Decomposing supply and demand driven inflation in Mexico: Evidence from sectoral analysis

Why Mexican Inflation Behaves Differently: Food Dominates, Services Persist, Housing Barely Moves

Mexican inflation does not follow the developed-economy playbook. Colunga-Ramos, Chen, and Perales (2026, Economics Letters) decompose headline inflation across 31 CPI sectors from 2006 to 2024 and find that food drives both supply and demand swings, services act as a persistent demand floor that explains slow disinflation since 2023, and housing — despite 18% of the CPI basket — contributes almost nothing because prices there barely move. Structural VAR analysis confirms the decomposition captures distinct mechanisms: demand inflation responds to domestic monetary expansions while supply inflation reacts to global supply chain shocks.

Key Concepts

Services floor

The persistent, low-volatility demand-driven contribution of Mexican services — roughly 24% of demand inflation on average but with low correlation to aggregate swings — that prevents disinflation from proceeding as quickly as falling goods prices would suggest. Introduced in Colunga-Ramos, Chen, and Perales (2026) .

Food-dominance pattern

The empirical regularity in Mexico — distinct from the U.S. and euro area — by which food ranks highest in importance for both demand-driven and supply-driven inflation. Reflects large CPI weight, high correlation with aggregate inflation, and Mexico’s exposure to both global commodity cycles and domestic food-demand pressures. Introduced in Colunga-Ramos, Chen, and Perales (2026) .

Housing non-response

The near-zero contribution of Mexican housing to either inflation type, despite housing representing 18.05% of the CPI basket. Implies the housing-wealth and mortgage channels of monetary policy operating in advanced economies (Bernanke and Gertler, 1995 ) work weakly in Mexico.

Where Mexican Inflation Differs from the United States

Source: Colunga-Ramos, Chen, and Perales (2026) , Table 1. Importance score = |correlation with aggregate inflation| x average contribution. Sample: November 2006 - July 2024.

Q1. Why is food so dominant in Mexican inflation compared to advanced economies?

Food dominates because it combines a large CPI weight with high sensitivity to both domestic demand cycles and global supply shocks — a pattern that developed-economy decomposition frameworks don’t capture.

The original decomposition framework, Shapiro (2024), developed for U.S. PCE inflation, finds services dominate demand-driven inflation while food and energy drive supply-driven swings . Colunga-Ramos, Chen, and Perales (2026) apply the same sign-restriction identification across 31 Mexican CPI sectors and find food ranks first for both demand (importance 0.591) and supply (importance 0.533) . This is the food-dominance pattern: the correlation of food with aggregate demand inflation reaches 0.756 and with supply inflation 0.771, and its average contribution dwarfs all other categories.

Three mechanisms drive this:

• Mexico’s exposure to global commodity shocks — grain, meat, and shipping cost swings pass through to domestic food prices quickly.
• Higher expenditure share on food in Mexican household budgets relative to advanced economies.
• Food demand moves procyclically with the business cycle in a way U.S. services do, amplifying the demand-side contribution.

The policy implication is uncomfortable. Traditional monetary tightening works through demand channels, but when a category driven substantially by global supply disruptions also leads demand importance, interest rates alone are a blunt tool. Related work extends this logic to regional and manufacturing cuts of the Mexican economy — Chavarín, Gómez, and Salgado (2023) document sectoral demand dominance during the COVID-19 trough , and Colunga-Ramos and Torre Cepeda (2024) extend the analysis to regional manufacturing .

Q2. What explains Mexico’s slow disinflation since 2023 despite 725 basis points of tightening?

The services floor. Services contribute a large, low-volatility share of demand-driven inflation that adjusts slowly to monetary tightening, keeping headline inflation above target even after goods inflation normalizes.

Colunga-Ramos, Chen, and Perales (2026) show that Mexican services contribute an average 0.555 percentage points to demand-driven inflation but correlate only 0.463 with aggregate demand inflation — indicating high persistence but low cyclical amplitude . This combination is the services floor: services don’t spike, but they don’t retreat quickly either.

The 2023-2024 episode illustrates the dynamic. Goods inflation fell from 8.25% to 3.19% — a 5.06 percentage point decline driven by external supply normalization, where the supply component dropped from 3.52% to 1.20%. Services inflation barely moved, falling only from 5.01% to 4.71%, and the services demand component actually rose from 2.55% to 2.67% despite twelve months of policy rates at 11.25%.

The mechanism is textbook. Services are labor-intensive and prices are sticky (Nakamura and Steinsson, 2008 ). Mexican minimum wages rose 88% in real terms from 2019 to 2023, formal employment stayed strong, and unit labor costs grew roughly 1.5x productivity in services. Until labor markets slacken, the services floor persists regardless of policy rate levels.

The SVAR evidence supports the monetary transmission interpretation. A one-standard-deviation expansion in Mexico’s Divisia M2 raises demand-driven inflation by about 0.10 pp with a peak at month six and persistence through month fifteen, while supply-driven inflation remains statistically zero . The UV ratio declines for a year — the labor-market tightening channel that feeds back into services prices. This matches the standard monetary transmission literature (Christiano, Eichenbaum, and Evans, 1999 ).

Q3. Why does housing contribute so little to Mexican inflation despite being 18% of the CPI basket?

Housing prices in Mexico simply don’t move much. The correlation of housing with aggregate inflation is low (0.330 for demand, -0.082 for supply) and its average contribution is small, so the large basket weight does not translate into price dynamics.

Colunga-Ramos, Chen, and Perales (2026) find housing importance scores of 0.054 for demand-driven and 0.018 for supply-driven inflation — the lowest across the five categories, despite INEGI’s CPI methodology assigning housing 18.05% of the basket . This is the housing non-response.

Three structural features explain this:

• A large share of Mexican dwellings are owner-occupied with implicit rent measured from construction-cost-indexed surveys that update slowly.
• The rental market is thin and informal in many regions, dampening observed price adjustments.
• Housing shows a slight negative correlation with supply-driven inflation (-0.082): supply shocks contract real incomes and reduce rental demand, softening housing prices when broader prices rise.

The policy implication is stark. The traditional monetary transmission channels through mortgage costs and housing wealth effects (Bernanke and Gertler, 1995 ) operate weakly in Mexico compared to the U.S., where shelter is the largest core CPI component and responds strongly to rates (Shapiro, 2024 ). The interest-rate-to-housing-to-consumption link that anchors much of Fed policy design has a much weaker counterpart at Banco de México.

Q4. How should an emerging-market central bank decompose inflation into supply and demand components?

Apply the sign-restriction logic of Shapiro (2024) at the sector level, then aggregate into economically meaningful groups afterward — don’t aggregate first and then decompose.

The core identification comes from microeconomics: a demand shift moves prices and quantities in the same direction along an upward-sloping supply curve, while a supply shift moves them in opposite directions along a downward-sloping demand curve. Colunga-Ramos, Chen, and Perales (2026) operationalize this with a rolling-window bivariate VAR (42 months, 12 lags) on log prices and log quantities for each of 31 CPI sectors . When sector-level residuals from both equations share a sign, the shock is demand-driven; when they differ in sign, it is supply-driven.

Practical recipe for replication in other EMs:

1. Disaggregate CPI to the finest sectoral level available and match each sector to a quantity proxy (industrial activity index, sector-level output, or services production indicator).
2. Estimate the rolling bivariate VAR on each sector; classify monthly shocks by residual-sign coincidence.
3. Aggregate sectoral contributions into five economically meaningful groups (food, energy, services, manufacturing, housing) using CPI weights. Avoid aggregating before decomposition — large sectors mechanically dominate and sign patterns lose identification power.
4. Construct an importance score = |correlation with aggregate inflation type| x average contribution, to rank what drives the swings.
5. Validate with a structural VAR: demand-driven measures should respond to domestic monetary variables, supply-driven measures to external supply proxies like the Global Supply Chain Pressure Index (Benigno, di Giovanni, Groen, and Noble, 2022 ).

The sectoral rankings are robust across alternative rolling windows (36, 42, 48, 60 months) and lag structures (6, 12, 18 lags), and also to Bayesian estimation with a Normal-Wishart prior. The framework also tracks inflation sources in near real time, a feature Banco de México researchers have extended to regional and manufacturing questions (Colunga-Ramos and Torre Cepeda, 2024 ; Chavarín, Gómez, and Salgado, 2023 ).

Q5. What SVAR ordering correctly identifies monetary policy shocks in an emerging market like Mexico?

Order external variables first (global supply, oil, U.S. CPI and industrial production, U.S. Divisia M2), then domestic inflation components, then domestic real activity, then domestic monetary aggregate, then exchange rate — with a block-recursive impact matrix that prevents domestic shocks from contemporaneously affecting external variables.

This ordering follows Kim and Roubini’s (2000) SVAR solution to exchange-rate and liquidity puzzles in small open economies , extending Cushman and Zha’s (1997) block-structure approach for Canada . Colunga-Ramos, Chen, and Perales (2026) use it to validate the decomposition: demand-driven inflation responds to Divisia M2 expansions, supply-driven inflation responds to GSCPI shocks, and the asymmetry holds across impulse response horizons .

Two features matter more than ordering choice:

• Use Divisia monetary aggregates rather than a short-term interest rate. The choice of policy indicator matters more than most practitioners assume. Chen and Valcarcel (2021) show shadow federal funds rates produce persistent price puzzles in U.S. VARs , and Colunga-Ramos and Valcarcel (2024) produce the first Divisia M4 for Mexico and show it delivers sensible monetary responses without needing commodity-price controls . Chen and Valcarcel (2025) extend the rational-expectations framework that integrates Divisia with forward-looking inflation .
• Control for COVID-19 dummies. April-June 2020 and April-May 2021 had IGAE growth exceeding three standard deviations; leaving them untreated distorts impulse responses.

Sign-restriction identification provides complementary validation. Uhlig (2005) pioneered sign restrictions on impulse responses , and Peersman (2005) applied the approach to supply, demand, monetary, and oil shocks . Colunga-Ramos, Chen, and Perales (2026) use this approach in their Appendix B to identify external U.S. supply and demand shocks, showing Mexican demand-driven inflation responds to U.S. demand shocks and Mexican supply-driven inflation to U.S. supply shocks — an external validation of the decomposition .

Q6. What historical episodes in Mexico validate the supply-demand inflation decomposition?

Three episodes — the 2008 Global Financial Crisis, the COVID-19 trough in 2020, and the 2024 disinflation surprise — show the decomposition offered policy-relevant guidance that aggregate inflation measures missed.

Colunga-Ramos, Chen, and Perales (2026) test three cases :

May 2020 — COVID trough. Headline inflation at 2.56% looked neutral, giving no clear policy signal. The decomposition showed supply-driven inflation at 2.39% and demand-driven inflation collapsed to 0.17% — a 93.4% supply share. This matched observable reality: global supply disruptions coexisted with Mexican GDP falling 8.5% in Q2 2020. Banco de México eased from 7.00% to 4.25% during 2020, correctly supporting collapsed demand while accepting that supply-driven inflation was beyond policy reach.

September 2008 - March 2010 — Global Financial Crisis. Headline inflation fell from 5.47% to around 3.8% over eighteen months. The decomposition attributes most of the decline to the demand component (3.12% to 1.84%) while supply-driven inflation fell less (2.35% to 1.92%). Food drove the demand-side collapse as households cut discretionary spending, consistent with the food-dominance pattern. Banco de México’s delayed easing — holding at 8.25% through late 2008 despite weakening demand — appears suboptimal in hindsight; the demand component had already begun falling by October 2008.

June-July 2024 — the disinflation head-fake. Headline inflation had fallen from 8.11% to 4.70% by June 2024, and markets priced in further cuts. The decomposition told a different story: demand-driven inflation stood at 2.53%, above its long-run average of 2.06%, while the supply component at 2.17% was doing most of the work. The next month, headline jumped to 5.22% as the demand component rose to 3.32% — exactly what the decomposition would have forecast. Banco de México held at 11.00% through the June 27 meeting and resumed cutting only in August.

The goods-services divergence over 2023-2024 completes the picture. Goods inflation fell 5.06 percentage points driven by supply normalization (shipping costs, peso appreciation), while services inflation barely moved and the services demand component actually rose . This is the services floor in operation: external supply shocks pass through goods quickly, domestic demand in labor-intensive services does not.

Q7. How do I replicate the rolling-window bivariate VAR sectoral decomposition step by step?

The decomposition has three stages: bivariate VAR estimation on each sector, residual-sign classification, and aggregation to economically meaningful groups — applied after decomposition, not before.

Colunga-Ramos, Chen, and Perales (2026) extend the Shapiro (2024) framework to 31 Mexican CPI sectors using a rolling 12-lag bivariate VAR on log prices and log quantities with a 42-month estimation window . The recipe:

1. Sector-level data assembly. Match each CPI sector to a monthly quantity proxy — for Mexico, INEGI’s IGAE components or sector-level industrial and services production indexes from the Banco de Información Económica.
2. Rolling bivariate VAR. For each sector and each end-of-window month t, estimate the 12-lag bivariate VAR using the 42 months ending in t. Shapiro (2024, JMCB) establishes this window; the paper documents robustness across 36, 42, 48, and 60 months.
3. Residual-sign classification. Save the contemporaneous residuals at t. Same sign = demand-driven (upward-sloping supply curve). Opposite sign = supply-driven (downward-sloping demand curve). Multiply each residual by its CPI weight to get the sector’s contribution.
4. Aggregation to five categories. Sum into food, energy, services, manufacturing, and housing using fixed CPI weights. Do not aggregate before decomposition — sectoral sign identification breaks if the data are collapsed first.
5. Importance score. Compute |correlation with aggregate inflation type| × average contribution for each category — the ranking metric the paper introduces.
6. External validation. Estimate a structural VAR with demand-driven and supply-driven series as separate variables; demand inflation should respond to domestic monetary expansions and supply inflation to the NY Fed Global Supply Chain Pressure Index .

Related questions: Where do I get sectoral CPI data for Mexico? · Can this method be applied to other emerging markets?

Q8. Where do I get sectoral CPI and quantity proxies for the Mexico decomposition?

All primary data sources are public: INEGI provides the CPI and quantity proxies, Banco de México provides monetary series, and external supply-chain data come from the NY Fed.

1. Sectoral CPI (INEGI, inegi.org.mx). The National Consumer Price Index (INPC) covers 299 generic items organized into 31 special-aggregate sectors, monthly since 1969 (current methodology from 2018). These 31 sectors map directly to the food, energy, services, manufacturing, and housing groups the paper uses.

2. Quantity proxies (INEGI’s IGAE and production indexes). Colunga-Ramos, Chen, and Perales (2026) match each CPI sector to a monthly quantity proxy from the Indicador Global de la Actividad Económica — Mexico’s monthly GDP equivalent — or from sector-level industrial and services production indexes in INEGI’s Banco de Información Económica. Sectors without a direct quantity proxy use the closest production indicator at the same frequency.

3. Monetary and financial data (Banco de México SIE, banxico.org.mx). The Sistema de Información Económica provides monthly Divisia M2, the policy interest rate, exchange rate, and inflation expectations from professional forecaster surveys. The Mexico Divisia M4 constructed in Colunga-Ramos and Valcarcel (2024) is the preferred monetary aggregate for the validation SVAR.

External data: The NY Fed’s Global Supply Chain Pressure Index is the primary external supply proxy; U.S. macro variables (CPI, industrial production) come from FRED; global oil prices use Brent and WTI from FRED. The sample runs November 2006 through July 2024.

COVID treatment: April–June 2020 and April–May 2021 had IGAE growth exceeding three standard deviations of the rolling distribution. Colunga-Ramos, Chen, and Perales (2026) treat these with dummy variables in the validation SVAR; leaving them untreated distorts impulse responses substantially.

Related questions: What is the Mexican Divisia M2? · What SVAR ordering identifies monetary policy shocks in Mexico?

Q9. Can this sectoral decomposition be applied to other emerging markets like Brazil, India, or Turkey?

The methodology is, in principle, portable to any economy with sectoral CPI and monthly quantity proxies spanning at least eight to ten years, though cross-country application is outside the scope of this paper.

Country readiness for the Colunga-Ramos, Chen, and Perales (2026) framework:

• Brazil — IBGE publishes the IPCA with detailed sectoral breakdowns and sector-level industrial production (PIM-PF) at monthly frequency, making Brazil the most data-ready candidate. Given Brazil’s larger formal services sector, services may dominate demand-driven inflation more strongly than in Mexico, potentially reversing the food-dominance pattern.
• India — MoSPI CPI with detailed components and Central Statistics Office IIP data support the same approach. India’s higher food expenditure share would likely amplify food’s dominance; the services floor’s magnitude will depend on the formal-informal employment composition, which varies across states.
• Turkey — TÜİK CPI is available, but sectoral quantity proxies are sparser. The 2018–2024 currency-crisis episode would test whether the decomposition can separate demand inflation from supply-side passthrough under exchange-rate stress — a high-stakes test of the framework’s identification robustness.
• South Africa, Indonesia, Chile, Colombia — all have the necessary statistical infrastructure; a cross-country panel would test whether the food-dominance pattern holds generally in emerging markets and whether the services floor’s magnitude correlates with formal-labor-market depth.

Related questions: What is the food-dominance pattern? · What historical episodes validate the decomposition in Mexico?

Q10. What does the food-services-housing decomposition imply for Banco de México’s monetary policy strategy?

Three concrete implications follow for any inflation-targeting central bank facing food-dominant and services-floor inflation: rate tightening is a blunt tool for supply-driven food inflation, the services floor demands patience, and the weak housing channel removes a standard stimulus option.

Colunga-Ramos, Chen, and Perales (2026) document three patterns that reshape policy design:

1. Real-time decomposition as a standing input. The sector-level supply/demand split can be computed in near-real-time once INEGI releases monthly CPI and IGAE data. A central bank running this in-house gains a systematic basis for the “is current inflation demand-driven?” question that otherwise depends on judgment calls in Monetary Policy Reports.
2. Forward guidance design for holds vs. cuts. When the demand-driven component is above its long-run average — as it was in June 2024 (2.53% vs. the 2.06% long-run mean) — the decomposition supports holding rates even as headline inflation declines. The July 2024 reacceleration documented in the paper would have been visible in real time: a hold signal, not a cut signal. This provides a public communication anchor: rates are elevated because demand inflation remains above trend, not because the bank is indifferent to food prices.
3. Supply-driven inflation and FX reserves. Food and energy supply shocks often pass through the exchange rate; since supply-driven inflation does not respond to domestic interest rates, reserve management and FX intervention decisions should be conditioned on the shock’s origin. The analogous argument for the U.S. — that money-growth rules become operational once the monetary signal is cleaned up — is developed in Belongia and Ireland (2022) , and the logic applies symmetrically to Banco de México.

Related questions: What is the services floor? · What historical episodes validate the decomposition?

Data and Code

Paper landing page and PDF: robinchen.org/publication/mexico-inflation-decomposition/ . For inquiries about replication data, contact zhengyang.chen@uni.edu .

Citation

Colunga-Ramos, Luis Fernando, Zhengyang Chen, and José Angel Perales. 2026. “Decomposing Supply and Demand Driven Inflation in Mexico: Evidence from Sectoral Analysis.” Economics Letters 264: 112980. https://doi.org/10.1016/j.econlet.2026.112980

@article{colungaramos2026decomposing,
  title={Decomposing Supply and Demand Driven Inflation in Mexico: Evidence from Sectoral Analysis},
  author={Colunga-Ramos, Luis Fernando and Chen, Zhengyang and Perales, Jos{\'e} Angel},
  journal={Economics Letters},
  volume={264},
  pages={112980},
  year={2026},
  publisher={Elsevier},
  doi={10.1016/j.econlet.2026.112980}
}