## World — Online Annex (ch2onlineannex)

## Source details

**Canonical URL:** [World — Online Annex (ch2onlineannex)](https://www.imf.org/-/media/files/publications/weo/2026/october/english/ch2onlineannex.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/weo/2026/october/english/ch2onlineannex.pdf.md)
- [Structured JSON version](/-/media/files/publications/weo/2026/october/english/ch2onlineannex.pdf.json)

---

### Relative prices, real wages, and inflation inequality under alternative treatments
- Alternative world-commodity-price classification:
  - The relative prices of essentials rises steadily, reaching 10 percent by the tenth quarter.
  - Real wages hold up during the first year before falling to 4 percent below their pre-onset level.
  - The inflation gap between quintiles 1 and 5 turns positive after the fourth quarter and reaches 2½ log points by the tenth quarter.
- Oil supply news shock (Känzig 2021), normalized to a 150 percent increase in the world oil price on impact:
  - Raises the relative prices of essentials by 9 percent within two quarters and by 13 percent by the eighth quarter.
  - Real wages fall to a trough 9½ percent below their pre-shock level in the seventh and eighth quarters and recover to only 4½ percent below that level by the twelfth quarter.
  - The inflation gap between quintile-specific price indices widens to 3 log points by the seventh quarter.
- Euro area pure monetary policy surprise (Jarociński and Karadi 2020), normalized to a 25-basis-point easing surprise:
  - A 25-basis-point easing surprise leads neither to a significant drop in real wages nor to a significant increase in the gap in the inflation experienced by low-income households relative to that experienced by high-income households.
  - This suggests that the baseline results largely capture global supply shocks rather than demand-driven episodes.
- Estimation and presentation notes:
  - Results are from local-projection estimates on a quarterly panel of countries over 2005–25 with country fixed effects.
  - Shaded areas indicate 90 percent confidence intervals based on country-clustered standard errors.
  - The relative price of basic necessities is the CPI for food, transport, and housing-related expenses, including utilities, relative to the CPI for all other items.
  - Real wages are nominal wages deflated by headline CPI.

### Cross-country distribution of CPI weights and income-specific CPI construction
- Cross-country distribution of CPI weights (basic necessities):
  - Basic necessities composition includes "Food and non-alcoholic beverages", "Housing, water, and home energy", and "Transport".
  - Boxplot edges represent interquartile range, the line inside each box indicates the median country, and whiskers denote the 90th and 10th percentiles.
  - Lower-income households allocate a larger share of their consumption basket to basic necessities, making them disproportionately affected by rises in the price of basic necessities.
  - Country groups shown: AEs (advanced economies), EMs (emerging markets), LIDCs (low-income developing countries).
- Distributional Consumer Price Index (income-specific CPI) construction and budget shares:
  - Definition (per Jaravel 2026):
    - For household quintile q in economy i in quarter t, the income-specific price index P_i,t^q is constructed using chained fixed-basket indices that weight COICOP division price indices by that quintile’s budget shares, with t_b denoting the quarter of the most recent measurement of expenditure weights at or before t.
    - P_i,t^q = P_i,t^b_q * (sum_c w_i,c^q(t_b) p_i,c,t) / (sum_c w_i,c^q(t_b) p_i,c,t_b) (expression preserved as in source).
  - Budget shares:
    - Expenditure shares by COICOP division and quintile are spliced from three sources in priority order: Eurostat household budget surveys; shares compiled by Kpodar and Liu (2022); and shares computed from household consumption-expenditure microdata in the Luxembourg Income Study.
    - The priority is applied cell by cell; spliced shares are carried forward between observations.
  - Coverage:
    - The index is computed only for country-quarters in which all twelve division price indices are available.

### Adjusted poverty rates using income-specific CPI
- Adjustment factor and counterfactuals:
  - ω_i,t = (P_i,t^Q1 / P_i,base^Q1) / (P_i,t^head / P_i,base^head); values above one mean first-quintile households faced higher cumulative inflation than the average household since the base year.
  - H_i,t^cf(z) = H_i,t(z ⋅ ω_i,t)
  - ΔH_i,t(z) = H_i,t(z ⋅ ω_i,t) − H_i,t(z)
  - Headcount ratios are converted to counts using population data and then aggregated.
- Data and anchors:
  - Poverty headcounts are taken from the World Bank’s Poverty and Inequality Platform, using lined-up annual estimates, at the international line of $2.15 per day in 2017 purchasing-power-parity terms.
  - The baseline anchors the wedge at 2015.
  - Coverage requires an economy to have both a headline and a quintile-1 index in the base year, and a poverty headcount in the World Bank dataset.
- Aggregation:
  - The additional people in poverty for poverty line z equals sum_over_i∈S pop_i,t ⋅ ΔH_i,t(z), aggregated over the country sample S.

### Novel narrative database on fiscal policy interventions (1996–2025 EIU reports)
- Coverage and exclusions:
  - Quarterly Economist Intelligence Unit (EIU) reports from 1996 to 2025 for 26 advanced economies, 36 emerging markets, and 14 low-income countries.
  - Excludes monetary policy; broader policies that affect living costs only indirectly; military operations; humanitarian interventions by foreign governments; policies by supranational or non-governmental organizations; and interventions from sub-national authorities.
- Classification dimensions for each policy measure:
  - Commodity sector: whether addressing food and energy purchasing power and specific sub-products targeted.
  - Channel: (i) price-suppressing (e.g., reductions in VAT, sales taxes, excise duties, customs duties or import tariffs; subsidies to producers; introduction/tightening of price controls); (ii) income-supporting (e.g., cash, semi-cash—vouchers or discounts—or in-kind transfers to households; increases in public wages or pensions; reductions in income or corporate taxes); (iii) other measures (e.g., export bans; release of strategic reserves; government loans and guarantees to producers; equity injections into firms—including nationalization; public investment in productive capacity).
  - Timing: policies are assigned to the quarter in which they are introduced.
- Contextual variables:
  - Whether the country was experiencing elevated food and energy price inflation; social and economic consequences of price pressures; population groups affected.
- LLM extraction and validation:
  - LLM extraction produced standardized tabular output with variable name, coded answer, self-assessed confidence (1 to 100), justification (up to 500 words), and the quote (exact sentence(s) from the report).
  - A second LLM procedure assessed when a new measure was introduced versus mentions of ongoing, withdrawn, or not-implemented measures.
  - Human-validation covered a random sample of 21 percent of the countries covered in EIU reports.
  - Trained reviewers independently evaluated seven key variables.
  - Overall agreement between human- and LLM-produced assessments was 98 percent, with rates ranging from 94 to 100 percent across policy types.
- Caveats:
  - The database may understate less prominent or gradual measures not prominently discussed in EIU reports.
  - The validation addresses false positives but not false negatives, since reviewers assess coded entries rather than missing policies.

### Policy effectiveness: empirical approach and estimation details
- Objective:
  - Assess the effect of newly introduced price-suppressing and income-supporting measures on inflation, inflation expectations, aggregate consumption, and the fiscal balance.
- Identification and treatments:
  - Policies are typically adopted when cost-of-living pressures are acute; the analysis combines two-way fixed-effects local projections with a doubly robust augmented inverse-propensity-weighted (AIPW) estimator to address selection on observables.
  - Three mutually exclusive treatments: price-suppressing measures; income-supporting measures; simultaneous use of both types.
  - Reference category: no new policy introduced.
- Two-stage estimation:
  - First stage (propensity model):
    - Multinomial logit models Pr(W_it = w) using regressors including four quarterly lags of headline, energy and food inflation; controls Z_it such as a fiscal-space index, food-import exposure, indicators of government ideology and public support for intervention, report-derived measures of social stress and whether relief is targeted at low-income households or pensioners, the actual and expected monetary-policy stance, and the expected direction of inflation; and calendar-year effects δ_year(t),w.
    - The fitted arm probabilities are truncated at their 5th and 95th percentiles and enter the second stage as inverse-propensity weights.
  - Second stage (policy responses):
    - Local projection outcome model for horizons h = 0, 1, ..., 8:
      - y_i,t+h − y_i,t−1 = α_i + δ_t + Σ_w θ_w(h) 1{W_it = w} + X_it' φ + Z_it' π + u_it,
      - Dependent variable is cumulative change from the first quarter before adoption to quarter h after adoption, winsorized at the 1st and 99th percentiles.
      - θ_w(h) gives the response to each policy arm relative to no new policy; α_i and δ_t are country and time fixed effects.
      - Standard errors clustered by country.
- Doubly-robust AIPW estimator:
  - Uses fitted outcomes μ̂_w,it(h) and fitted arm probabilities p̂_w,it to construct the effect τ̂_w(h) as an average of an augmentation term and re-weighted residuals, correcting for selection into treatment.

### Theoretical welfare framework and calibration (welfare impact of cost-of-living crises)
- Model setting and household problem:
  - Deterministic, discrete time; continuum of households; J consumption goods; K+1 assets (real and nominal).
  - Household types determined by age a and group g; within type households identical and earn labor income w_t.
  - Assets pay nominal or real dividends {D_k,t} and trade at prices {Q_k,t}; assets subject to convex adjustment costs χ_k(ΔN_k,t^ag).
  - Government transfers (or taxes if negative) T_t^ag to ag-type households.
  - Representative ag household maximizes present-value utility Σ_t β^t U(C_t^ag, L_t^ag, {N_k,t^ag}_k∈ℛ, {N_k,t^ag}_k∈ℕ) subject to a period-by-period budget constraint expressed in real terms (x̃_t ≡ x_t / P_t), with implicitly-defined consumption aggregator and explicit notation preserved from source.
- Good-specific taxes and subsidies:
  - Observed price p_j,t = p_j,t* (1 + τ_j,t), where p_j,t* is pre-tax price and τ_j,t captures good-specific taxes (positive) or subsidies (negative).
- Money-metric welfare gain to a shock dz:
  - Defined as dV_ag/dz divided by λ_0 (marginal utility of dollar consumption at time zero), interpretable as willingness to pay measured in time-zero dollars.
  - Expression for dV_ag/dz / λ_0 decomposes welfare change into components:
    - Loss in real income.
    - Loss from composition of consumption basket.
    - Changes in nominal and real wealth.
    - Changes in dividends and asset-price terms.
    - Indexation of nominal income.
    - Role of price-suppressing and income-supporting measures.
  - Online Annex Figure 2.4.1 (referenced) shows the relative importance of each component.
- Calibration:
  - Framework calibrated to the 2021–23 crisis in a sample of 18 European countries (calibration reference preserved as in source).
- Key statements on welfare decomposition:
  - 2.7 in the main text shows the total welfare losses implied by this equation, as well as the ones that would have occurred absent fiscal policy interventions (last component of the equation).
  - Due to data availability, the calibration excludes the role of indexation of nominal income.
- Data sources and notes for the welfare decomposition figure:
  - Sources: Consensus Economics; European Central Bank, Household Finance and Consumption Survey; European Commission, Weekly Oil Bulletin; Eurostat, Household Budget Surveys; Jordà and Taylor 2019; national sources; Sgaravatti and others 2023; and IMF staff calculations.
  - The figure presents a welfare decomposition for the 2021–23 Cost-of-Living Crisis (Percent of three-year income) across income quintiles for a panel of 18 European countries.

### Model-Based Assessment of Policy Tradeoffs During Cost-of-Living Crises (GDN model)
- Data sources and measurement assumptions:
  - Realized monthly paths of aggregate prices are taken from Eurostat HICP.
  - One-year ahead inflation expectations as of December 2020 are from Consensus Economics.
  - Two- and three-year ahead inflation expectations are assumed to be equal to average inflation over 2015-2019.
  - Consumption composition at the 4-digit COICOP level is computed from the Eurostat Household Budget Survey (2015); consumption weights use 2015 data and consumption levels are updated to end of 2020 using realized inflation by 4-digit category between end of 2015 and end of 2020.
  - (Gross) income and households’ balance sheet positions come from the Household Finance and Consumption Survey (ECB), calibrated to 2017 and updated with realized inflation until end of 2020. Nominal and real variable definitions preserved as in source.
  - Changes in bond, share and house prices in response to the unexpected rise in inflation use elasticities from Pallotti and others (2024): -0.726 (bonds), -0.41 (shares), and -0.138 (house prices).
  - Data on fiscal policy interventions (price-suppressing and income-supporting) implemented in response to the energy shock of 2022 are obtained from Sgaravatti and others (2023).
- Model overview and structure:
  - Global Dynamic Network (GDN) model, quarterly, two regions (Home and RoW). Separate Home–RoW pairs calibrated for a representative advanced economy, emerging market economy, and low-income country.
  - Each region contains 12 production sectors, two heterogeneous household groups, a government, and a monetary authority.
  - Production uses a nested CES structure with heterogeneous capital types (Agriculture, Mining, Oil-Extraction, Other).
  - Sectoral Phillips-curve relation:
    - log(π_{r,s,t}/π_{r,s,t}^{ref}) = β log(π_{r,s,t+1}/π_{r,s,t+1}^{ref}) + κ_{r,s} log(MC_{r,s,t}/P_{r,s,t})
    - Parameter κ_{r,s} calibrated using frequency of price adjustment 훿_{r,s} from Ozhan and others (Forthcoming), except for food and energy-related sectors (Online Annex Table 2.5.5).
- Households and aggregation:
  - Two household types per region: High-income households (HIHHs) and Low-income households (LIHHs).
  - LIHH population shares used in calibrations:
    - LIHHs account for 20 percent of households in the advanced economy calibration.
    - LIHHs account for 40 percent in the emerging market and low-income developing country calibrations.
  - Consumption aggregate and monetary-policy-targeted aggregate consumer price:
    - C_{r,h,t} = ( ∑_{j=1}^{12} (α_{r,j→h}^{c})^{1/e_{r}^{c}} C_{r,j→h,t}^{(e_{r}^{c}-1)/e_{r}^{c}} )^{e_{r}^{c}/(e_{r}^{c}-1)}
    - π_{r,t} = λ_{r,L} π_{r,L,t} + λ_{r,H} π_{r,H,t}
- Monetary policy, fiscal constraint, and welfare:
  - Taylor-rule focused on headline CPI with forward-looking weight and interest-rate smoothing; full equation preserved in source.
  - Government budget constraint and public-debt fiscal rule preserved as in source.
  - Welfare: Per-period utility is CRRA over a Cobb-Douglas composite of consumption and leisure; welfare effects of policies evaluated relative to counterfactual without fiscal intervention and expressed as permanent percentage change in no-policy lifetime consumption.
- Mapping household expenditure to model sectors:
  - COICOP mapped to GLORIA industries following HBS-to-GLORIA method of Merkle and Dolphin (2025).
  - Advanced economy calibration uses 112 four-digit COICOP categories → 60 GLORIA industries → aggregated to 12 model sectors; EM and LIC calibrations map 12 one-digit COICOP divisions directly to 12 sectors.
- Sector and parameter details:
  - Online Annex Table 2.5.4. Net Exports of Food and Energy Products (Percent of home GDP):
    - Advanced Economy: Food −0.70; Energy −1.46
    - Emerging Market: Food −0.03; Energy −1.94
    - Low-Income Developing Country: Food 5.19; Energy −3.37
  - Selected parameter notes (Online Annex Table 2.5.5) preserve elasticity tuples and price-adjustment frequency parameter δ_{s,r} descriptions.

### Baseline global shock assumptions and policy scenario design
- Shock calibration (Online Annex Table 2.5.6):
  - Baseline energy shock: calibrated to raise the year-over-year global energy price by about 70 percent on impact.
  - Baseline food shock: calibrated to raise the global food price by about 45 percent on impact.
  - These increases correspond to roughly 2 to 2¼ historical standard deviations of the relevant global price series.
  - Baseline half-life: approximately two quarters.
  - Alternative (more persistent) half-life: about 13½ quarters.
- Design of policy scenarios:
  - Four temporary fiscal instruments compared one at a time:
    - Targeted transfers to low-income households (LIHHs).
    - Untargeted transfers to both household groups.
    - Consumer subsidies on downstream food or energy goods (analogue of temporary reductions in VAT or excise taxes).
    - Producer subsidies on intermediate inputs from the sector in which the shock originates.
  - Policy-relevant bundles:
    - Food bundle: household purchases from agriculture (S01) and food manufacturing (S03).
    - Energy bundle: petroleum and chemical products (S05) and utilities (S09).
  - Benchmark “equal-protection” exercise:
    - Aims at containing the impact on households from higher consumer prices by one half.
    - Transfers set to cover one-half of the increase in the cost of purchasing pre-shock quantities of the relevant bundle (scale χ_k = 0.5 for targeted transfers).
    - Subsidy rates chosen endogenously so the increase in the corresponding household bundle price is one-half of the increase under the no-policy baseline.
    - Consumer and producer subsidies assumed to be fully passed through to final purchasers and intermediate users, respectively.

### Policy instrument definitions and fiscal channels
- Targeted transfers:
  - For bundle k, transfers to LIHHs proportional to the increase in the cost of purchasing the pre-shock quantity of the LIHHs.
  - Scale χ_k set to 0.5 so that the transfer compensates for one-half of the additional bill.
  - Transfer expressed in nominal units and enters the LIHHs’ budget constraint directly.
- Untargeted transfers:
  - Same aggregate compensation distributed across both household types using household share λ_h and χ_k = 0.5 scaling.
- Consumer subsidies:
  - Lower effective purchaser price of final food- or energy-related goods and services.
  - Nominal price paid by households with consumer subsidy rate τ_r,e,t^hh is: P_r,e,t^hh ≡ (1 − τ_r,e,t^hh) P_r,e,t + τ_r,e,t^hh P_r,e.
  - Government finances difference between market price and subsidized price on actual household purchases; subsidy applies to both household groups.
- Producer subsidies:
  - Reduce effective price paid by firms for inputs from the mining and quarrying sector (S02) for energy shocks.
  - For energy-related sectors e (S05 and S09), effective intermediate-input price entering firm pricing block: P_r,e,t^f ≡ (1 − τ_r,e,t^f) P_r,e,t + τ_r,e,t^f P_r,e.
  - Fiscal cost equals the subsidy wedge multiplied by eligible intermediate-input use.
- Producer-subsidy export channel and fiscal implications:
  - Producer subsidies reduce domestic downstream marginal costs and final prices, reduce export prices and stimulate foreign demand.
  - Empirical result (advanced-economy calibration, energy shock): producer subsidies generate a sharp increase in exports from downstream energy-related sectors S05 and S09; consumer subsidies leave those exports broadly unchanged or slightly lower.
  - Export leakage: producer subsidies direct part of fiscal support to foreign purchasers, requiring a substantially larger intervention than consumer subsidies to achieve the same reduction in the domestic retail-price increase.

### Persistence, de-anchoring, international spillovers, and comparative results
- Persistence and fiscal costs:
  - Greater persistence raises fiscal costs because consumer and producer subsidy rates must remain elevated for longer, raising public debt and taxes required during consolidation—especially under producer subsidies.
  - Targeted transfers become more costly under greater persistence, but their relative advantage increases because support remains concentrated on LIHHs.
  - Under a common fiscal envelope, the persistent shock widens the welfare advantage of targeted transfers for LIHHs.
- De-anchored inflation expectations:
  - Baseline assumes inflation expectations remain anchored.
  - Alternative specification allows large increases in salient food and energy prices to affect the reference inflation entering sectoral price setting via a de-anchoring factor: π_r,s,t^deanchor = (π_r,t^food / π̅_r)^{ν_food} (π_r,t^energy / π̅_r)^{ν_energy}.
  - Elasticities ν_food and ν_energy calibrated so that, after the advanced economy energy shock, non-energy inflation rises by about 0.85 percentage point more than under anchored expectations.
  - De-anchoring produces a much larger and more front-loaded non-energy inflation response; consumer subsidies, by lowering the retail energy-price increase, dampen the shift in inflation on impact and associated second-round effects.
- International spillovers and simultaneous policies:
  - Simultaneous consumer subsidies by Home and RoW, each targeting one-half of pass-through to retail energy bundle, weaken the demand contraction that would otherwise help clear the global market and raise the global energy price relative to the no-policy baseline.
  - In the calibration reported in the main text, the peak difference is about 3 percentage points.
  - The higher global price partly offsets domestic price relief and lowers the welfare gain from domestic consumer subsidies.
- Comparative quantitative results (selected findings):
  - For the global food shock in advanced-economy calibration: consumer subsidies cost almost 7 times as much as targeted transfers while providing a similar near-term improvement in LIHH consumption.
  - In low-income country calibration, containing pass-through by one-half costs roughly 22 times more under producer subsidies than under targeted transfers; comparable ratios reported as about 16 times in advanced economies and 14 times in emerging markets.
  - Across advanced economies, emerging markets, and low-income developing countries, subsidies reduce inflation more sharply than transfers but require substantially more public borrowing.
  - Under a common fiscal envelope, targeted transfers produce the largest welfare gains for LIHHs.
  - Consumer subsidies provide weaker welfare support for LIHHs despite near-term effects on consumption.
  - The fiscal cost of consumer subsidies rises markedly with the economy’s exposure to food prices, whereas targeted transfers provide comparable protection to vulnerable households at a much lower fiscal cost.

*Source: International Monetary Fund. Online Annex 2.4–2.5, World Economic Outlook, October 2026 (ch2onlineannex).*

### 1. World

### 1. World

### Relative prices, real wages, and inflation inequality under alternative treatments
- Under the alternative world-commodity-price classification:
  - The relative prices of essentials rises steadily, reaching 10 percent by the tenth quarter.
  - Real wages hold up during the first year before falling to 4 percent below their pre-onset level.
  - The inflation gap between quintiles 1 and 5 turns positive after the fourth quarter and reaches 2½ log points by the tenth quarter.
- Oil supply news shock (Känzig 2021), normalized to a 150 percent increase in the world oil price on impact:
  - Raises the relative prices of essentials by 9 percent within two quarters and by 13 percent by the eighth quarter.
  - Real wages fall to a trough 9½ percent below their pre-shock level in the seventh and eighth quarters and recover to only 4½ percent below that level by the twelfth quarter.
  - The inflation gap between quintile-specific price indices widens to 3 log points by the seventh quarter.
- Euro area pure monetary policy surprise (Jarociński and Karadi 2020), normalized to a 25-basis-point easing surprise:
  - A 25-basis-point easing surprise leads neither to a significant drop in real wages nor to a significant increase in the gap in the inflation experienced by low-income households relative to that experienced by high-income households.
  - This suggests that the baseline results largely capture global supply shocks rather than demand-driven episodes.
- Estimation and presentation notes:
  - Results are from local-projection estimates on a quarterly panel of countries over 2005–25 with country fixed effects.
  - Shaded areas indicate 90 percent confidence intervals based on country-clustered standard errors.
  - The relative price of basic necessities is the CPI for food, transport, and housing-related expenses, including utilities, relative to the CPI for all other items.
  - Real wages are nominal wages deflated by headline CPI.

### Cross-country distribution of CPI weights by income quintile (basic necessities)
- Basic necessities composition:
  - Includes "Food and non-alcoholic beverages", "Housing, water, and home energy", and "Transport".
- Cross-country distribution depiction:
  - Boxplot edges represent interquartile range, the line inside each box indicates the median country, and whiskers denote the 90th and 10th percentiles.
- Implication:
  - Lower-income households allocate a larger share of their consumption basket to basic necessities, making them disproportionately affected by rises in the price of basic necessities.
- Country groups shown:
  - AEs (advanced economies), EMs (emerging markets), LIDCs (low-income developing countries).

### Distributional Consumer Price Index (income-specific CPI) construction and budget shares
- Definition (per Jaravel 2026):
  - For household quintile q in economy i in quarter t, the income-specific price index P_i,t^q is constructed using chained fixed-basket indices that weight COICOP division price indices by that quintile’s budget shares, with t_b denoting the quarter of the most recent measurement of expenditure weights at or before t.
  - P_i,t^q = P_i,t^b_q * (sum_c w_i,c^q(t_b) p_i,c,t) / (sum_c w_i,c^q(t_b) p_i,c,t_b) (expression preserved as in source).
- Budget shares:
  - Expenditure shares by COICOP division and quintile are spliced from three sources in priority order: Eurostat household budget surveys; shares compiled by Kpodar and Liu (2022); and shares computed from household consumption-expenditure microdata in the Luxembourg Income Study.
  - The priority is applied cell by cell; spliced shares are carried forward between observations.
- Coverage:
  - The index is computed only for country-quarters in which all twelve division price indices are available.

### Adjusted poverty rates using income-specific CPI
- Adjustment factor:
  - ω_i,t = (P_i,t^Q1 / P_i,base^Q1) / (P_i,t^head / P_i,base^head); values above one mean first-quintile households faced higher cumulative inflation than the average household since the base year.
- Counterfactual headcount and additional poverty attributable to inflation inequality:
  - H_i,t^cf(z) = H_i,t(z ⋅ ω_i,t)
  - ΔH_i,t(z) = H_i,t(z ⋅ ω_i,t) − H_i,t(z)
  - Headcount ratios are converted to counts using population data and then aggregated.
- Data and anchors:
  - Poverty headcounts are taken from the World Bank’s Poverty and Inequality Platform, using lined-up annual estimates, at the international line of $2.15 per day in 2017 purchasing-power-parity terms.
  - The baseline anchors the wedge at 2015.
  - Coverage requires an economy to have both a headline and a quintile-1 index in the base year, and a poverty headcount in the World Bank dataset.
- Aggregation note:
  - The additional people in poverty for poverty line z equals sum_over_i∈S pop_i,t ⋅ ΔH_i,t(z), aggregated over the country sample S.

### Novel narrative database on fiscal policy interventions (1996–2025 EIU reports)
- Coverage:
  - Quarterly Economist Intelligence Unit (EIU) reports from 1996 to 2025 for 26 advanced economies, 36 emerging markets, and 14 low-income countries.
- Policy perimeter and exclusions:
  - Excludes monetary policy; broader policies that affect living costs only indirectly (e.g., trade liberalization, labor market deregulation); military operations; humanitarian interventions by foreign governments; policies by supranational or non-governmental organizations; and interventions from sub-national authorities.
- Classification dimensions for each policy measure:
  - Commodity sector: whether addressing food and energy purchasing power and specific sub-products targeted.
  - Channel: (i) price-suppressing (e.g., reductions in VAT, sales taxes, excise duties, customs duties or import tariffs; subsidies to producers; introduction/tightening of price controls); (ii) income-supporting (e.g., cash, semi-cash—vouchers or discounts—or in-kind transfers to households; increases in public wages or pensions; reductions in income or corporate taxes); (iii) other measures (e.g., export bans; release of strategic reserves; government loans and guarantees to producers; equity injections into firms—including nationalization; public investment in productive capacity).
  - Timing: policies are assigned to the quarter in which they are introduced.
- Contextual variables recorded:
  - Whether the country was experiencing elevated food and energy price inflation; social and economic consequences of price pressures; population groups affected.
- LLM extraction method:
  - A structured prompt defined all variables and classification categories; the model produced standardized tabular output with variable name, coded answer, self-assessed confidence (1 to 100), justification (up to 500 words), and the quote (exact sentence(s) from the report).
  - A second LLM procedure assessed when a new measure was introduced versus mentions of ongoing, withdrawn, or not-implemented measures.
- Validation:
  - Human-validation exercise covered a random sample of 21 percent of the countries covered in EIU reports.
  - Trained reviewers independently evaluated seven key variables.
  - Overall agreement between human- and LLM-produced assessments was 98 percent, with rates ranging from 94 to 100 percent across policy types.
- Caveats:
  - The database may understate less prominent or gradual measures not prominently discussed in EIU reports.
  - The validation addresses false positives but not false negatives, since reviewers assess coded entries rather than missing policies.

### Policy effectiveness: empirical approach and estimation details
- Objective:
  - Assess the effect of newly introduced price-suppressing and income-supporting measures on inflation, inflation expectations, aggregate consumption, and the fiscal balance.
- Identification challenge:
  - Policies are typically adopted when cost-of-living pressures are acute; the analysis combines two-way fixed-effects local projections with a doubly robust augmented inverse-propensity-weighted (AIPW) estimator to address selection on observables.
- Treatments:
  - Three mutually exclusive treatments: price-suppressing measures; income-supporting measures; simultaneous use of both types.
  - Reference category: no new policy introduced.
- Two-stage estimation:
  - First stage (propensity model):
    - Multinomial logit models Pr(W_it = w) using regressors including four quarterly lags of headline, energy and food inflation; controls Z_it such as a fiscal-space index, food-import exposure, indicators of government ideology and public support for intervention, report-derived measures of social stress and whether relief is targeted at low-income households or pensioners, the actual and expected monetary-policy stance, and the expected direction of inflation; and calendar-year effects δ_year(t),w.
    - The fitted arm probabilities are truncated at their 5th and 95th percentiles and enter the second stage as inverse-propensity weights.
  - Second stage (policy responses):
    - Local projection outcome model for horizons h = 0, 1, ..., 8:
      - y_i,t+h − y_i,t−1 = α_i + δ_t + Σ_w θ_w(h) 1{W_it = w} + X_it' φ + Z_it' π + u_it,
      - Dependent variable is cumulative change from the first quarter before adoption to quarter h after adoption, winsorized at the 1st and 99th percentiles.
      - θ_w(h) gives the response to each policy arm relative to no new policy; α_i and δ_t are country and time fixed effects.
      - Standard errors clustered by country.
- Doubly-robust AIPW estimator:
  - Uses fitted outcomes μ̂_w,it(h) and fitted arm probabilities p̂_w,it to construct the effect τ̂_w(h) as an average of an augmentation term and re-weighted residuals, correcting for selection into treatment.

### Theoretical welfare framework and calibration (welfare impact of cost-of-living crises)
- Model setting:
  - Deterministic, discrete time; continuum of households; J consumption goods; K+1 assets (real and nominal).
  - Household types determined by age a and group g; within type households identical and earn labor income w_t.
  - Assets pay nominal or real dividends {D_k,t} and trade at prices {Q_k,t}; assets subject to convex adjustment costs χ_k(ΔN_k,t^ag).
  - Government transfers (or taxes if negative) T_t^ag to ag-type households.
  - Representative ag household maximizes present-value utility Σ_t β^t U(C_t^ag, L_t^ag, {N_k,t^ag}_k∈ℛ, {N_k,t^ag}_k∈ℕ) subject to a period-by-period budget constraint expressed in real terms (x̃_t ≡ x_t / P_t), with implicitly-defined consumption aggregator and explicit notation preserved from source.
- Good-specific taxes and subsidies:
  - Observed price p_j,t = p_j,t* (1 + τ_j,t), where p_j,t* is pre-tax price and τ_j,t captures good-specific taxes (positive) or subsidies (negative).
- Money-metric welfare gain to a shock dz:
  - Defined as dV_ag/dz divided by λ_0 (marginal utility of dollar consumption at time zero), interpretable as willingness to pay measured in time-zero dollars.
  - Expression for dV_ag/dz / λ_0 decomposes welfare change into components:
    - Loss in real income.
    - Loss from composition of consumption basket.
    - Changes in nominal and real wealth.
    - Changes in dividends and asset-price terms.
    - Indexation of nominal income.
    - Role of price-suppressing and income-supporting measures.
  - Online Annex Figure 2.4.1 (referenced) shows the relative importance of each component.
- Calibration:
  - Framework calibrated to the 2021–23 crisis in a sample of 18 European countries (calibration reference preserved as in source). 

*International Monetary Fund | October 2026 — Online Annex (ch2onlineannex - 1. World)*

### 2.7 in the main text shows the total welfare losses

### Welfare Decomposition for the 2021–23 Cost-of-Living Crisis

### Key statements
- 2.7 in the main text shows the total welfare losses implied by this equation, as well as the ones that would have occurred absent fiscal policy interventions (last component of the equation).
- Due to data availability, the calibration excludes the role of indexation of nominal income.

### Calibration
- Following Pallotti and others (2024), the framework is calibrated to the 2021-23 crisis in a sample of 18 European countries (see Online Annex Table 2.1.1 for the sample composition).
- Changes in the price level 풅 풍풏푷 풕 풅풛. Changes in the price level are computed as deviations from the expected trend growth in prices in the absence of the shock.
- Flow variables (such as income) are discounted using deviations of monthly inflation relative to expectations for the entire period 2021-23 as follows:

- 푑 푙표푔 푃 푑푧 푌 а푔 =∑ [ 푙표푔 푃 푡 푃 퐷푒푐2020 −푙표푔( 푃 푡 푃 퐷푒푐2020 ) ∗ ] × 푌 а푔,푝푟푒퐶푟푖푠푖푠 12 2023 푡= 2021

### Figure and welfare channels (Online Annex Figure 2.4.1)
- The figure presents a welfare decomposition for the 2021–23 Cost-of-Living Crisis (Percent of three-year income) across income quintiles for a panel of 18 European countries between 2021–.
- Channels plotted in the chart:
  - Income Quintile (x-axis)
  - Consumption basket
  - Fiscal: price intervention
  - Fiscal: transfers
  - Purchasing power
  - Wealth channel
  - Total welfare (with fiscal)
  - Total welfare (no fiscal)

### Data sources and notes
- Sources: Consensus Economics; European Central Bank, Household Finance and Consumption Survey; European Commission, Weekly Oil Bulletin; Eurostat, Household Budget Surveys; Jordà and Taylor 2019; national sources; Sgaravatti and others 2023; and IMF staff calculations.
- Note: The chart plots each channel affecting the welfare impact of households along the income quintiles from a panel of 18 European countries between 2021–.

*Online Annex Figure 2.4.1; Sources: Consensus Economics; European Central Bank; European Commission; Eurostat; Jordà and Taylor 2019; national sources; Sgaravatti and others 2023; and IMF staff calculations.*

### 23. The country average is weighted by 2019 GDP at purchasing power parity.

### Online Annex 2.5. Model-Based Assessment of Policy Tradeoffs During Cost-of-Living Crises

### Data sources and measurement assumptions
- Realized monthly paths of aggregate prices are taken from Eurostat HICP.  
- One-year ahead inflation expectations as of December 2020 are from Consensus Economics.  
- Two- and three-year ahead inflation expectations are assumed to be equal to average inflation over 2015-2019.  
- Consumption composition at the 4-digit COICOP level is computed from the Eurostat Household Budget Survey (2015); consumption weights use 2015 data and consumption levels are updated to end of 2020 using realized inflation by 4-digit category between end of 2015 and end of 2020.  
- (Gross) income and households’ balance sheet positions come from the Household Finance and Consumption Survey (ECB), calibrated to 2017 and updated with realized inflation until end of 2020. Nominal variables include deposits (da2101), bonds (da2103), money owed to households (da2107), monthly flow of mortgage and non-mortgage payments including interest payments (dl2000). Real variables include real estate (da1110 and da1120) and equity shares (da2105).  
- Changes in bond, share and house prices in response to the unexpected rise in inflation use elasticities from Pallotti and others (2024): -0.726 (bonds), -0.41 (shares), and -0.138 (house prices).  
- Interest-rate data come from ECB MIR database, IMF IFS and national sources.  
- Data on fiscal policy interventions (price-suppressing and income-supporting) implemented in response to the energy shock of 2022 are obtained from Sgaravatti and others (2023).  

### Counterfactual derivations for energy price measures and transfers
- Counterfactual price change for energy product e in month t and country c:
  - ∆price_{e,t}^{c} = Total fiscal budget on price-suppressing measures_{e,t} / Monthly consumption of e_{preCrisis}
  - e ∈ {electricity, natural gas, gasoline}; households are implicitly assumed to have kept energy consumption unchanged during the crisis period.  
- Monthly budget allocations: budget cost divided by number of months measure is implemented (complemented with national sources when missing in Sgaravatti and others 2023).  
- Monthly consumption of electricity and natural gas (kWh) and gasoline (liters) by country obtained from Eurostat and European Commission Weekly Oil Prices.  
- Monthly retail price series constructed by combining retail prices in early 2021 with monthly evolution of corresponding 4-digit COICOP HICP components. Adding ∆price_{e,t}^{c} to the price time series yields the counterfactual price series absent policy intervention.  
- Counterfactual transfers: total budget for each transfer measure divided by number of months in effect to yield monthly allocation; monthly allocations aggregated across active transfer measures; resulting amount divided by targeted population group size using detailed UN population data.

### Model overview and structure
- Model: Global Dynamic Network (GDN) model, quarterly, two regions (Home and RoW). Separate Home–RoW pairs calibrated for a representative advanced economy, emerging market economy, and low-income country. Each region contains 12 production sectors, two heterogeneous household groups, a government, and a monetary authority.  
- Production: Each sector produces a differentiated good combining value added and a composite of intermediate inputs from every domestic and foreign sector with a nested CES structure. Capital is heterogeneous and classified in four categories: agriculture capital, mining capital, oil-extraction capital, and other physical capital. The calibrated input-output matrix fixes steady-state expenditure shares 훼_{va} and 훼_{ic} for each source sector and country.  
- Sectoral price setting: Firms are monopolistically competitive with sector-specific nominal rigidity. Sectoral Phillips-curve relation:
  - log(π_{r,s,t}/π_{r,s,t}^{ref}) = β log(π_{r,s,t+1}/π_{r,s,t+1}^{ref}) + κ_{r,s} log(MC_{r,s,t}/P_{r,s,t})
  - Parameter κ_{r,s} calibrated using frequency of price adjustment 훿_{r,s} from Ozhan and others (Forthcoming), except for food and energy-related sectors (Online Annex Table 2.5.5). Reference rate fixed at economy-wide inflation target without de-anchored inflation expectations.  

### Households and aggregation
- Two household types per region:
  - High-income households (HIHHs): overlapping generations (Blanchard 1985; Yaari 1965), supply labor, receive profits and labor income, pay taxes, and hold domestic government bonds and an internationally traded asset.  
  - Low-income households (LIHHs): liquidity-constrained, no access to financial assets; consumption pinned down by current labor income, share of distributed income, and fiscal transfers.  
- LIHH population shares used in calibrations:
  - LIHHs account for 20 percent of households in the advanced economy calibration.  
  - LIHHs account for 40 percent in the emerging market and low-income developing country calibrations.  
- Consumption aggregate for household h in region r:
  - C_{r,h,t} = ( ∑_{j=1}^{12} (α_{r,j→h}^{c})^{1/e_{r}^{c}} C_{r,j→h,t}^{(e_{r}^{c}-1)/e_{r}^{c}} )^{e_{r}^{c}/(e_{r}^{c}-1)}
  - Weights α_{c} constructed from household budget surveys.  
- Aggregate consumer price targeted by monetary policy:
  - π_{r,t} = λ_{r,L} π_{r,L,t} + λ_{r,H} π_{r,H,t}
  - λ_{r,L} is LIHH population share; π_{r,L,t}, π_{r,H,t} are inflation indices faced by LIHHs and HIHHs.

### Monetary policy, fiscal constraint, and welfare
- Monetary policy: Taylor-rule focused on headline CPI with weight on projected inflation:
  - log i_{r,t} = ρ_{r} log i_{r,t-1} + (1−ρ_{r}){ (1−φ_{r}) log π_{r,t} + φ_{r} log π_{r,t+1} + log R^{n} } + (1−ρ_{r}) γ_{r}{ (1−φ_{r}) log π_{r,t} + φ_{r} log π_{r,t+1} − log π̅_{r} }
  - i_{r,t} nominal policy rate; R^{n} neutral real rate; π̅_{r} inflation target; parameters ρ_{r}, φ_{r}, γ_{r} denote interest-rate smoothing, forward-looking inflation weight, and inflation-response coefficient.  
- Government budget constraint:
  - P_{r,t}^{G} G_{r,t} + T_{r,t} + SUBEXP_{r,t} + (1+i_{r,t}^{G}) B_{r,t-1} = TAXREV_{r,t} + B_{r,t}
  - P_{r,t}^{G} G_{r,t} nominal government expenditure; T_{r,t} transfers; SUBEXP_{r,t} subsidy expenditures; TAXREV_{r,t} tax revenue; B_{r,t} government debt issuance.  
- Public debt and fiscal rule:
  - def_{r,t}^{tar} = def_{r,t}^{pol} + ψ_{r} ( B_{r,t}/NGDP_{r,t} − b_{r}^{*} )
  - def_{r,t}^{tar} deficit-to-GDP target; def_{r,t}^{pol} policy expenditure over GDP under fiscal space; b_{r}^{*} pre-shock debt-to-GDP ratio; ψ_{r} chosen to imply a debt half-life of approximately four to five years.  
- Welfare: Per-period utility is CRRA over a Cobb-Douglas composite of consumption and leisure:
  - U_{r,h,t} = ( C_{r,h,t}^{α_{r,h}} (L̅_{r,h,t} − L_{r,h,t})^{1−α_{r,h}} )^{1−1/σ_{r,h}}
  - Lifetime welfare discounts continuation value using household-specific discount factor; for asset-holding households includes survival probability from OLG structure. Welfare effects of policies evaluated relative to counterfactual without fiscal intervention; results expressed as permanent percentage change in no-policy lifetime consumption that makes household indifferent between policy and no-policy scenarios.

### Mapping household expenditure to model sectors
- COICOP (consumption purpose) mapped to GLORIA industries (production) to produce household-specific expenditure weights and sectoral trade positions used in steady state. Procedure follows HBS-to-GLORIA method of Merkle and Dolphin (2025).  
- Country-level COICOP budget shares are combined using 2023 current-US-dollar GDP weights to form representative household baskets.  
- Advanced economy calibration:
  - Uses 112 four-digit COICOP categories, mapped to 60 GLORIA industries, then aggregated to 12 model sectors; distinguishes bottom 20 percent from upper 80 percent of households.  
- Emerging market and low-income calibrations:
  - Use 12 one-digit COICOP divisions mapped directly to 12 model sectors; distinguish bottom 40 percent from upper 60 percent of households.  
- Aggregation and partitioning:
  - For a sector with at least one COICOP match, group-specific sectoral expenditure r_{k,g} = f_{k} ( ∑_{i} C_{k i} h_{i g} / ∑_{i} C_{k i} t_{i} ), where t_{i} = ∑_{g} h_{i g}; GLORIA fixes total household final-demand envelope f_{k} while HBS data split that envelope across household groups.  
  - 60-to-12 aggregation for advanced economies uses binary partition matrix A: f_{12} = A f_{60}, R_{12} = A R_{60}. EM and LIC profiles mapped directly to 12 sectors.

### Sector and concordance details (model sector definitions and mappings)
- Online Annex Table 2.5.2: Aggregation of 60 GLORIA industries to 12 model sectors (S01–S12) with constituent industries listed (e.g., S01 Primary activities: Agriculture, forestry, and fishing; S02 Mining and quarrying; S03 Food, beverages, and tobacco manufacturing; S05 Petroleum, coke, and chemicals; S09 Utilities and waste; S12 ICT, finance, government, education, health, arts, and other services).  
- Online Annex Table 2.5.3: Division-level COICOP–GLORIA concordance used in direct mapping; examples:
  - COICOP 01 Food and Non-Alcoholic Beverages → S01, S03, S11
  - COICOP 04 Housing, Water, Electricity, Gas, and Other Fuels → S01, S02, S04, S05, S06, S07, S08, S09, S10, S11, S12

### Calibration, parameters, and shock process
- Baseline calibration designed to generate rapid pass-through of adverse commodity-supply shocks to prices of food- and energy-related goods and services. Price-adjustment parameters and substitution elasticities aligned with Ozhan and others (Forthcoming) and reaggregated for 12-sector calibration. Online Annex Table 2.5.5 summarizes key parameters.  
- Online Annex Table 2.5.4. Net Exports of Food and Energy Products in Model Economies (Percent of home GDP):
  - Advanced Economy: Food −0.70; Energy −1.46  
  - Emerging Market: Food −0.03; Energy −1.94  
  - Low-Income Developing Country: Food 5.19; Energy −3.37  
- Selected parameter notes (Online Annex Table 2.5.5):
  - Elasticities of substitution tuples (e_{r,s}^{z}, e_{r,s}^{ic}, e_{r,s}^{tic}, e_{r,s}^{c}) reported by sector; example entries:
    - Agriculture (S01): (0.1, 0.1, 0.1, 0.5); capital use 0.99; capital type Agriculture Capital.  
    - Energy-related (S05, S09): (0.1, 0.1, 0.8, 0.5); capital use 0.99; capital type Oil-Extraction Capital.  
  - Price-adjustment frequency parameter δ_{s,r}: a value of 0.99 implies nearly flexible prices. For sectors outside S01, S02, S03, S05, S09, δ varies between 0.1 and 0.4.  
- Shock process representing cost-of-living crises:
  - Temporary negative TFP shocks in sector supplying the relevant global commodity in both Home and RoW (global supply disruption). Energy shock lowers productivity in mining and quarrying (S02); food shock lowers productivity in agriculture (S01). Productivity follows AR(1):
    - log A_{r,s,t} = ρ_{s} log A_{r,s,t−1} + ε_{r,s,t}
  - Innovation ε_{r,s,t} scaled separately for mining and quarrying and agriculture (scaling details provided in model specification).

*Source: International Monetary Fund. Online Annex 2.5, "Model-Based Assessment of Policy Tradeoffs During Cost-of-Living Crises", World Economic Outlook, October 2026.*

### Annex Table 2.5.6). The baseline energy shock is calibrated to raise the year-over-year global

### Baseline Global Shock Assumptions (Online Annex Table 2.5.6)

### Shock calibration
- Baseline energy shock: calibrated to raise the year-over-year global energy price by about 70 percent on impact.
- Baseline food shock: calibrated to raise the global food price by about 45 percent on impact.
- These increases correspond to roughly 2 to 2¼ historical standard deviations of the relevant global price series.
- Baseline half-life: approximately two quarters.
- Alternative (more persistent) half-life: about 13½ quarters.

### Design of policy scenarios
- Four temporary fiscal instruments compared one at a time:
  - Targeted transfers to low-income households (LIHHs).
  - Untargeted transfers to both household groups.
  - Consumer subsidies on downstream food or energy goods (analogue of temporary reductions in VAT or excise taxes).
  - Producer subsidies on intermediate inputs from the sector in which the shock originates.
- Policy-relevant bundles:
  - Food bundle: household purchases from agriculture (S01) and food manufacturing (S03).
  - Energy bundle: petroleum and chemical products (S05) and utilities (S09).
- Benchmark “equal-protection” (equal-price-suppression) exercise:
  - Aims at containing the impact on households from higher consumer prices by one half.
  - Transfers set to cover one-half of the increase in the cost of purchasing pre-shock quantities of the relevant bundle.
  - Subsidy rates chosen endogenously so the increase in the corresponding household bundle price is one-half of the increase under the no-policy baseline.
  - Consumer and producer subsidies assumed to be fully passed through to final purchasers and intermediate users, respectively.

### Policy instrument definitions and scaling
- Targeted transfers:
  - For bundle k, transfers to LIHHs proportional to the increase in the cost of purchasing the pre-shock quantity of the LIHHs.
  - Scale χ_k set to 0.5 so that the transfer compensates for one-half of the additional bill.
  - Transfer expressed in nominal units and enters the LIHHs’ budget constraint directly.
- Untargeted transfers:
  - Same aggregate compensation distributed across both household types.
  - Payments to household h use household share λ_h and the same χ_k = 0.5 scaling.
- Consumer subsidies:
  - Lower effective purchaser price of final food- or energy-related goods and services.
  - Nominal price paid by households with consumer subsidy rate τ_r,e,t^hh is: P_r,e,t^hh ≡ (1 − τ_r,e,t^hh) P_r,e,t + τ_r,e,t^hh P_r,e.
  - Government finances difference between market price and subsidized price on actual household purchases; subsidy applies to both household groups.
- Producer subsidies:
  - Reduce effective price paid by firms for inputs from the mining and quarrying sector (S02) for energy shocks.
  - For energy-related sectors e (S05 and S09), effective intermediate-input price entering firm pricing block: P_r,e,t^f ≡ (1 − τ_r,e,t^f) P_r,e,t + τ_r,e,t^f P_r,e.
  - Fiscal cost equals the subsidy wedge multiplied by eligible intermediate-input use.

### Producer-subsidy export channel and fiscal implications
- Producer subsidies lower input costs, propagating through production networks to:
  - Reduce domestic downstream marginal costs and domestic final prices.
  - Reduce export prices and stimulate foreign demand.
- Empirical result (advanced-economy calibration, energy shock):
  - Producer subsidies generate a sharp increase in exports from downstream energy-related sectors S05 and S09.
  - Consumer subsidies leave those exports broadly unchanged or slightly lower.
- Export leakage:
  - Producer subsidies direct part of fiscal support to foreign purchasers.
  - Because of export leakage, producer subsidies require a substantially larger intervention than consumer subsidies to achieve the same reduction in the domestic retail-price increase.

### Persistence, fiscal costs, and welfare comparisons
- Greater persistence raises fiscal costs because consumer and producer subsidy rates must remain elevated for longer, raising public debt and taxes required during consolidation—especially under producer subsidies.
- Targeted transfers become more costly under greater persistence, but their relative advantage increases because support remains concentrated on LIHHs.
- Under a common fiscal envelope, the persistent shock widens the welfare advantage of targeted transfers for LIHHs.
- De-anchored inflation expectations:
  - Baseline assumes inflation expectations remain anchored.
  - Alternative specification allows large increases in salient food and energy prices to affect the reference inflation entering sectoral price setting via a de-anchoring factor: π_r,s,t^deanchor = (π_r,t^food / π̅_r)^{ν_food} (π_r,t^energy / π̅_r)^{ν_energy}.
  - Elasticities ν_food and ν_energy calibrated so that, after the advanced economy energy shock, non-energy inflation rises by about 0.85 percentage point more than under anchored expectations.
  - De-anchoring produces a much larger and more front-loaded non-energy inflation response; consumer subsidies, by lowering the retail energy-price increase, dampen the shift in inflation on impact and associated second-round effects.

### International spillovers and simultaneous policies
- Simultaneous consumer subsidies by Home and RoW, each targeting one-half of pass-through to retail energy bundle, weaken the demand contraction that would otherwise help clear the global market and raise the global energy price relative to the no-policy baseline.
  - In the calibration reported in the main text, the peak difference is about 3 percentage points.
  - The higher global price partly offsets domestic price relief and lowers the welfare gain from domestic consumer subsidies.

### Comparative quantitative results (selected findings)
- Cost comparisons and fiscal multipliers (summary of reported comparisons):
  - For the global food shock in advanced-economy calibration: consumer subsidies cost almost 7 times as much as targeted transfers while providing a similar near-term improvement in LIHH consumption.
  - In low-income country calibration, containing pass-through by one-half costs roughly 22 times more under producer subsidies than under targeted transfers; comparable ratios reported as about 16 times in advanced economies and 14 times in emerging markets.
- Cross-country patterns:
  - Across advanced economies, emerging markets, and low-income developing countries, subsidies reduce inflation more sharply than transfers but require substantially more public borrowing.
  - Under a common fiscal envelope, targeted transfers produce the largest welfare gains for LIHHs.
  - Consumer subsidies provide weaker welfare support for LIHHs despite near-term effects on consumption.
  - The fiscal cost of consumer subsidies rises markedly with the economy’s exposure to food prices, whereas targeted transfers provide comparable protection to vulnerable households at a much lower fiscal cost.

---


_Source: https://www.imf.org/-/media/files/publications/weo/2026/october/english/ch2onlineannex.pdf_