample)
- Budget brand (K-2): sample mean PRs 21 per pack of 20 cigarettes; minimum PRs 6; maximum PRs 60.
- Premium brand (Gold Leaf): sample mean PRs 66; minimum PRs 37; maximum PRs 135.
- Mid-range brands (Capstan, Morven Gold): prices ranged from minimum PRs 14 to maximum PRs 70 per pack.
- GST: standard rate currently set at 17 percent.

### Tax mechanism and administration notes
- Two types of excises: specific (quantity-based) and ad valorem (value-based).
- Tobacco manufacturers pay taxes based on the "recommended retail price"; tax liability created at time of sale to distributors.
- Unit root testing: price variables non-stationary in levels; regression models estimated in first differences (Im-Pesaran-Shin (2003) panel unit root test applied; results available upon request).

---

### III.   ESTIMATION STRATEGY AND RESULTS

### Econometric model
- Estimated in log differences following Benedek, De Mooij, Keen, and Wingender (2015).
- Dependent variable: ∆ln(price_ct) where price is retail price of a pack of 20 cigarettes including all taxes.
- Key regressor: ∆ln(tax_ct) where tax is combined cigarette taxes (ad valorem + specific excises + GST) measured in rupees.
- Control: ∆ln(cpi_ct) (consumer price index).
- Model estimated separately for each brand to capture brand-specific pass-through dynamics; lags/leads j ∈ (-12,12) tested; baseline includes contemporaneous and t+1 coefficients.
- Fixed effects estimation with robust standard errors clustered at the city level; city fixed effects (η_c) and time effects (μ_t) included.

### Interpretation of coefficients
- β_j measures fraction of tax change passed through to consumer prices at month t+j.
- Cumulative β = 1 implies full pass-through; >1 overshifting; <1 undershifting.

### Main results (Table 1 summary)
- Average pass-through: a one-rupee increase in taxes → PRs 0.8 increase in retail prices (undershifting).
- Speed: over 90 percent of pass-through contemporaneous; whole adjustment within two months.
- Brand-specific contemporaneous tax pass-through coefficients:
  - Budget (K-2): 0.70*** (standard error 0.034)
  - Mid-range (Capstan): 0.69*** (standard error 0.022)
  - Mid-range (Morven Gold): 0.78*** (standard error 0.023)
  - Premium (Gold Leaf): 0.82*** (standard error 0.060)
- t+1 coefficients:
  - K-2: 0.03 (0.08)
  - Capstan: 0.00 (0.093)
  - Morven Gold: 0.06 (0.057)
  - Gold Leaf: 0.16* (0.098)
- CPI coefficients:
  - K-2: 0.01 (0.042)
  - Capstan: 0.11*** (0.033)
  - Morven Gold: 0.00 (0.031)
  - Gold Leaf: 0.06** (0.027)
- Regression fit and sample details:
  - Number of observations per regression: 4680
  - Number of cities: 40
  - R^2: K-2 0.56; Capstan 0.75; Morven Gold 0.77; Gold Leaf 0.67
  - F-statistics significant with p-values 0.00
- Alternative robustness checks excluding outliers and cities with incomplete data (balanced panel 2004–2015 with 25 cities) deliver a similar pattern with higher coefficients (Appendix Table 2).

### Interpretation and contributing factors
- Tax pass-through almost complete for the premium brand (estimated ~0.99 in other summary passages) and significantly higher than for budget and mid-range brands, indicating different demand elasticities across product tiers.
- Possible drivers of undershifting at the budget end:
  - Competitive market pressures at budget end.
  - Changing consumption patterns and greater awareness of health risks.
  - Smuggling and illicit domestic production increasing competitive pressure.
    - Annual cigarette consumption fell from 84.1 billion sticks in 2009 to 82.4 billion in 2015.
    - Illicit cigarettes reportedly priced at average PRs 27 per pack (below minimum price PRs 44 and below cheapest brand price PRs 60).
    - Nielsen (2016): share of illicit cigarettes doubled from 14 percent in 2010 to 28 percent in 2015.

---

### IV.   CONCLUSION AND POLICY IMPLICATIONS

### Conclusions
- Tax pass-through to retail cigarette prices in Pakistan:
  - Fast: over 90 percent contemporaneous; adjustment completes within two months.
  - Mostly incomplete in magnitude: average pass-through ~0.8 (PRs 0.8 increase in retail price per PRs 1 tax increase).
  - Variation across price spectrum: near-complete pass-through for premium brands; lower pass-through for budget and mid-range brands.

### Policy recommendations
- Tax design:
  - Adopt a simpler structure of taxation without tiers to better influence relative prices across the price bands and reduce incentives to switch down to cheaper brands.
- Administration and enforcement:
  - Strengthen administrative efforts against smuggling and illicit domestic production of cigarettes to reduce downward pressure on legal market prices.
- Consider market heterogeneity:
  - Tax policy should account for cigarette market conditions across the price spectrum and for consumers’ price sensitivity, brand switching behavior, and manufacturers’ pricing strategies (brand repositioning, differential tax shifting, cross-brand price subsidy), which affect revenue collection and public health objectives.

*Source: IMF working paper content (I. INTRODUCTION through IV. CONCLUSION, dataset description, estimation results, and policy discussion) as provided in the supplied content unit.*

### Appendix Table 1. List of Cities

### Appendix Table 1. List of Cities

### Table structure and brands
- Table header (preserve exact text):
  - BudgetPremium
  - BrandK-2CapstanMorven GoldGold Lea
  - f

### Tax pass-through estimates (preserve exact layout and numeric values)
- Tax pass-through
  - Contemporaneous
    - 0.77*** (0.037)
    - 0.73*** (0.022)
    - 0.81*** (0.023)
    - 0.89*** (0.027)
  - t+10.
    - 0.03 (0.088)
    - 0.00 (0.096)
    - 0.11* (0.056)
    - 0.18** (0.093)
  - CPI
    - 0.0 (0.054)
    - 0.1 0** (0.040)
    - 0.04 (0.036)
    - 0.11*** (0.031)

### Regression controls and sample
- Fixed effects: Yes, Yes, Yes, Yes
- Number of observations: 3400, 3400, 3400, 3400
- Number of cities: 25, 25, 25, 25

### Goodness-of-fit and significance
- R2: 0.57, 0.77, 0.81, 0.74
- F-statistic: 31.2, 78.8, 100.2, 68.5
- [p-value]: 0.00, 0.00, 0.00, 0.00

### Notes and sources (preserve exact wording)
- Note: Robust standard errors are reported in parantheses. ***, **, and * denote significance at the 1, 5, and 10 percent levels. All regressions include a constant term, which is not displayed in the table. (Balanced panel excluding cities with incomplete data)
- Source: Pakistan Bureau of Statistics.
- Sources: Author's calculations.

*Appendix Table 1. List of Cities — Source PDF content*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16179.pdf_