## 1. Distribution of Welfare Impact from Tariff Increase Across Households

## Source details

**Canonical URL:** [1. Distribution of Welfare Impact from Tariff Increase Across Households](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp0809.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp0809.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp0809.pdf.json)

---

### Context and shocks in early 2004
- Madagascar experienced severe cyclone damage to the rice harvest and market infrastructure in February and March 2004.
- Rice harvest typically occurs March through June.
- World price of rice increased by 43 percent in Bangkok dollar terms.
- Malagasy franc (FMG) depreciated by 58 percent relative to the dollar.
- Import parity price of rice increased by 113 percent between January and August 2004.
- Rice is an important staple; such price increases imply substantial household welfare impacts.

### Policy options and trade-offs considered
- Two main policy responses debated:
  - Decrease the import tax on rice.
  - Rely on direct transfers to poor households.
- Rice taxes at start of 2004:
  - Import tariff of 20 percent and VAT of 21 percent (levied on the import tax-inclusive price), combining for a net tax rate of 45 percent.
- Arguments summarized:
  - Reduce tariff: lower domestic rice price quickly benefits rice-consuming poor households, but risks revenue loss, balance-of-payments effects, and political pressure for wider tariff cuts.
  - Direct transfers: can be well-targeted to the poor but require time to design and implement an effective safety net.

### Analytical framework and approach
- Partial equilibrium model with two agents: households and government.
- Household lump-sum income and profit functions specified in equations (1)–(3) (as presented).
- Social welfare: Bergson-Samuelson social welfare over H households (equation (4)).
- Evaluation focuses on "equal-revenue" comparisons: compare welfare effect of reducing tariff by one unit of revenue forgone with allocating that unit to transfers targeted to the poor (revenue-neutral comparison).

### Rice tariff reform — analytical findings
- Social welfare effect of marginal tariff change derived using Roy’s Identity and Shepard’s lemma.
- Revenue effect depends on tariff rate τi and elasticities η for demand, production, and imports.
- Definition: social welfare cost of raising one unit of revenue via tariff denoted as tλ (equation (5)).
- Key insight: if s
i
η = 0 (no quantity responses), revenue elasticity = 1; when s
i
η < 0 revenue elasticity < 1 and welfare cost of raising revenue via tariffs rises due to deadweight loss.

### Targeted transfers — analytical findings
- Social welfare impact of transfers expressed via household marginal utilities and marginal budget shares.
- Revenue cost of transfers includes direct transfers and second-round revenue effects from increased rice consumption.
- Net revenue cost of a unit transfer is less than unity when tax rate positive and rice is normal.
- Administrative/operating costs introduced as proportion ρ of program budget; welfare impact per unit revenue allocated to transfers adjusted by ρ (equation (6)).
- ρ interpreted as share of transfers in total program costs; inverse of ρ is efficiency cost of transfers.

### Welfare weights and inequality aversion
- Atkinson (1970) constant elasticity social welfare function used: βh = (yh/yk)^(-ε) with βk = 1.
- ε captures aversion to inequality; empirical evaluations use ε = 1 to 5.
- Household per capita consumption used as welfare measure.
- Poverty / welfare group definitions:
  - "extreme poor": bottom welfare decile;
  - "moderate poor": deciles 2–3;
  - "poor": bottom three deciles;
  - "middle income": deciles 4–7;
  - "high income": top three deciles.

### Empirical context and rice market facts (EPM2001-based)
- Net tax on rice imports at start of 2004: 45 percent (domestic price 1.45 times world price).
- Household composition and trading patterns:
  - 76 percent of households classified as rural.
  - Around 90 percent of poor households are rural.
  - Urban: 87 percent of households are purchasers of rice.
  - Rural: 66 percent are net purchasers of rice.
  - Urban poor: around 65 percent net purchasers; urban high-income: over 90 percent net purchasers.
  - Rural poor: around 64 percent net purchasers; rural high-income: around 70 percent net purchasers.
- Big rice cultivator prevalence:
  - Urban big cultivators ≈ 3.6 percent of urban households.
  - Rural big cultivators ≈ 35 percent of rural population.
- Price baseline and counterfactual:
  - 2005 price about 5000 FMG per kg.
  - A 10 percent price decrease = 500 FMG per kg.

### Numerical welfare effects of a 10 percent rice price decrease
- Urban net purchasers: welfare increase between 4–6 percent for most net purchasers; falls to between 2–4 percent for high-income households.
- Urban net sellers (big rice producers): welfare decrease greater than 2 percent for moderately poor and middle income households; falls towards zero for extreme poor and high welfare households.
- Rural net purchasers: welfare increases between 2.5–7 percent for poor households; less than 3 percent for high-income households.
- Rural net sellers: decreases progressive — poor households lose between 1–3 percent; high-income households lose between 5–15 percent.
- Household consumption percentiles:
  - Bottom 10 percent: monthly per capita consumption < 25,000 FMG.
  - Bottom 30 percent: consumption < around 40,000 FMG.
  - Top 30 percent: consumption > 135,000 FMG.

### Distributional impact of tariff reductions
- Aggregate targeting performance (Table 1 summary): top five deciles together gain 97.8 FMG out of every 100 FMG lost from tariff revenue — indicating poor targeting of tariff reduction benefits.
- Aggregate effects per 100 FMG revenue forgone by household classification:
  - Landless households (urban + rural) together gain 128 FMG.
  - Big rice cultivators lose 81 FMG.
  - Non-rice farmers and small rice cultivators gain 53 FMG.
- Conclusion: lower rice tariffs redistribute welfare mainly from higher income net producers to higher income net consumers, with little absolute impact on lower income groups because of low absolute rice trading levels among the poor.

### Efficiency impact and Marginal Cost of Public Funds (MCF)
- MCF of raising one unit of revenue via tariffs equals inverse of revenue elasticity.
- Illustrative example:
  - If production is 80 percent of consumption (imports 20 percent), production elasticity 0.2, consumption elasticity -0.3 → import elasticity ≈ -2.3.
  - With initial tax rate 30 percent in that example, MCF = 3.23 (inverse of 1 + 0.3 * -2.3); deadweight loss associated with raising one unit of revenue is 2.23.
- MCF behavior:
  - MCF negative when import share is 10 percent and for higher elasticities of production and consumption, implying tariff on wrong side of Laffer curve (decreasing tariff can increase revenue).
  - Range for MCF considered (excluding Pareto-improving cases): 1.20–6.25.
  - Three MCF values used for combined analysis: 1.20 (LOW), 3.70 (MEDIUM), 6.25 (HIGH).

- Selected cells from Table 2 (Marginal Cost of Funds, τ = 0.3):
  - Demand;Production elasticities = 0.1;0.1:
    - Import Share=10% → 2.33 (1.90 s
η=−)
    - Import Share=20% → 1.37 (0.90 s
η=−)
    - Import Share=30% → 1.20 (0.57 s
η=−)
  - 0.1;0.2:
    - Import Share=10% → 6.25 (2.80 s
η=−)
    - Import Share=20% → 1.64 (1.30 s
η=−)
    - Import Share=30% → 1.32 (0.80 s
η=−)
  - 0.2;0.1:
    - Import Share=10% → -7.14 (3.80 s
η=−)
    - Import Share=20% → 2.17 (1.80 s
η=−)
    - Import Share=30% → 1.51 (1.13 s
η=−)
  - 0.3;0.2:
    - Import Share=10% → -2.27 (4.80 s
η=−)
    - Import Share=20% → 3.23 (2.30 s
η=−)
    - Import Share=30% → 1.78 (1.47 s
η=−)
  - 0.3;0.3:
    - Import Share=10% → -1.41 (5.70 s
η=−)
    - Import Share=20% → 5.26 (2.70 s
η=−)
    - Import Share=30% → 2.04 (1.70 s
η=−)

### Combined distributional and efficiency results (numerical)
- Pure distributional metric Dtλ and adjusted welfare impacts for MCFs tLλ (1.20), tMλ (3.70), tHλ (6.25) across ε values (Table 3):
  - ε = 0:
    - Dtλ = 1.000; tLλ = 1.200; tMλ = 3.700; tHλ = 6.250
  - ε = 1:
    - Dtλ = 0.172; tLλ = 0.207; tMλ = 0.637; tHλ = 1.077
  - ε = 2:
    - Dtλ = 0.050; tLλ = 0.059; tMλ = 0.183; tHλ = 0.310
  - ε = 5:
    - Dtλ = 0.019; tLλ = 0.023; tMλ = 0.071; tHλ = 0.191
- Interpretation:
  - Dtλ declines rapidly with ε, indicating tariff reduction benefits are poorly targeted to low-income groups.
  - Aggregate welfare benefits per revenue unit forgone rise with assumed MCF but fall with greater inequality aversion.

### Targeted transfers — empirical targeting and administrative costs
- Evidence on targeting performance:
  - Median program (Coady, Grosh, and Hoddinott, 2004): “poor” received only 25 percent more than population share → if 30 percent poor, 37.5 percent of transfers go to the poor.
  - Means-tested programs (median): poor received 1.5 times population share → with 30 percent poverty, poor receive 45 percent.
  - Top ten performing means-tested programs (median): poor received ≈ 2 times population share → with 30 percent poverty, poorest 30 percent receive 60 percent.
- Administrative cost evidence:
  - Share of administrative costs in total program budget ranges from 10 percent to 40 percent.
  - Fiscal cost of distributing one unit of welfare to beneficiaries ranges from 1.11–1.67 units.

### Proxy-means targeting simulated for Madagascar (EPM2001)
- Method: regress per capita household consumption on household characteristics to create predicted consumption score; beneficiaries are households below threshold.
- Targeting errors:
  - Undercoverage and leakage highest near poverty line.
  - Undercoverage highest in urban areas; leakage highest in rural areas.
  - Net effect: rural non-poor included at expense of urban poor.
  - Suggested mitigation: separate proxy-means scores for urban and rural areas and allocate program places proportional to urban/rural poverty shares.
- Regression details:
  - Independent variables include geographic location; head’s gender, age, education, sectoral employment; household size and composition; housing types and materials; housing area; water and lighting source; consumer durables.
  - r-squared = 0.69 based on 4,857 household observations.

### Welfare impact of proxy-means targeted transfers (Table 4)
- Share of transfers accruing to poorest households:
  - Nearly 30 percent of transfers accrue to the poorest households: each one unit of revenue transferred through the program results in 0.3 going to the bottom of the welfare distribution.
- Administrative efficiency scenarios:
  - Share of budget allocated to transfers (one minus administrative cost share):
    - Low-efficiency (L): 0.6
    - Medium-efficiency (M): 0.75
    - High-efficiency (H): 0.9
  - Multiplied by 1.057 to capture second-round revenue effects (using 0.18 as marginal budget share for rice).
- Final-row summary: benefit to lowest welfare households per unit revenue allocated to program ranges from 0.19 to 0.28.
- Table 4 entries (Dmλ and mLλ mMλ mHλ):
  - ε = 0: Dmλ = 1.000; mLλ = 0.634; mMλ = 0.793; mHλ = 0.951
  - ε = 1: Dmλ = 0.608; mLλ = 0.386; mMλ = 0.482; mHλ = 0.579
  - ε = 2: Dmλ = 0.442; mLλ = 0.280; mMλ = 0.350; mHλ = 0.420
  - ε = 5: Dmλ = 0.297; mLλ = 0.189; mMλ = 0.236; mHλ = 0.283

### Tariffs versus transfers — comparative findings
- Scenarios:
  - Tariffs: MCF scenarios tHigh = 6.25, tMedium = 3.70, tLow = 1.20.
  - Transfers: administrative cost shares mHigh = 10 percent, mMedium = 25 percent, mLow = 40 percent.
- Key comparative results:
  - If efficiency gains from reducing rice tariffs are low, tariff reduction is dominated by even low-efficiency transfer programs.
  - Tariff reductions superior only if efficiency gains are high (e.g., tHigh) and at lower ε (e.g., ε = 1) where welfare weights place relatively larger weight on gains to higher income groups.
  - At higher ε (greater inequality aversion), transfers rank above tariff reductions.
  - For ε emphasizing the poor (e.g., ε ≥ 2), direct targeted transfers outperform tariff reductions.
  - Efficiency gains from tariff reductions are marginal and expected to decrease non-linearly with the tariff rate.

- Policy sequencing and practical recommendations:
  - Prefer switching from rice tariffs to other taxes and using revenue increases to finance transfer programs when other tax instruments have lower MCF.
  - Initial tariff reductions can yield substantial poverty reduction since marginal efficiency gains are likely larger when reducing tariffs from high levels.
  - Recommended sequence: initially reduce tariffs, gradually replace lost revenue with other tax sources (broader tax reform), and use additional revenues to develop targeted transfer programs.
  - Short-term: ad hoc measures to mitigate adverse impacts on poor net rice sellers, including measures to increase land productivity, yields, and crop diversification; consider conditioning transfers on participation in extension services.
  - Long-term: integrate transfers into a comprehensive targeted safety net with high coverage.

### Partial versus general equilibrium considerations
- Partial equilibrium analysis does not capture indirect revenue effects via other taxed/subsidized goods or welfare effects via factor markets.
- Indirect revenue effects likely limited in Madagascar because main substitutes/complements (maize and cassava) are not traded internationally and are not taxed/subsidized.
- Labor market effects concern:
  - If rice production is unskilled labor-intensive, lower rice prices may reduce unskilled wages.
  - If poorest households depend heavily on unskilled wages and cannot find alternative employment, wage effects could make them net losers — making tariff reductions less attractive distributionally and reinforcing transfers’ dominance for poverty alleviation.
- Overall implication: combine lower tariffs with increases in other taxes and direct safety-net interventions for effective poverty policy.

### Summary conclusions
- Substantial efficiency gains from tariff reductions can exist but tend to accrue mainly to higher income households.
- Poor net rice sellers lose from price decreases.
- Well-designed and implemented targeted direct transfers are likely more cost-effective for poverty alleviation.
- Recommended financing approach: shift revenue raising from rice tariffs to alternative, more efficient tax instruments to fund transfers.
- Short-term policy: targeted transfers to poor net rice sellers, possibly conditioned on extension participation to raise productivity.
- Long-term policy: integrate transfers into a comprehensive targeted safety net with high coverage.

*Authors’ calculations and analysis based on EPM2001 as presented in Section V of the source document.*

### 1. Distribution of Welfare Impact from Tariff Increase Across Households ..........................16

### 1. Distribution of Welfare Impact from Tariff Increase Across Households

### Context and shocks in early 2004
- Madagascar experienced severe cyclone damage to the rice harvest and to crucial market infrastructure in February and March 2004.
- The rice harvest typically occurs during the months March through June.
- The world price of rice increased by 43 percent in Bangkok dollar terms.
- The Malagasy franc (FMG) depreciated by 58 percent relative to the dollar.
- The import parity price of rice increased by 113 percent between January and August 2004.
- Rice is an important staple in Madagascar, so such price increases can be expected to have a substantial welfare impact on households.

### Policy debate and policy options considered
- Two main policy responses debated:
  - Decrease the import tax on rice to lower domestic rice prices.
  - Rely on direct transfers to poor households.
- At the time, rice imports were subject to:
  - an import tax of 20 percent, and
  - a value-added tax of 21 percent (levied on the import tax-inclusive price),
  - combining for a net tax rate of 45 percent.
- Arguments in favor of reducing the rice tariff:
  - Lower domestic price of rice would quickly benefit poor households for whom rice forms a substantial proportion of total consumption.
- Arguments against reducing tariffs:
  - Adverse effects on tax revenues and the balance of payments.
  - Potential political pressure for wider tariff reductions.
- Arguments in favor of direct transfers:
  - Targeted support to poor households; however, developing a well designed and implemented safety net would take some time.

### Objective and analytical framework of the paper
- The paper evaluates the relative merits of the two policy responses (rice tariff reforms versus targeted transfers) from the perspective of:
  - distributional impacts,
  - efficiency implications,
  - revenue implications.
- Structure of the paper:
  - Section II: presents a simple partial equilibrium model to integrate distributional, efficiency, and revenue impacts of both policies into a net welfare evaluation.
  - Section III: uses this framework to evaluate each policy in turn and compare them.
  - Section IV: discusses caveats related to using a partial equilibrium model and potential implications for policy conclusions.

*Source: _wp0809 - 1. Distribution of Welfare Impact from Tariff Increase Across Households ..........................16*

### Section V provides a brief summary of the results and their policy implications.

### _wp0809 - Section V provides a brief summary of the results and their policy implications.

### Partial equilibrium model and framework
- Model structure and assumptions:
  - Two agents: households and the government.
  - Rural agricultural households are included in household sector; other producers implicitly use constant returns to scale technology with fixed producer and factor prices.
  - Household welfare captured by indirect utility function V(p, y).
  - Household lump-sum income: (, )    .ypAmπ=+ (equation (1)).
  - Profit function: (, )    ...pApqcfπ=− (equation (2)).
  - Government revenue from rice import tariffs and lump-sum taxation: equation (3) R = ∑ h ( … ) (as specified).
  - Import tax per unit: *iii ppt−= (domestic price increases one-for-one with tariff under specification).
- Social welfare specification:
  - Bergson-Samuelson social welfare function over H households: 11 [(   ,),.....,(   ,),.....,(   ,)].
hhH H
WWVpyVpyV py= (equation (4)).
- Evaluation approach:
  - Focus on "equal-revenue" expenditure reforms: compare welfare effect of a unit decrease in tariff (reducing revenue by one unit) with allocating an extra unit of revenue to transfers to households identified as “poor” (revenue-neutral comparison).

### Rice tariff reform (analytical results)
- Social welfare and revenue effects:
  - Social welfare effect of marginal tariff change derived using Roy’s Identity and Shepard’s lemma (see derivations in text).
  - Revenue effect involves tariff rate iτ and price elasticities ),,(
s
i
q
i
x
i
ηηη for demand, production, and imports respectively.
  - Interpretation: s
ii
ητ as marginal deadweight loss from unit increase in tariff.
- Social welfare cost of raising one unit of revenue via tariff (definition):
  - Defined as t
λ in equation (5): ().
(1)
hh
hhD E
h
tRtt
s
h
ii
inv
βθ
λβθ   η   λλ
τη
==≡
+
∑
∑
- Key comparative insight:
  - If 0
=
s
i
η (no quantity responses), revenue elasticity is unity; when 0
<
s
i
η revenue elasticity < 1 and welfare cost of raising revenue is greater due to deadweight loss.

### Targeted transfers (analytical results)
- Social welfare impact of transfers:
  - dW/dm derived: h
hhh
hh
hh
WWV
dmdmdm
mVy
β
∂∂∂
==
∂∂∂
∑∑.
- Revenue effect of transfer program:
  - Includes direct transfers and second-round revenue effects from increased rice consumption via marginal budget share (MBS): ()
h
hh
iii
h
hh
i
tpx
R
dmdmdm
mp   m
∂
∂
=−
∂∂
∑∑.
  - If MBS constant across households, bracketed term rewritten and defines marginal tax propensity for rice.
  - Net revenue cost of unit transfer is less than unity when tax rate positive and rice normal.
- Social welfare benefit per unit revenue allocated to transfers:
  - Expressed as ()1
hh
hh
h
m
h
h
ii
i
h
dm
px
dm
m
β
λγβφ
τ
== (see text).
  - h
φ is share of each household in total transfers; γ is inverse of bracketed marginal tax propensity term.
- Administrative/operating costs:
  - Incorporate fixed proportion ρ of total program budget as operating costs; social welfare impact per unit revenue allocated to program becomes: ..
hhh
hhD
hh
mm
h
h
h
dmdm
dmB
β
λγγβφρλγρ==≡ (equation (6)).
  - ρ interpreted as share of transfers in total program costs; inverse of ρ is efficiency cost of transfers.

### Welfare weights
- Specification:
  - Atkinson (1970) constant elasticity social welfare function: (/)
hkh
yy
ε
β≡, with reference household k having 1
k
β=.
  - ε captures "aversion to inequality": 0ε= implies unity weights; 1ε= implies welfare weight = 0.5 for twice-income household; 2ε= implies 0.25 for twice-income household; as ε→∞ Rawlsian maxi-min.
- Empirical choices:
  - Evaluate welfare impact for ε values 1ε= to 5ε=.
  - Household per capita consumption used as welfare measure.
  - Definitions used in paper:
    - "extreme poor": bottom welfare decile;
    - "moderate poor": deciles 2–3;
    - "poor": bottom three welfare deciles;
    - "middle income": deciles 4–7;
    - "high income": top three deciles.
- Visualization note:
  - Non-parametric regressions super-impose vertical lines at the 10th, 30th, 70th, and 90th percentiles.

### Empirical context and rice market facts (EPM2001-based)
- Tariff and tax status at start of 2004:
  - Net tax on rice imports: 45 percent (domestic price 1.45 times world price).
  - Component taxes: import tariff of 20 percent and a value-added tax (VAT) of 21 percent levied on import tax-inclusive price.
- Household composition and trade patterns:
  - Approximately 76 percent of households classified as rural.
  - Around 90 percent of poor households are rural.
  - Urban: 87 percent of households are purchasers of rice.
  - Rural: 66 percent are net purchasers of rice.
  - Urban: around 65 percent of poor households are net purchasers; over 90 percent of high-income households are net purchasers.
  - Rural: around 64 percent of poor households are net purchasers; around 70 percent of high-income households are net purchasers.
- Rice cultivator classifications and prevalence:
  - Big rice cultivators constitute around 3.6 percent of all urban households; in rural areas big rice cultivators constitute around 35 percent of total population.
- Price baseline and counterfactual used:
  - 2005 price about 5000 FMG per kg.
  - A 10 percent price decrease = 500 FMG per kg.

### Welfare effects of a 10 percent rice price decrease (numerical impacts)
- Urban net purchasers:
  - Welfare increase between 4–6 percent for most net rice purchasers; falls to between 2–4 percent for high-income households.
- Urban net sellers (big rice producers):
  - Decrease in welfare greater than 2 percent for moderately poor and middle income households; falls towards zero for extreme poor and high welfare households.
- Rural net purchasers:
  - Welfare increases between 2.5–7 percent for poor households; less than 3 percent for high-income households.
- Rural net sellers:
  - Decreases progressive: poor households experience decreases between 1–3 percent; high-income households experience decreases between 5–15 percent.
- Note on household consumption percentiles:
  - Bottom 10 percent have monthly per capita consumption less than 25,000 FMG;
  - Bottom 30 percent consumption less than around 40,000 FMG;
  - Top 30 percent consumption greater than 135,000 FMG.

### Distributional impact of tariff reductions
- Aggregate targeting performance:
  - Table 1 analysis: top five deciles together gain 97.8 FMG out of every 100 FMG lost from tariff revenue—indicating poor targeting of tariff reduction benefits.
- By household classification (aggregate effects per 100 FMG revenue forgone):
  - Landless households (urban + rural) together gain 128 FMG.
  - Big rice cultivators lose 81 FMG.
  - Non-rice farmers and small rice cultivators gain 53 FMG.
- Conclusion:
  - Lower rice tariffs redistribute welfare from higher income net producers to higher income net consumers with little absolute impact on lower income groups due to low absolute rice trading levels among the latter.

### Efficiency impact and marginal cost of public funds (MCF)
- Concept:
  - Marginal social cost of raising one unit of revenue via tariffs equals inverse of revenue elasticity (equation (5) context).
- Illustrative numeric example (from text):
  - If production is 80 percent of total consumption (imports 20 percent), production elasticity 0.2, consumption elasticity -0.3, then import elasticity approximately -2.3.
  - With initial tax rate 30 percent (as defined in that example), MCF = 3.23 (inverse of 1 + 0.3 * -2.3). Deadweight loss associated with raising one unit of revenue is 2.23.
- Behavior of MCF:
  - MCF negative when import share is 10 percent and for higher elasticities of production and consumption — indicates tariff on wrong side of Laffer curve (decreasing tariff can increase revenue).
  - Range for MCF considered (excluding Pareto-improving cases): 1.20–6.25.
  - For combining distributional and efficiency impacts, the paper considers MCF values: 1.20 (LOW), 3.70 (MEDIUM), and 6.25 (HIGH).

- Table 2 summary (selected MCF cells preserved as presented; note authors’ calculations based on EPM2001):
  - Marginal Cost of Funds (0.3τ=) for demand and production elasticities:
    - 0.1;0.1: Import Share=10% → 2.33 (1.90 s
η=−); Import Share=20% → 1.37 (0.90 s
η=−); Import Share=30% → 1.20 (0.57 s
η=−).
    - 0.1;0.2: Import Share=10% → 6.25 (2.80 s
η=−); Import Share=20% → 1.64 (1.30 s
η=−); Import Share=30% → 1.32 (0.80 s
η=−).
    - 0.2;0.1: Import Share=10% → -7.14 (3.80 s
η=−); Import Share=20% → 2.17 (1.80 s
η=−); Import Share=30% → 1.51 (1.13 s
η=−).
    - 0.3;0.2: Import Share=10% → -2.27 (4.80 s
η=−); Import Share=20% → 3.23 (2.30 s
η=−); Import Share=30% → 1.78 (1.47 s
η=−).
    - 0.3;0.3: Import Share=10% → -1.41 (5.70 s
η=−); Import Share=20% → 5.26 (2.70 s
η=−); Import Share=30% → 2.04 (1.70 s
η=−).

### Combined distributional and efficiency results (numerical)
- Pure distributional impact D
t
λ and adjusted welfare impacts for alternative MCFs and inequality aversion levels (Table 3; authors’ calculations based on EPM2001):
  - Columns: D
t
λ | tL
λ | tM
λ | tH
λ corresponding to MCFs 1.20, 3.70, 6.25 respectively.
  - Inequality Aversion 0ε=:
    - D
t
λ = 1.000; tL
λ = 1.200; tM
λ = 3.700; tH
λ = 6.250.
  - Inequality Aversion 1ε=:
    - D
t
λ = 0.172; tL
λ = 0.207; tM
λ = 0.637; tH
λ = 1.077.
  - Inequality Aversion 2ε=:
    - D
t
λ = 0.050; tL
λ = 0.059; tM
λ = 0.183; tH
λ = 0.310.
  - Inequality Aversion 5ε=:
    - D
t
λ = 0.019; tL
λ = 0.023; tM
λ = 0.071; tH
λ = 0.191.
- Interpretation:
  - Pure distributional metric D
t
λ declines rapidly with ε indicating benefits of tariff reduction are poorly targeted to low-income groups.
  - Final three columns show aggregate welfare benefits per unit revenue forgone increase with assumed MCF (efficiency gains from tariff reductions); magnitude also declines with greater inequality aversion.

### Main empirical conclusions and policy implications (as presented)
- Tariff reductions:
  - Tend to generate sizable efficiency gains (large reductions in deadweight loss) when import base is elastic and imports small share of supply; these efficiency gains increase the welfare return to tariff reduction when measured in aggregate.
  - Distributional gains from tariff reductions are poorly targeted to the poor: top deciles capture the vast majority of gains; D
t
λ for 5ε= is 0.019 (only 0.019 units of each revenue unit forgone accrues to the lowest parts of the welfare distribution).
- Targeted transfers:
  - Welfare impact of transfers depends critically on targeting (h
φ) and program efficiency (ρ).
  - Transfers that are well-targeted to households with high welfare weights produce greater social welfare benefits per unit revenue than poorly targeted transfers.
- Trade-off:
  - Comparison between tariff reductions and targeted transfers must account for both distributional targeting (D
t
λ, and analogous transfer metric) and efficiency (MCF for tariff changes; 1/ρ and marginal tax propensity for transfers).
- Practical considerations:
  - Operating costs and leakages (parameter ρ) materially reduce the welfare benefit of transfers; accounting for MBS (marginal budget share) and second-round revenue effects is essential.
  - In some parameterizations (import share low and certain elasticities), tariff decreases can be Pareto-improving (increase revenue and household welfare simultaneously).

*Italic: Authors’ calculations and analysis based on EPM2001 as presented in Section V of the source document.*

### 1.20 to 6.25 for low to high efficiency gains. For each value of MCF, the column gives the

### _wp0809 - 1.20 to 6.25 for low to high efficiency gains. For each value of MCF, the column gives the

### Targeted Transfers: effectiveness, targeting performance, and administrative costs
- Targeting effectiveness depends on: identification of poor households and low administrative cost delivery.
- Even best programs are imperfect: leakage to non-poor and incomplete coverage of the poor.
- Median program targeting performance (Coady, Grosh, and Hoddinott, 2004): the “poor” received only 25 percent more than their population share.
  - If 30 percent classified as poor, this implies 37.5 percent of transfers go to the poor.
- Means-tested programs (median): poor received 1.5 times their population share.
  - With 30 percent poverty rate, poor receive 45 percent of total transfers.
- Top ten performing means-tested programs (median): poor received approximately twice their population share.
  - With 30 percent poverty rate, poorest 30 percent would receive 60 percent of transfers.
- Administrative cost evidence (limited) suggests wide variability:
  - Share of administrative costs in total program budget ranges from 10 percent to 40 percent.
  - Fiscal cost of distributing one unit of welfare to all beneficiary households ranges from 1.11–1.67 units.

### Proxy-means targeting simulated for Madagascar (EPM2001)
- Simulation method:
  - Regress per capita household consumption on household characteristics to create a predicted consumption per capita “score.”
  - Households below a threshold of the score are program beneficiaries.
  - Targeting errors: errors of omission (excluded poor) and errors of inclusion (included non-poor).
- Distributional pattern of targeting errors (Figure 5 summary):
  - Undercoverage and leakage highest close to the poverty line.
  - Undercoverage highest in urban areas.
  - Leakage highest in rural areas.
  - Net effect: rural non-poor included at the expense of urban poor.
  - Suggested mitigation: separate proxy-means scoring for urban and rural areas and allocate program places proportional to urban/rural poverty shares.
- Underlying regression details:
  - Dependent variable: per capita household consumption.
  - Independent variables: geographic location; gender, age, education and sectoral employment of head; household size and composition; housing types and materials; housing area; water and lighting source; possession of consumer durables.
  - r-squared for regression: 0.69 based on 4,857 household observations.

### Welfare impact of proxy-means targeted transfers (Table 4)
- First-column (share of transfers accruing to various target populations):
  - Nearly 30 percent of transfers accrue to the poorest households: every one unit of revenue transferred through the program results in 0.3 going to the bottom of the welfare distribution.
- Administrative efficiency scenarios and second-round revenue adjustment:
  - Share of budget allocated to transfers (one minus administrative cost share) taken as:
    - Low-efficiency program (L): 0.6
    - Medium-efficiency programs (M): 0.75
    - High-efficiency program (H): 0.9
  - Multiplied by 1.057 to capture second-round revenue effects, derived using 0.18 as the marginal budget share for rice.
- Final-row summary (benefit to lowest welfare households per one unit of revenue allocated to the program):
  - Ranges from 0.19 to 0.28.

- Table 4 entries (Welfare Impact of Proxy-Means Targeted Transfers):
  - Row labels epsilon: 0ε=, 1ε=, 2ε=, 5ε=
  - Columns Dmλ mLλ mMλ mHλ
  - Values:
    - 0ε=: 1.000, 0.634, 0.793, 0.951
    - 1ε=: 0.608, 0.386, 0.482, 0.579
    - 2ε=: 0.442, 0.280, 0.350, 0.420
    - 5ε=: 0.297, 0.189, 0.236, 0.283

### Tariffs, Transfers, or Both? — comparison of welfare impact per unit revenue
- Scenarios considered (three efficiency levels for each instrument):
  - Tariffs: revenue elasticities considered are 6.25 (tHigh), 3.70 (tMedium), and 1.20 (tLow).
  - Direct transfers: administrative cost shares considered are 10 percent (mHigh), 25 percent (mMedium), and 40 percent (mLow).
- Key comparative findings (Figure 6 summary):
  - If efficiency gains from reducing rice tariffs are low, tariff reduction is dominated by even low-efficiency direct transfer programs.
  - Tariff reductions are superior only if:
    - Efficiency gains are high (e.g., revenue elasticity near tHigh), and
    - At lower values of ε where welfare weights place relatively larger weight on gains to higher income groups (e.g., 1ε=).
  - At higher values of ε (greater inequality aversion), tariffs become less attractive because transfers focus gains on the bottom of the welfare distribution.
  - For ε values emphasizing the poor (e.g., 2ε>), direct targeted transfers rank above tariff reductions.
  - Efficiency gains from tariff reductions are upper-bounds (marginal changes); they are expected to decrease non-linearly with the tariff rate.
- Policy sequencing and caveats:
  - Preferable to switch from rice tariffs to other taxes and use revenue increases to finance transfer programs when other tax instruments have lower MCF.
  - Initial tariff reductions can yield substantial poverty reduction since efficiency gains are likely higher when reducing tariffs from high levels.
  - Recommended sequencing: initially reduce tariffs, gradually replace with other tax revenues (broader tax reform), and use extra revenues to develop targeted transfer programs.
  - Short-term: ad hoc measures to mitigate adverse impacts on poor net sellers of rice (small-holder rice producers), including measures to increase land productivity, yields, and crop diversification; condition transfers on participation in extension services.
  - Long-term: integrate transfers into comprehensive targeted safety net with high coverage.

### Partial versus general equilibrium considerations
- Partial equilibrium limitation: earlier analysis does not capture indirect revenue effects via changing demand and supply for other taxed/subsidized goods and welfare effects via factor markets.
- Indirect revenue effects likely limited in Madagascar because main substitutes/complements (maize and cassava) are not traded internationally and not taxed/subsidized.
- Labor market effects:
  - If rice production is relatively unskilled labor-intensive, lower rice prices may reduce unskilled wages.
  - If poorest households rely heavily on unskilled wages and cannot find alternative employment, wage effects could make them net losers.
  - Such wage effects would make tariff reductions even less attractive distributionally and reinforce dominance of transfers for poverty alleviation.
- Overall implication: lower tariffs combined with increases in other taxes and direct safety-net interventions remains the most effective response.

### Summary and conclusions
- Main concerns: relative efficiency, distributional, and revenue implications of tariffs and transfers for poverty alleviation.
- Core conclusions:
  - Substantial efficiency gains from tariff reductions exist, but accrue mainly to higher income households.
  - Poor net rice sellers lose from price decreases.
  - Well designed and implemented targeted direct transfers are likely to be substantially more cost-effective for poverty alleviation.
  - Financing approach: switch revenue raising from rice tariffs to alternative, more efficient tax instruments to fund transfers.
  - Short-term focus: transfers to poor net rice sellers, potentially conditioned on extension participation to raise agricultural productivity.
  - Long-term: integrate into a comprehensive targeted safety net with high coverage.

### Appendix I: The Madagascar EPM Household Survey (EPM2001) — data and poverty definition
- Survey details:
  - Organized by INSTAT at the end of 2001.
  - Nationwide and comprehensive: socioeconomic characteristics, consumption, health, education, income sources, time allocation, occupation, and agricultural production.
  - Stratified sample representative at national and provincial level and urban-rural divide.
  - Total sample: 5,080 households.
  - About 2,500 households had land in cultivation.
  - Households in rural areas accounted for 2,040 households in the sample.
- Consumption aggregate:
  - Calculated for the year prior to survey and incorporates: food autoconsumption, purchased food, food gifts and payments in-kind, education and health care expenditures, imputed and actual housing costs, expenditures on consumer durables, other non-food expenses.
  - Deflated for regional price differences using survey price information and the Paasche index method.
- Poverty line and rates:
  - Poverty line evaluated in 2001 at approximately 988.6 FMG per person per year (corresponding to US$0.42 per day).
  - Estimated that almost 70 percent of the Malagasy population was poor.
  - Poverty rate: rural areas 77 percent; urban areas 44 percent.

### Table 5 (Mean Per Capita Consumption and Welfare Weights) — selected entries
- Based on per capita monthly household consumption (FMG ‘000), adjusted for regional price differences, sample after dropping bottom/top 1 percent outliers: sample decreases from 5,080 to 4,955 households.
- Selected rows (Mean Per Capita Consumption; Ratio of Mean to Bottom; welfare weights for ε values 0.5, 1.0, 2.0, 3.0, 4.0, 5.0):
  - Bottom: 19.86; Ratio 1.00; weights: 1.00, 1.00, 1.00, 1.00, 1.00, 1.00
  - 2nd: 29.60; Ratio 1.49; weights: 0.82, 0.67, 0.45, 0.30, 0.20, 0.14
  - 3rd: 37.61; Ratio 1.88; weights: 0.73, 0.53, 0.28, 0.15, 0.08, 0.04
  - 4th: 46.47; Ratio 2.33; weights: 0.66, 0.43, 0.18, 0.08, 0.03, 0.01
  - 5th: 57.12; Ratio 2.86; weights: 0.59, 0.35, 0.12, 0.04, 0.01, 0.01
  - 6th: 70.80; Ratio 3.55; weights: 0.53, 0.28, 0.08, 0.02, 0.01, 0.00
  - 7th: 89.16; Ratio 4.47; weights: 0.47, 0.22, 0.05, 0.01, 0.00, 0.00
  - 8th: 117.99; Ratio 5.90; weights: 0.41, 0.17, 0.03, 0.00, 0.00, 0.00
  - 9th: 165.44; Ratio 8.34; weights: 0.35, 0.12, 0.01, 0.00, 0.00, 0.00
  - Top: 299.43; Ratio 15.41; weights: 0.25, 0.06, 0.00, 0.00, 0.00, 0.00

*Source: Authors’ calculations based on EPM2001 (excerpted content provided).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp0809.pdf_