## wpiea2023205-print-pdf

## Source details

**Canonical URL:** [wpiea2023205-print-pdf](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023205-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023205-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023205-print-pdf.pdf.json)

---

### 1.1 Literature Review — Literature and contribution
- Builds on Shapiro (2022) novel methodology for decomposing inflation into demand and supply components using monthly PCE data (1990 to 2023); this methodology is adopted in the study.
- Extends Sheremirov (2022) which categorized inflation into transitory and persistent demand-driven and supply-driven components for the U.S.
- Expands limited cross-country implementations (Gonçalves and Koester (2022) for aggregate Euro Area; OECD (2022) for eight advanced economies) by applying the approach to 32 advanced and emerging market economies.
- First systematic examination, to the authors’ knowledge, of how the slope of the Phillips curve varies when inflation is demand-driven vs supply-driven across 32 countries (building on U.S.-focused evidence from Bergholt, Furlanetto, and Vaccaro-Grange (2023)).
- Demonstrates that monetary policy strongly and significantly affects demand-driven inflation while supply-driven inflation is less responsive to monetary policy shocks.
- Shows supply chain pressures (using GSCPI from Benigno et al. (2022)) contributed to inflation across 32 countries, increasing supply-driven inflation with no significant impact on demand-driven inflation.

### 1.1 Data (summary of sources and scope)
- Sectoral PCE data at quarterly frequency from Haver Analytics and Eurostat; monthly PCE series available for the U.S. but quarterly series used for cross-country consistency.
- Sample span: from the 1990s to 2023Q2.
- Coverage: 32 countries (23 advanced economies and 9 emerging market economies).
- Sectoral series are seasonally adjusted (either by the source or by the authors) using X13-ARIMA-SEATS; allows for additive outliers and level shifts.
- Note on item coverage: BEA provides 99 price and quantity items for the U.S.; several European countries have only 4 sectors. For countries with few sectors, two items may account for over 80% of total expenditure.
- Validation data:
  - Phillips curve exercise: quarterly output gap (HP filter), one-year ahead inflation expectations from Consensus Forecasts (CF), import prices from HAVER.
  - Monetary policy transmission: quarterly monetary policy shock series from Deb et al. (2023).
  - Oil shocks: oil supply shocks from Baumeister and Hamilton (2019).
  - Supply chain pressures: GSCPI from Benigno et al. (2022).

### 1.1 Inflation decomposition methodology
- Follows Shapiro (2022) and Sheremirov (2022): classify inflation in each expenditure item as demand-driven or supply-driven based on sign co-movement of price and quantity shocks.
- Estimation:
  - For each country c and expenditure item i, estimate a VAR in first differences with p = 4 lags:
    y_cit = sum_{h=1}^p C_cij y_ci,t-h + ν_cit,
    where y_cit = (Δp_cit, Δq_cit) and ν_cit = (ν^p_cit, ν^q_cit).
  - Baseline classification: item is demand-driven if ν^p_cit * ν^q_cit ≥ 0; supply-driven if ν^p_cit * ν^q_cit < 0.
  - Aggregate demand-driven and supply-driven inflation series computed as weighted sums of item-level inflation rates using year-to-date expenditure share weights ω_cit.
  - Aggregate inflation π_ct = π^d_ct + π^s_ct.
- Caveats:
  - Classification is noisy: sectors may be simultaneously affected by demand and supply shocks; the observable reduced-form sign pattern reflects the more prevalent shock.
  - Limited number of expenditure items in some countries can make decomposition sensitive to classification of a small number of sectors.

### Validation: narrative historical evidence
- Pandemic era (since 2019Q1):
  - At 2020Q1 lockdown outset, downward pressure on prices primarily via the demand channel; demand-driven factors had a negative contribution throughout 2020.
  - As reopening and supportive fiscal/monetary policies occurred, demand-driven inflation surged through end-2022.
  - Supply-driven inflation remained positive and relatively stable through the pandemic; its relative contribution was small during initial 2021 pick-up and began rising after 2022Q1, potentially linked to commodity price pressures after Russia’s invasion of Ukraine in early 2022.
  - U.S. and Canada: PCE inflation peaked in early 2022; subsequent slowdown was due to both supply-driven and demand-driven components.
  - Asian countries (except Japan): inflation decline began towards end-2022 with decreasing contributions from both demand and supply factors.
- Great Financial Crisis (GFC):
  - Supply-driven inflation trajectory aligns with commodity price cycle peaking mid-2008 then sharply declining.
  - Decline in demand-driven inflation beginning towards end-2008 consistent with buffer-stock saving behavior under income uncertainty (Carroll (1997)) and empirical evidence that uncertainty during the GFC increased saving rates and reduced consumption and GDP growth (Mody, Ohnsorge, and Sandri (2012)).
- Commodity price drop episode (2014–2016):
  - Crude oil prices fell to under $30 in February 2016 from around $106 in June 2014.
  - Similar declines in metals and agricultural products; decomposition shows a significant decline in supply-driven inflation across regions.

### Robustness checks
- Alternative specifications considered:
  - (i) 8 lags in VAR;
  - (ii) Lag length chosen by AIC;
  - (iii) Estimation sample ending in 2019Q4;
  - (iv) 10-year rolling window estimation;
  - (v) One-step-ahead forecast errors;
  - (vi) Classifying sectors with either price or quantity residuals smaller than 0.1 standard deviations as ambiguous;
  - (vii) Classifying the smallest 10% of price and quantity residuals as ambiguous;
  - (viii) Estimation in levels with linear deterministic trend;
  - (ix) Detrending level series with Hamilton regression filter (Hamilton2018);
  - (x) Detrending level series with Hodrick-Prescott filter with smoothing parameter 1600.
- Purpose: sensitivity to lag order, pandemic outliers, parameter instability, contemporaneous information only, concerns about items with small residuals, alternative trend specifications.
- Results:
  - Correlations between baseline and alternative specifications are substantially high, with the mean across countries exceeding 0.9 for most specifications (Tables A5 and A6).
  - Figures (A1) and (A2) show point-wise maxima and minima across specifications; baseline trends are robust to alternatives.

### Applications: demand and supply channels — overview
- Uses decomposed series to study channels with differential impacts on demand-driven vs supply-driven inflation: Phillips curve, monetary policy transmission, oil price shocks, and global supply chain pressures.

### 4.1 Demand channel — Phillips curve findings
- Model estimated: hybrid Phillips curve
  π_{j,c,t} = β1 ŷ_{c,t} + β2 π^E_{c,t} + β3 π^m_{c,t} + Σ_{k=1}^4 γ_k π_{j,c,t-k} + ε_{c,t},
  where j ∈ {demand, supply, aggregate}, ŷ_{c,t} is the output gap (HP filter), π^E one-year ahead inflation expectations, and π^m import prices; controls include country and time fixed effects and four lags of dependent variable.
- Main empirical findings:
  - Demand-driven inflation: positive and statistically significant relationship with output gap; coefficient 0.0536 across 28 countries.
  - Supply-driven inflation: Phillips curve coefficient turns negative when estimating using the supply-driven component (no explicit numeric value reported in text for the negative coefficient).
  - Aggregate inflation: weaker (0.0342) and less significant relationship with output gap compared to demand-driven inflation.
- Interpretation:
  - Supply shocks bias the empirical Phillips curve downward because they move prices and output in opposite directions.
  - Removing supply-side disturbances from aggregate inflation reveals a stronger and significant Phillips curve.
  - The empirical Phillips curve relationship weakens during periods with pronounced supply-side shocks (e.g., post-pandemic).
- Further results:
  - Heterogeneity: interacting output gap with AE and EM dummies (Table A8) suggests the Phillips curve is stronger and more significant when supply-side inflation is removed, in both advanced economies (AEs) and emerging markets (EMs).
  - Time variation: 60-quarter rolling window estimates (Figure A9c) show weakening Phillips curve across countries over time; Figure A9a indicates the weakening in aggregate Phillips curve is driven by declining sensitivity of demand-driven inflation to the output gap recently.
  - Robustness: findings hold with additional controls for trade-partner cost pressures (Table 1 columns 4–6) and when using Hamilton (2018) filter for output gap (Table A7).

### 4.2 Demand Channel 2: Monetary Policy Transmission — objective and methodology
- Objective: Test whether monetary policy transmission is stronger for demand-driven inflation than for supply-driven inflation.
- Identification of shocks: Externally identified monetary policy shock series from Deb et al. (2023).
- Sample: 22 countries, period 1990Q1 to 2019Q4.
- Empirical method: Jordà (2005) local projections estimating responses for an eight quarters horizon using the specification
  - π_{j,c,t+h} = β_h Shock_MP_{c,t-1} + γ Z_{c,t} + α_c + α_t + ε_{c,t}
  - where π_j denotes demand-driven or supply-driven inflation (j ∈ {demand, supply}), Shock_MP_{c,t-1} are Deb et al. (2023) shocks, α_c are country fixed effects, α_t are time fixed effects, and Z_{c,t} includes 4 quarter lags of the dependent variable and the monetary policy shocks to address autocorrelation.
- Inference: Standard errors clustered in countries; confidence intervals at 90 percent reported in figures.

### 4.2 Main empirical findings — monetary policy effects
- Monetary policy tightening (100 bps shock) effects:
  - Demand-driven inflation:
    - Declines gradually over two years following a tightening shock.
    - Effects are significant across each horizon and highly persistent.
  - Supply-driven inflation:
    - Monetary policy shocks have no significant impact on supply-driven inflation series.
- Interpretation:
  - Monetary policy transmission on inflation is stronger when price changes are driven by demand-side factors, consistent with macroeconomic theory.
  - Supply-side factors can weaken monetary policy transmission, particularly when supply shocks dominate demand shocks.
- Robustness:
  - Results are robust to including controls for:
    - (i) change in nominal effective exchange rate to control for exchange rate pass-through (Figure A8);
    - (ii) change in cyclically adjusted primary balance to control for fiscal policy change (Figure A5);
    - (iii-iv) output gap and real GDP growth to control for other demand relevant factors (Figures A6 and A7).

### Key statistics and reported coefficients (from Table 1 excerpts and related tables)
- Output gap coefficients on inflation (selected entries reported):
  - ȳ̂_{i,t}: 0.0536 ∗∗∗ (0.0173); -0.0348 ∗ (0.0189); 0.0342 ∗ (0.0185); 0.0602 ∗∗∗ (0.0191); -0.0367 ∗ (0.0205); 0.0398 ∗ (0.0197)
- Inflation expectations:
  - π_E_{i,t}: 0.109 ∗∗∗ (0.00658); 0.280 ∗∗∗ (0.0743); 0.430 ∗∗∗ (0.0656); 0.110 ∗∗∗ (0.00672); 0.282 ∗∗∗ (0.0766); 0.433 ∗∗∗ (0.0662)
- Import prices:
  - π_m_{i,t}: 0.0863 (0.237); 1.012 ∗∗ (0.428); 0.959 ∗∗ (0.391); 0.148 (0.241); 0.989 ∗∗ (0.441); 1.001 ∗∗ (0.416)
- Lag coefficients (selected):
  - π_d_{i,t-1}: 0.831 ∗∗∗ (0.0676); 0.826 ∗∗∗ (0.0679)
  - π_s_{i,t-1}: 0.618 ∗∗∗ (0.0965); 0.617 ∗∗∗ (0.0984)
  - π_agg_{i,t-1}: 0.653 ∗∗∗ (0.0930); 0.649 ∗∗∗ (0.0924)
- Δppi_w_{i,t}: 0.00143 (0.00450); 0.00666 (0.00673); 0.0103 (0.00798)
- Sample and model fit (as reported):
  - Number of Observations: 2302 2302 2302 2174 2174 2174
  - No of Country: 28 28 28 27 27 27
  - R^2: 0.862, 0.894, 0.942, 0.860, 0.895, 0.942
- Significance notation:
  - ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01

### Implications
- Policy implication: Monetary policy is more effective at dampening inflation when inflation is driven by demand-side factors; when supply shocks dominate, monetary policy transmission to inflation is weaker, complicating stabilization.
- Research implication: The inflation decomposition provides a tool to distinguish demand- vs supply-driven inflation for policy design and for future research into inflation dynamics.

*Source: wpiea2023205-print-pdf*

### 1.1 Literature Review

### 1.1 Literature Review

### Literature and contribution
- Builds on Shapiro (2022) novel methodology for decomposing inflation into demand and supply components using monthly PCE data (1990 to 2023); this methodology is adopted in the study.
- Extends Sheremirov (2022) which categorized inflation into transitory and persistent demand-driven and supply-driven components for the U.S.
- Expands limited cross-country implementations (Gonçalves and Koester (2022) for aggregate Euro Area; OECD (2022) for eight advanced economies) by applying the approach to 32 advanced and emerging market economies.
- First systematic examination, to the authors’ knowledge, of how the slope of the Phillips curve varies when inflation is demand-driven vs supply-driven across 32 countries (building on U.S.-focused evidence from Bergholt, Furlanetto, and Vaccaro-Grange (2023)).
- Demonstrates that monetary policy strongly and significantly affects demand-driven inflation while supply-driven inflation is less responsive to monetary policy shocks.
- Shows supply chain pressures (using GSCPI from Benigno et al. (2022)) contributed to inflation across 32 countries, increasing supply-driven inflation with no significant impact on demand-driven inflation.

### Data (summary of sources and scope)
- Sectoral PCE data at quarterly frequency from Haver Analytics and Eurostat; monthly PCE series available for the U.S. but quarterly series used for cross-country consistency.
- Sample span: from the 1990s to 2023Q2.
- Coverage: 32 countries (23 advanced economies and 9 emerging market economies).
- Sectoral series are seasonally adjusted (either by the source or by the authors) using X13-ARIMA-SEATS; allows for additive outliers and level shifts.
- Note on item coverage: BEA provides 99 price and quantity items for the U.S.; several European countries have only 4 sectors. For countries with few sectors, two items may account for over 80% of total expenditure.
- Validation data:
  - Phillips curve exercise: quarterly output gap (HP filter), one-year ahead inflation expectations from Consensus Forecasts (CF), import prices from HAVER.
  - Monetary policy transmission: quarterly monetary policy shock series from Deb et al. (2023).
  - Oil shocks: oil supply shocks from Baumeister and Hamilton (2019).
  - Supply chain pressures: GSCPI from Benigno et al. (2022).

### Inflation decomposition methodology
- Follows Shapiro (2022) and Sheremirov (2022): classify inflation in each expenditure item as demand-driven or supply-driven based on sign co-movement of price and quantity shocks.
- Estimation:
  - For each country c and expenditure item i, estimate a VAR in first differences with p = 4 lags:
    y_cit = sum_{h=1}^p C_cij y_ci,t-h + ν_cit,
    where y_cit = (Δp_cit, Δq_cit) and ν_cit = (ν^p_cit, ν^q_cit).
  - Baseline classification: item is demand-driven if ν^p_cit * ν^q_cit ≥ 0; supply-driven if ν^p_cit * ν^q_cit < 0.
  - Aggregate demand-driven and supply-driven inflation series computed as weighted sums of item-level inflation rates using year-to-date expenditure share weights ω_cit.
  - Aggregate inflation π_ct = π^d_ct + π^s_ct.
- Caveats:
  - Classification is noisy: sectors may be simultaneously affected by demand and supply shocks; the observable reduced-form sign pattern reflects the more prevalent shock.
  - Limited number of expenditure items in some countries can make decomposition sensitive to classification of a small number of sectors.

### Validation: narrative historical evidence
- Pandemic era (since 2019Q1):
  - At 2020Q1 lockdown outset, downward pressure on prices primarily via the demand channel; demand-driven factors had a negative contribution throughout 2020.
  - As reopening and supportive fiscal/monetary policies occurred, demand-driven inflation surged through end-2022.
  - Supply-driven inflation remained positive and relatively stable through the pandemic; its relative contribution was small during initial 2021 pick-up and began rising after 2022Q1, potentially linked to commodity price pressures after Russia’s invasion of Ukraine in early 2022.
  - U.S. and Canada: PCE inflation peaked in early 2022; subsequent slowdown was due to both supply-driven and demand-driven components.
  - Asian countries (except Japan): inflation decline began towards end-2022 with decreasing contributions from both demand and supply factors.
- Great Financial Crisis (GFC):
  - Supply-driven inflation trajectory aligns with commodity price cycle peaking mid-2008 then sharply declining.
  - Decline in demand-driven inflation beginning towards end-2008 consistent with buffer-stock saving behavior under income uncertainty (Carroll (1997)) and empirical evidence that uncertainty during the GFC increased saving rates and reduced consumption and GDP growth (Mody, Ohnsorge, and Sandri (2012)).
- Commodity price drop episode (2014–2016):
  - Crude oil prices fell to under $30 in February 2016 from around $106 in June 2014.
  - Similar declines in metals and agricultural products; decomposition shows a significant decline in supply-driven inflation across regions.

### Robustness checks
- Alternative specifications considered:
  - (i) 8 lags in VAR;
  - (ii) Lag length chosen by AIC;
  - (iii) Estimation sample ending in 2019Q4;
  - (iv) 10-year rolling window estimation;
  - (v) One-step-ahead forecast errors;
  - (vi) Classifying sectors with either price or quantity residuals smaller than 0.1 standard deviations as ambiguous;
  - (vii) Classifying the smallest 10% of price and quantity residuals as ambiguous;
  - (viii) Estimation in levels with linear deterministic trend;
  - (ix) Detrending level series with Hamilton regression filter (Hamilton2018);
  - (x) Detrending level series with Hodrick-Prescott filter with smoothing parameter 1600.
- Purpose: sensitivity to lag order, pandemic outliers, parameter instability, contemporaneous information only, concerns about items with small residuals, alternative trend specifications.
- Results:
  - Correlations between baseline and alternative specifications are substantially high, with the mean across countries exceeding 0.9 for most specifications (Tables A5 and A6).
  - Figures (A1) and (A2) show point-wise maxima and minima across specifications; baseline trends are robust to alternatives.

### Applications: demand and supply channels
- Uses decomposed series to study channels with differential impacts on demand-driven vs supply-driven inflation: Phillips curve, monetary policy transmission, oil price shocks, and global supply chain pressures.

4.1 Demand channel — Phillips curve findings
- The empirical approach uses a hybrid Phillips curve:
  π_j,c,t = β1 ŷ_c,t + β2 π^E_c,t + β3 π^m_c,t + Σ_{k=1}^4 γ_k π_{j,c,t-k} + ε_c,t,
  where j ∈ {demand, supply, aggregate}, ŷ_c,t is the output gap (HP filter), π^E one-year ahead inflation expectations, and π^m import prices; controls include country and time fixed effects and four lags of dependent variable.
- Estimation results (Table 1):
  - Demand-driven inflation: positive and statistically significant relationship with output gap; coefficient 0.0536 across 28 countries.
  - Supply-driven inflation: Phillips curve coefficient turns negative when estimating using the supply-driven component (no explicit numeric value reported in text for the negative coefficient).
  - Aggregate inflation: weaker (0.0342) and less significant relationship with output gap compared to demand-driven inflation.
- Interpretation:
  - Supply shocks bias the empirical Phillips curve downward because they move prices and output in opposite directions.
  - Removing supply-side disturbances from aggregate inflation reveals a stronger and significant Phillips curve.
  - The empirical Phillips curve relationship weakens during periods with pronounced supply-side shocks (e.g., post-pandemic).
- Further results:
  - Heterogeneity: interacting output gap with AE and EM dummies (Table A8) suggests the Phillips curve is stronger and more significant when supply-side inflation is removed, in both advanced economies (AEs) and emerging markets (EMs).
  - Time variation: 60-quarter rolling window estimates (Figure A9c) show weakening Phillips curve across countries over time; Figure A9a indicates the weakening in aggregate Phillips curve is driven by declining sensitivity of demand-driven inflation to the output gap recently.
  - Robustness: findings hold with additional controls for trade-partner cost pressures (Table 1 columns 4–6) and when using Hamilton (2018) filter for output gap (Table A7).

*Source: wpiea2023205-print-pdf - 1.1 Literature Review*

### 4.2 Demand Channel 2: Monetary Policy Transmission

### 4.2 Demand Channel 2: Monetary Policy Transmission

### Objective and methodology
- Objective: Test whether monetary policy transmission is stronger for demand-driven inflation than for supply-driven inflation.
- Identification of shocks: Externally identified monetary policy shock series from Deb et al. (2023).
- Sample: 22 countries, period 1990Q1 to 2019Q4.
- Empirical method: Jordà (2005) local projections estimating responses for an eight quarters horizon using the specification
  - π_{j,c,t+h} = β_h Shock_MP_{c,t-1} + γ Z_{c,t} + α_c + α_t + ε_{c,t}
  - where π_j denotes demand-driven or supply-driven inflation (j ∈ {demand, supply}), Shock_MP_{c,t-1} are Deb et al. (2023) shocks, α_c are country fixed effects, α_t are time fixed effects, and Z_{c,t} includes 4 quarter lags of the dependent variable and the monetary policy shocks to address autocorrelation.
- Inference: Standard errors clustered in countries; confidence intervals at 90 percent reported in figures.

### Main empirical findings
- Monetary policy tightening (100 bps shock) effects:
  - Demand-driven inflation:
    - Declines gradually over two years following a tightening shock.
    - Effects are significant across each horizon and highly persistent.
  - Supply-driven inflation:
    - Monetary policy shocks have no significant impact on supply-driven inflation series.
- Interpretation:
  - Monetary policy transmission on inflation is stronger when price changes are driven by demand-side factors, consistent with macroeconomic theory.
  - Supply-side factors can weaken monetary policy transmission, particularly when supply shocks dominate demand shocks.
- Robustness:
  - Results are robust to including controls for:
    - (i) change in nominal effective exchange rate to control for exchange rate pass-through (Figure A8);
    - (ii) change in cyclically adjusted primary balance to control for fiscal policy change (Figure A5);
    - (iii-iv) output gap and real GDP growth to control for other demand relevant factors (Figures A6 and A7).

### Key statistics and reported coefficients (Table 1 excerpts)
- Reported coefficients and standard errors (as presented):
  - ȳ̂_{i,t}: 0.0536 ∗∗∗ (0.0173); -0.0348 ∗ (0.0189); 0.0342 ∗ (0.0185); 0.0602 ∗∗∗ (0.0191); -0.0367 ∗ (0.0205); 0.0398 ∗ (0.0197)
  - π_E_{i,t}: 0.109 ∗∗∗ (0.00658); 0.280 ∗∗∗ (0.0743); 0.430 ∗∗∗ (0.0656); 0.110 ∗∗∗ (0.00672); 0.282 ∗∗∗ (0.0766); 0.433 ∗∗∗ (0.0662)
  - π_m_{i,t}: 0.0863 (0.237); 1.012 ∗∗ (0.428); 0.959 ∗∗ (0.391); 0.148 (0.241); 0.989 ∗∗ (0.441); 1.001 ∗∗ (0.416)
  - Lag coefficients:
    - π_d_{i,t-1}: 0.831 ∗∗∗ (0.0676); 0.826 ∗∗∗ (0.0679)
    - π_s_{i,t-1}: 0.618 ∗∗∗ (0.0965); 0.617 ∗∗∗ (0.0984)
    - π_agg_{i,t-1}: 0.653 ∗∗∗ (0.0930); 0.649 ∗∗∗ (0.0924)
  - Δppi_w_{i,t}: 0.00143 (0.00450); 0.00666 (0.00673); 0.0103 (0.00798)
- Sample and model fit:
  - Number of Observations: 2302 2302 2302 2174 2174 2174 (as reported)
  - No of Country: 28 28 28 27 27 27 (as reported)
  - R^2: 0.862, 0.894, 0.942, 0.860, 0.895, 0.942 (as reported)
- Significance notation:
  - ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01

### Implications
- Policy implication: Monetary policy is more effective at dampening inflation when inflation is driven by demand-side factors; when supply shocks dominate, monetary policy transmission to inflation is weaker, complicating stabilization.
- Research implication: The inflation decomposition provides a tool to distinguish demand- vs supply-driven inflation for policy design and for future research into inflation dynamics.

*Source: 4.2 Demand Channel 2: Monetary Policy Transmission (extracted from the provided IMF PDF chapter content).*

### References

### wpiea2023205-print-pdf - References

### References (bibliography highlights)
- Auer, Raphael, Claudio EV Borio, and Andrew J Filardo. “The globalisation of inflation: the growing importance of global value chains”. In: (2017).
- Ball, Laurence M, Daniel Leigh, and Prachi Mishra. Understanding us inflation during the covid era. Tech. rep. National Bureau of Economic Research, 2022.
- Baumeister, Christiane and James D Hamilton. “Structural interpretation of vector autoregressions with incomplete identification: Revisiting the role of oil supply and demand shocks”. In: American Economic Review 109.5 (2019), pp. 1873–1910.
- Benigno, Gianluca et al. “The GSCPI: A new barometer of global supply chain pressures”. In: FRB of New York Staff Report 1017 (2022).
- Blanchard, Olivier. “The Phillips curve: back to the ’60s?” In: American Economic Review 106.5 (2016), pp. 31–34.
- Blanchard, Olivier J and Ben S Bernanke. What Caused the US Pandemic-Era Inflation? Tech. rep. National Bureau of Economic Research, 2023.
- Carrière-Swallow, Yan et al. “Shipping costs and inflation”. In: Journal of International Money and Finance 130 (2023), p. 102771.
- Christiano, Lawrence J, Martin Eichenbaum, and Charles L Evans. “Nominal rigidities and the dynamic effects of a shock to monetary policy”. In: Journal of political Economy 113.1 (2005), pp. 1–45.
- Galí, Jordi. Monetary policy, inflation, and the business cycle: an introduction to the new Keynesian framework and its applications. Princeton University Press, 2015.
- Hamilton, James D. “Why you should never use the Hodrick-Prescott filter”. In: Review of Economics and Statistics 100.5 (2018), pp. 831–843.
- Jordà, Òscar. “Estimation and inference of impulse responses by local projections”. In: American economic review 95.1 (2005), pp. 161–182.
- McLeay, Michael and Silvana Tenreyro. “Optimal inflation and the identification of the Phillips curve”. In: NBER Macroeconomics Annual 34.1 (2020), pp. 199–255.
- Romer, Christina D and David H Romer. “A new measure of monetary shocks: Derivation and implications”. In: American economic review 94.4 (2004), pp. 1055–1084.
- Smets, Frank and Raf Wouters. “An estimated dynamic stochastic general equilibrium model of the euro area”. In: Journal of the European economic association 1.5 (2003), pp. 1123–1175.
- Additional cited working papers, technical reports, PhD theses, and commentary pieces as listed in the source References section.

### Appendix — Figures (robustness and notes)
- Figure A1: GFC Inflation Decomposition - Robustness
  - Panels (a) Year-on-year demand-driven inflation and (b) Year-on-year supply-driven inflation shown for USA, Europe, Asia.
  - Notes: Solid blue lines show demand- and supply-driven inflation rates from baseline specification. Red dashed bands show pointwise maxima and minima across alternative specifications described in section 3.3.
  - Europe includes FRA, GER, ITA, DEN, SWE. Asia includes AUS, NZL, THA, PHL, KOR.
- Figure A2: Post 2019Q1 Inflation Decomposition - Robustness
  - Panels (a) demand-driven and (b) supply-driven inflation for USA, Europe, Asia over Q1-19 to Q1-23 (axes show %pts).
  - Notes: Solid blue lines show baseline; red dashed bands show pointwise maxima and minima across alternative specifications described in section 3.3.
  - Europe includes FRA, GER, ITA, DEN, SWE. Asia includes AUS, NZL, THA, PHL, KOR, IDN.
- Figure A3: Inflation Decomposition For 2015 “mini-recession”
  - Plots for USA, Europe, Asia from Q1-13 to Q1-17; Notes: For Europe and Asia median values across countries are plotted. Europe includes FRA, GER, ITA, DEN, SWE. Asia includes AUS, NZL, THA, PHL, KOR.
- Figures A4–A8: Monetary Policy Transmission on Demand- vs Supply-Driven Inflation — Robustness checks
  - A4 controls for the change in net effective exchange rate. Panels A and B present the response of demand- and supply-driven inflation against monetary policy shocks. Confidence intervals at 90 percent; standard errors clustered in countries.
  - A5 and A8 include the change in cyclically-adjusted primary balance to control for fiscal policy. Panels A and B show responses to monetary policy shocks. Confidence intervals at 90 percent; standard errors clustered in countries.
  - A6 includes the country-level output gap series. Panels A and B show responses to monetary policy shocks. Confidence intervals at 90 percent; standard errors clustered in countries.
  - A7 includes country-level real GDP growth series. Panels A and B show responses to monetary policy shocks. Confidence intervals at 90 percent; standard errors clustered in countries.
- Figure A9: Phillips Curve Slope: Rolling Windows
  - Panels (a) Demand-Driven Inflation, (b) Supply-Driven Inflation, (c) Aggregate Inflation.
  - Notes: Panels show the Phillips curve coefficients in 60-quarters rolling windows. The date on y-axis denotes the end of window. Confidence intervals are at 90 percent and standard errors are clustered in countries.

### Appendix — Tables (data, correlations, robustness, and estimation results)
- Table A1: Data Description (columns: Country, PCE Data Source, Seasonal Adjustment, Number of Sectors, Start Date, End Date)
  - Examples (exact entries preserved): Australia haverauthor 26 Q3 1985   Q1 2023; France haversource 18 Q1 1990   Q2 2023; UnitedStates haversource 99 Q1 1988   Q2 2023. Notes: Start date varies across countries due to data availability.
- Table A2: PCE CPI Correlations
  - Selected country correlations (as reported): Thailand 0.925 0.995; Philippines 0.89 0.972; Australia 0.938 0.999; NewZealand 0.916 0.997; Japan 0.897 0.999; SouthKorea 0.951 0.999; Taiwan 0.857 0.98; Indonesia 0.714 0.998; Canada 0.839 0.933; UnitedStates 0.956 0.998; UnitedKingdom 0.927 0.994; Germany 0.943 1; France 0.888 0.998; Italy 0.942 1; Sweden 0.972 0.96; Denmark 0.945 0.992; Netherlands 0.779 0.948; Finland 0.932 0.994; Mexico 0.988 0.998; SouthAfrica 0.906 0.977; Austria 0.923 1; Romania 0.937 0.981; Slovakia 0.934 0.986; Norway 0.84 0.925; Ireland 0.782 0.975; Czechia 0.979 1; Estonia 0.96 0.999; Latvia 0.887 0.997; Malta 0.74 0.984; Luxembourg 0.876 0.953; Cyprus 0.888 0.996; Hungary 0.978 0.996; Mean 0.900910.98509.
  - Notes: Correlation between annual PCE and CPI inflation. Table reports the high correlations between measures of aggregate PCE inflation and those reported by national authorities.
- Table A3: Mean Cumulative Expenditure Share of 6 Largest Expenditure Items
  - Selected country rows (exact values): Thailand 0.13 0.21 0.27 0.33 0.38 0.43; Philippines 0.34 0.47 0.58 0.69 0.77 0.82; Australia 0.18 0.29 0.37 0.43 0.49 0.54; UnitedStates 0.12 0.19 0.26 0.31 0.35 0.39; Germany 0.24 0.42 0.58 0.72 0.83 0.9; Netherlands 0.27 0.53 0.77 1 NaN NaN; Finland 0.5 0.82 0.91 1 NaN NaN; Mexico 0.45 0.8 0.86 0.91 0.95 0.98.
  - Notes: Table shows the average cumulative mean expenditure shares of the 6 largest PCE items within each country.
- Table A4: Specification tests for baseline model
  - Columns (1)-(11) show mean rejection rates of different specification tests across expenditure items in a given country. Column (12) reports the maximum eigenvalue of the companion matrix of the VAR across all expenditure items.
  - Sample entries (exact formatting preserved): Thailand 0.188 0.125 0.906 1 0.469 0.156 0.156 0.156 0.156 0.875 0.906 0.894; Philippines 0.083 0 0.833 0.833 0.25 0 0 0 0 0.917 0.667 0.831; UnitedStates 0.03 0.101 0.758 0.949 0.242 0.03 0.071 0.03 0.071 0.838 0.909 0.962.
  - Notes: Description of tests includes augmented Dickey-Fuller tests, Johansen maximum eigenvalue test, Ljung-Box test, ARCH test, Jarque-Bera test.
- Table A5: Demand-driven Inflation — Correlations between baseline and alternative specifications
  - Columns (i)–(x) list correlations of baseline demand-driven inflation with alternative robustness checks (Diff8Lags, DiffAIC, DiffNoCovid, RollingWindow, OneStepAhead, ThresholdNormal, ThresholdEmpiricalCDF, Level, DetrendHamilton, DetrendHP).
  - Selected country values (exact): Thailand 0.967 0.927 0.956 0.98 0.976 0.977 0.963 0.928 0.843 0.88; UnitedStates 0.935 0.944 0.923 0.96 0.982 0.994 0.956 0.942 0.859 0.821; Mean 0.87866 0.85053 0.88175 0.9261 0.93984 0.9705 0.95956 0.87469 0.79419 0.80234.
  - Notes: Table shows correlations between year-on-year demand-driven inflation rates from baseline specification (section 3.1) and robustness checks (section 3.3).
- Table A6: Supply-driven Inflation — Correlations between baseline and alternative specifications
  - Same robustness checks as Table A5 for supply-driven inflation.
  - Selected country values (exact): Thailand 0.969 0.929 0.957 0.956 0.942 0.992 0.975 0.922 0.853 0.865; UnitedStates 0.956 0.954 0.957 0.961 0.984 0.993 0.961 0.97 0.91 0.894; Mean 0.91147 0.89319 0.92541 0.93242 0.952 0.98003 0.97137 0.89759 0.81741 0.85013.
  - Notes: Table shows correlations between year-on-year supply-driven inflation rates from baseline specification (section 3.1) and robustness checks (section 3.3).
- Table A7: Phillips Curve Slope with Demand- vs. Supply-Driven Inflation (Hamilton Filter)
  - Regressions (columns (1)–(6)) report coefficients and standard errors (exact entries preserved): ̂y_i,t coefficient on π^d_i,t is 0.0309 ∗∗∗ (0.00461); on π^s_i,t is -0.00777 (0.00806) etc. π^E_i,t coefficient 0.121 ∗∗∗ (0.00912), 0.249 ∗∗∗ (0.0723), 0.423 ∗∗∗ (0.0554) across columns; π^m_i,t coefficients -0.0467 (0.237), 1.062 ∗∗ (0.396), 0.855 ∗∗ (0.411) etc.
  - Additional entries: π^d_i,t−1 0.851 ∗∗∗ (0.0497); π^s_i,t−1 0.680 ∗∗∗ (0.0686); π^agg_i,t−1 0.717 ∗∗∗ (0.0500). Number of Obs 2265 2265 2265 2149 2149 2149. Number of Country 28 28 28 27 27 27. R^2 values: 0.863 0.870 0.933 0.862 0.871 0.933.
  - Notes: Δppiw_i,t denotes change in trade-weighted partners’ producer price indexes. Standard errors clustered in countries. Significance notation: ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01.
- Table A8: Phillips Curve with Demand- vs. Supply-Driven Inflation: AE vs EM
  - Columns (1)–(3) present separate estimates. Selected coefficients (exact): ̂y_EM_i,t on π^d_i,t 0.101 ∗∗∗ (0.0252); on π^s_i,t -0.0594 (0.0350); ̂y_AE_i,t on π^d_i,t 0.0424 ∗∗∗ (0.0146); on π^s_i,t -0.0270 ∗ (0.0148); π^E_i,t 0.113 ∗∗∗ (0.00753), 0.281 ∗∗∗ (0.0762), 0.433 ∗∗∗ (0.0654); π^m_i,t 0.168 (0.244), 0.980 ∗∗ (0.436), 1.013 ∗∗ (0.420). Number of Obs 2174 2174 2174. Number of Country 27 27 27. R^2 0.862 0.895 0.942.
  - Notes: Standard errors clustered in countries. Significance notation: ∗ p<0.10, ∗∗ p<0.05, ∗∗∗ p<0.01.

*Content and data reproduced exactly as presented in the source document.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023205-print-pdf.pdf_