## wpiea2022168-print-pdf

## Source details

**Canonical URL:** [wpiea2022168-print-pdf](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022168-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022168-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022168-print-pdf.pdf.json)

---

### Executive Summary — Background and purpose
- Following the COVID-19 shock, price pressures intensified as demand recovered, supply chain distortions persisted, and commodity prices surged. Headline inflation spiked and core inflation began to rise.
- In the Middle East and Central Asia (ME&CA) region, inflation rates surged since mid-2020, driven mostly by external factors, particularly international food prices. The war in Ukraine triggered a further increase in commodity prices and higher inflation.
- Central banks need to assess upside risks to their baseline inflation forecasts; many emerging market central banks focus only on central inflation projections rather than the full distribution of risks.
- Key questions addressed:
  - What are the main drivers of core inflation in ME&CA countries?
  - Has the distribution of core inflation outcomes varied across time in the region?
  - How can central banks better deal with inflation risks, especially upside ones, when making monetary policy decisions and communicating risks to baseline forecasts?
  - Given recent price hikes, what are the main risks for inflation outcomes in the ME&CA region?

### Methodology
- Mean forecasts:
  - An augmented Phillips Curve is estimated via panel OLS to predict future core inflation at two horizons: Two-Quarter-Ahead (2Q-ahead) and Four-Quarter-Ahead (4Q-ahead).
  - Regressors included: current core inflation, output gap, inflation expectations, commodity prices, exchange rate, and a measure of underlying trend inflation (a less volatile gauge than CPI).
- Distributional forecasts:
  - A semi-parametric density estimation strategy used in the IMF Growth-at-Risk (GaR) model projects the distribution of future core inflation.
  - Quantile regressions estimate the impact of macro-financial regressors on different quantiles of 2Q-ahead and 4Q-ahead core inflation.
  - Focus on the right tail to capture upside risks (high inflation).
- Quantile regressions estimated for q ε {0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9}.
- Semi-parametric density constructed by fitting a t-skew distribution to estimated conditional quantiles via minimization of Euclidean distance; Basin-hopping algorithm used; quantile uncrossing applied when necessary.
- Projection dates for distributions: 2014 Q4, 2019 Q4, and latest datapoint (2021 Q3 for most countries).

### Key empirical findings — mean forecasts and drivers
- Current inflation is the main explanatory variable of future 2Q-ahead core inflation across the sample, indicating significant inertia in price setting.
- Inflation expectations have a stronger effect on 4Q-ahead core inflation than on 2Q-ahead.
- Commodity prices are important determinants of future inflation.
- Exchange rate effects are more muted, partly because some countries in the sample have a fixed exchange rate.
- The domestic output gap has limited explanatory power for future core inflation, pointing to a limited short-term trade-off between employment and inflation in the region.

### Benchmark panel OLS results — selected coefficients and statistics
- Panel OLS coefficients (standard errors in parentheses; significance: *** p<0.01, ** p<0.05, * p<0.1):
  - Core Inflation:
    - 2Q-ahead Core Inflation coefficient = 0.705*** (0.0391)
    - 4Q-ahead Core Inflation coefficient = 0.276*** (0.0518)
  - Inflation Expectations:
    - 2Q-ahead = 0.157** (0.0715)
    - 4Q-ahead = 0.337*** (0.0939)
  - Output Gap:
    - 2Q-ahead = −0.0364** (0.0175)
    - 4Q-ahead = 0.00103 (0.0236)
  - Exchange Rate:
    - 2Q-ahead = 0.000138*** (5.33e-05)
    - 4Q-ahead = 0.000412*** (7.10e-05)
  - Commodities:
    - 2Q-ahead = 0.303* (0.180)
    - 4Q-ahead = −0.0690 (0.257)
  - Constant:
    - 2Q-ahead = 1.057*** (0.340)
    - 4Q-ahead = 1.588*** (0.463)
  - Number of Observations: 2Q-ahead = 563; 4Q-ahead = 543
  - R-squared: 2Q-ahead = 0.530; 4Q-ahead = 0.209
  - Number of countries = 12
- Interpretation:
  - Current core inflation has the largest and most significant coefficient for 2Q-ahead, indicating near-term inertia.
  - Inflation expectations are more influential at 4Q-ahead.
  - Exchange rate significant at the 1 percent level at both horizons, larger for 4Q-ahead.
  - Output gap and commodity prices generally show lower statistical significance.

### Key empirical findings — quantile (distributional) results and nonlinearities
- Explanatory power of the model is typically maximized on the right-tail of the inflation distribution: high inflation outcomes are driven by a narrower set of strong drivers.
- For 2Q-ahead inflation:
  - Current core inflation has the strongest impact when inflation is high, indicating persistence and risk of self-reinforcing inflation dynamics.
  - Exchange rate and commodity prices have smaller average effects but relatively larger impacts at the right tail compared to central and lower quantiles.
- For 4Q-ahead inflation:
  - Inflation expectations play a larger role in explaining outcomes.
  - Transmission of commodity prices and exchange rate depreciation is more pronounced than in the short run.
- Quantile regression patterns:
  - Variables have more explanatory power in the right tail except in five countries: Armenia, Iran, Jordan, Morocco, and Pakistan.
  - Current core inflation: strong impact at center and right tail for 2Q-ahead; weaker across quantiles for 4Q-ahead.
  - Inflation expectations: more impact on the right tail for most countries; for 4Q-ahead, larger effects in middle quantiles (20 to 70).
  - Commodity prices: affect right tail in both OEs and OIs; especially large in Iran and UAE; pass-through stronger when inflation is high.
  - Exchange rate: larger effects on right tail in Algeria, Egypt, Jordan, and Pakistan for 2Q-ahead; stronger long-run (4Q-ahead) effects.
  - Output gap: no impact across quantiles for both horizons.

### Evolution of future inflation risks over time
- Parametric distributions fitted to estimated inflation quantiles at end-2014, end-2019, and end-2021.
- For 4Q-ahead inflation:
  - From 2014–2019, core inflation distributions shifted left and became centered around lower inflation levels in most ME&CA countries.
  - By end-2021, the distribution of future core inflation outcomes moved to the right in most countries; the surge in commodity prices had not yet fully passed through to core inflation at end-2021 in most countries, but future core inflation outcomes became more volatile and more skewed to the right.
- For 2Q-ahead since end-2019:
  - Four countries experienced a leftward shift: two OEs (Algeria and UAE) and two OIs (Pakistan and Tunisia).
  - Three OIs experienced a rightward shift: Armenia, Egypt, and Jordan.
  - Morocco and Oman maintained distributions centered around low inflation throughout the sample.
- Higher-order moments across time:
  - Skewness: Distributions became positively skewed in most countries for both horizons.
  - Kurtosis: Increased over time, especially among OEs, indicating fatter tails.
  - Variance: Dispersion widened in most OEs and OIs, especially between 2014 and 2019; variance increased more for 2Q-ahead than for 4Q-ahead over the recent period.
- Overall: Future core inflation outcomes have become more volatile and tilted to the right tail throughout the period, especially in OEs.

### Policy implications and recommendations
- The model provides a forward-looking Value-at-Risk (VaR) approach to assess inflation upside risks and potential de-anchoring of inflation expectations.
- Central banks could leverage this VaR / Inflation-at-Risk framework to:
  - Better communicate upside and downside risks to the inflation outlook and baseline forecasts.
  - Incorporate distributional information (not just point forecasts) into monetary policy decisions to mitigate inflation risks and prevent self-reinforcing inflation dynamics.
- Specific recommendations for central banks in ME&CA:
  - Give greater attention to inflation risks, especially upside risks, rather than relying solely on point forecasts.
  - Incorporate nonlinear effects into forecasting processes to reduce likelihood of persistent deviations from target by acting early on upside risks.
  - Use the VaR approach to decide on inflation risk tolerance and to guide policy responses (for example, reducing tail risk by increasing the policy rate when significant upside risk is projected).
    - Central banks could set thresholds as a value-at-risk of conditional inflation distributions (e.g., 90 percent, 95 percent).
    - In some cases, central banks might choose to "overshoot" policy rates to firmly anchor expectations.
  - Improve communication of risks:
    - Communicate full distributions or risk balances, not only point forecasts.
    - Reveal whether the balance of risks is tilted to the downside or upside and identify main factors driving the baseline using the full conditional distribution.
    - Examples: Morocco’s central bank (BAM) publishes fan charts showing probability bands for CPI (and real GDP) over 8 quarters; Pakistan’s Monetary Policy Committee provides an inflation range with a "2 percentage points amplitude".

### Appendix — Selected country-level OLS and quantile highlights
- OLS coefficients for four-quarter-ahead core inflation (country order listed): Algeria, Armenia, Bahrain, Egypt, Iran, Jordan, Kazakhstan, Lebanon, Morocco, Oman, Pakistan, Tunisia, UAE:
  - Core Inflation coefficients: -0.0922579, -0.0994558, 0.4845504***, 0.4195608***, -0.009004, -0.4040957**, 0.1745353, -0.3891954, -0.2948147*, 0.9016229***, -0.0346878, 0.4983559***, -0.4592576*
  - Inflation Expectations coefficients: -0.3246123***, -0.0935936, 0.3322314***, 0.1202242, 0.9124813***, 0.3558176***, 0.4445018***, -0.0972418, -0.0539157, -0.5619749***, 0.0601302, 0.1316716, 0.3272545*
  - Output Gap coefficients: 0.0002228, -0.0050338***, -0.0019300***, -0.0086566**, -0.0180458***, 0.0025377, 0.0004665, -0.0025730***, -0.0000332*, -0.0005709*, -0.000542
  - Exchange Rate coefficients: 1.0049073***, 0.3271818, 0, 0.2727087***, -0.3041564, 0.1328773***, -0.4653001***, -0.3184770**, 0.3877597***, 0.2252192***
  - Commodities coefficients: 0.5154859**, 0.6636534***, 0.8250012***, -0.2015048*, 1.3518269***, 0.1827964, -0.6409641***, -0.0881649, -0.1353215, 0.0037902, 0.6388697***, 0.5100771***, 1.5477106*
  - Constant terms: 0.0375197, 0.050977, 0.0444004, -0.2470369**, 0.1369073, -0.1026065, -0.1208117*, -0.3227842*, -0.0339935, 0.3056562**, 0.1038705, 0.2268401, 0.6642141
  - Number of Observations by country: 39, 40, 39, 48, 35, 56, 68, 31, 44, 57, 46, 40, 30
  - R-Squared by country: 0.70, 0.26, 0.40, 0.35, 0.77, 0.17, 0.33, 0.03, 0.30, 0.60, 0.32, 0.53, 0.16
  - BIC by country: 88.93, 112.58, 105.66, 141.66, 71.35, 175.05, 189.97, 88.18, 134.34, 129.93, 138.12, 95.28, 108.40
  - Significance notation in Appendix I: * p<0.2  ** p<0.1  *** p<0.05. HAC Standard Errors used.
- Quantile regression selected coefficients (examples across quantiles 10–90):
  - Current Core Inflation — Country Average: 0.059, 0.134, 0.112, 0.103, 0.070, 0.058, 0.079, 0.099, 0.128
  - Exchange Rate — Country Average: 0.096, 0.080, 0.143, 0.087, 0.106, 0.100, 0.077, 0.161, 0.158
  - Commodity Prices — Country Average: 0.239, 0.201, 0.229, 0.296, 0.304, 0.301, 0.337, 0.360, 0.387
  - Selected country examples:
    - Oman (Current Core Inflation quantiles): 0.773, 0.730, 0.718, 0.701, 0.744, 0.750, 0.740, 1.043, 1.176
    - Iran (Commodity Prices quantiles): 0.6468, 0.7827, 0.949, 1.267, 1.4096, 1.5157, 1.568, 1.7077, 1.3118
    - Algeria (Exchange Rate quantiles): 1.0153, 0.7835, 0.7651, 0.8355, 0.7238, 0.8726, 0.9156, 0.9504, 1.1547
- Notes:
  - Quantile coefficients for standardized regressors (except output gap) are interpretable as SD-to-SD effects on 2Q-ahead core inflation.
  - Industrial production is used instead of GDP for Jordan and Pakistan in some regressions.
  - For exchange rate coefficients, an increase implies depreciation; Bahrain, Oman, and UAE excluded from exchange rate panels due to fixed exchange rates.

### Data sources (variables and definitions)
- Core Inflation: Z-score of year over year change in core inflation — IMF MCD Regional Studies Division
- Output gap: Percent difference between nominal and potential GDP growth. Potential GDP growth calculated using a Hodrick-Prescott Filter — CEIC Data
- Industrial Production: Substitute for output gap. Percent difference between nominal and potential industrial production. Potential calculated using a Hodrick-Prescott Filter — Haver Analytics
- Nominal Effective Exchange Rate: Z-score of year over year change in nominal effective exchange rate — IMF International Financial Statistics
- Commodity Price Index: Z-score of year over year change in Goldman Sachs Global Commodity Price Index — Bloomberg
- Inflation Expectations: Z-score of 5-year ahead forecast of IMF CPI forecast — IMF World Economic Outlook

*Source: Executive Summary, IMF WORKING PAPERS — Inflation-at-Risk in the Middle East and Central Asia (wpiea2022168-print-pdf).*

### Executive Summary ......................................................................................................

### Executive Summary

### Background and purpose
- Following the COVID-19 shock, price pressures intensified as demand recovered, supply chain distortions persisted, and commodity prices surged. Headline inflation spiked and core inflation began to rise.
- In the Middle East and Central Asia (ME&CA) region, inflation rates surged since mid-2020, driven mostly by external factors, particularly international food prices. The war in Ukraine triggered a further increase in commodity prices and higher inflation.
- Central banks need to assess upside risks to their baseline inflation forecasts; many emerging market central banks focus only on central inflation projections rather than the full distribution of risks.
- Key questions addressed:
  - What are the main drivers of core inflation in ME&CA countries?
  - Has the distribution of core inflation outcomes varied across time in the region?
  - How can central banks better deal with inflation risks, especially upside ones, when making monetary policy decisions and communicating risks to baseline forecasts?
  - Given recent price hikes, what are the main risks for inflation outcomes in the ME&CA region?

### Methodology
- Mean forecasts: An augmented Phillips Curve is estimated via panel OLS to predict future core inflation at two horizons: Two-Quarter-Ahead (2Q-ahead) and Four-Quarter-Ahead (4Q-ahead).
  - Regressors included: current core inflation, output gap, inflation expectations, commodity prices, exchange rate, and a measure of underlying trend inflation (a less volatile gauge than CPI).
- Distributional forecasts: A semi-parametric density estimation strategy used in the IMF Growth-at-Risk (GaR) model (Prasad and al., 2019, and Lafarguette, 2019) projects the distribution of future core inflation.
  - Quantile regressions estimate the impact of macro-financial regressors on different quantiles of 2Q-ahead and 4Q-ahead core inflation.
  - Focus on the right tail to capture upside risks (high inflation).

### Key empirical findings — mean forecasts and drivers
- Current inflation is the main explanatory variable of future 2Q-ahead core inflation across the sample, indicating significant inertia in price setting.
- Inflation expectations have a stronger effect on 4Q-ahead core inflation than on 2Q-ahead.
- Commodity prices are important determinants of future inflation.
- Exchange rate effects are more muted, partly because some countries in the sample have a fixed exchange rate.
- The domestic output gap has limited explanatory power for future core inflation, pointing to a limited short-term trade-off between employment and inflation in the region.

### Key empirical findings — quantile (distributional) results and nonlinearities
- Explanatory power of the model is typically maximized on the right-tail of the inflation distribution: high inflation outcomes are driven by a narrower set of strong drivers.
- For 2Q-ahead inflation:
  - Current core inflation has the strongest impact when inflation is high, indicating a high degree of persistence and the risk of self-reinforcing inflation dynamics.
  - Exchange rate and commodity prices have smaller effects than current core inflation on average, but their relative impact is magnified at the right tail compared to central and lower quantiles.
- For 4Q-ahead inflation:
  - Inflation expectations play a larger role in explaining outcomes.
  - Transmission of commodity prices and exchange rate depreciation is more pronounced than in the short run.

### Evolution of future inflation risks over time
- Parametric distributions were fitted to estimated inflation quantiles at three points in time: end-2014 (middle of sample), end-2019 (pre-COVID-19), and end-2021.
- For 4Q-ahead inflation:
  - From 2014–2019, core inflation distributions shifted left and became centered around lower inflation levels in most ME&CA countries.
  - By end-2021, the distribution of future core inflation outcomes moved to the right in most countries. The surge in commodity prices had not yet fully passed through to core inflation at end-2021 in most countries, but future core inflation outcomes became more volatile and more skewed to the right.

### Policy implications and recommendations
- The model captures nonlinearities and provides a forward-looking Value-at-Risk (VaR) approach to assess inflation upside risks and the potential de-anchoring of inflation expectations.
- Central banks could leverage this VaR / Inflation-at-Risk framework to:
  - Better communicate upside and downside risks to the inflation outlook and baseline forecasts.
  - Incorporate distributional information (not just point forecasts) into monetary policy decisions to mitigate inflation risks and prevent self-reinforcing inflation dynamics.

*Source: Executive Summary, IMF WORKING PAPERS — Inflation-at-Risk in the Middle East and Central Asia.*

### Section IIII presents the results. And Section IV discusses policy implications and concludes.

### wpiea2022168-print-pdf - Section IIII presents the results. And Section IV discusses policy implications and concludes.

### Literature Review
- Adrian and al. (2019) transpose Value-at-Risk to macroeconomic forecasting to project the conditional distribution of future real GDP growth using a two-step semi-parametric approach: (i) quantile regressions of future GDP growth on current macro-financial indicators; (ii) fit a continuous t-skew distribution on the conditional quantiles.
- Lopez-Alido and al. (2020) apply quantile regressions to an augmented Phillips Curve for U.S. inflation and find that tight financial conditions carry substantial downside inflation risks; the conditional mean masks ample variability in the predictive distribution.
- Banerjee and al. (2020) use quantile panel regressions with fixed effects to estimate conditional quantiles of four-quarter-ahead headline inflation for a panel of advanced and EMEs; they fit a continuous skewed t-distribution to the estimated quantiles and find:
  - Upside inflation risks have generally declined over time.
  - Significant non-linearities in EMEs: large exchange rate depreciations associated with upside inflation risks; tightening financial conditions increase both up- and downside inflation risks.
- Evans and others (2015) document that the Federal Reserve accounts for uncertainty in policy setting and recommend a “risk management approach” where policy considers the dispersion of shocks around their means.

### Empirical Strategy
- Sample:
  - 12 EMEs in the ME&CA: six oil importers (OIs) — Armenia, Egypt, Jordan, Morocco, Pakistan, and Tunisia; six oil exporters (OEs) — Algeria, Bahrain, Iran, Kazakhstan, Oman, and UAE.
  - Data is quarterly and covers on average 45 quarters. Shorter periods are used for Armenia, Bahrain, Iran, Tunisia, and UAE due to data limitations.
- Target variable:
  - Primary focus on 2Q-ahead core inflation; 4Q-ahead also estimated with robustness checks in appendix.
  - Core inflation constructed by excluding food (food and non-alcoholic beverages) and energy (housing, water, electricity, gas and other fuels) from CPI, following IMF (2022) approach.
- Observed dynamics:
  - Egypt and Iran experienced larger upside inflation shocks driven by large currency depreciations in 2016 and 2018, respectively.
  - Algeria experienced a nominal depreciation in 2015 with a smaller inflation impact than Egypt and Iran.
  - Most countries maintained core inflation around five percent through the 2010s; Armenia and Morocco experienced near zero core inflation for several quarters; Bahrain, Oman, and UAE experienced negative core inflation during several consecutive quarters.
- Baseline econometric specification (augmented open-economy Phillips curve with fixed effects):
  - π_{i,t+2}^* = α + β_γ γ_{i,t} + β_π π_{i,t}^* + β_{π^e} π_{i,t}^{LT E} + β_n Δneer_{i,t} + β_p ΔP_{i,t}^* + FE
    - Where Y (or γ) is the output gap (or industrial production), π is current core prices, neer is the nominal effective exchange rate, P is the Goldman Sachs Commodity Price Index, and π^{LTE} is five-year ahead CPI inflation forecast from IMF’s WEO.
  - All variables (dependent and regressors except output gap) are normalized using z-score.
- Quantile regression setup:
  - Quantiles estimated for q ε {0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9}.
  - Quantile specification:
    - π_{i,t+2}^{*q} = α^q + β_γ^q γ_{i,t} + β_π^q π_{i,t}^* + β_{π^e}^q π_{i,t}^{LT E} + β_n^q Δneer_{i,t} + β_p^q ΔP_{i,t}^* + ε_{t+2}^q
  - Quantile regressions capture different slopes across points of the future core inflation distribution; country-by-country regressions use HAC standard errors and start with a confidence interval of 20 percent, noting significance at 10 percent and 5 percent.
- Semi-parametric density construction:
  - Fit continuous future core inflation distribution by minimizing Euclidean distance between empirical conditional quantiles and theoretical quantiles of a t-skew distribution (location, scale, skew parameters), using Basin-hopping algorithm in Python scipy.
  - Quantile uncrossing applied when necessary using Chernozhukov et al. (2014).
  - Fit criterion:
    - loc,scale,skew = argmin[ Σ{ tsk.quantile(q,loc,scale,skew) − q Q̂[π_{i,t+2}^{*q}] }^2 ]
- Projection dates for future core inflation distributions: (i) 2014 Q4, (ii) 2019 Q4, (iii) latest datapoint (2021 Q3 for most countries).

### Empirical Results — Benchmark Regressions and Key Findings
- Panel OLS (fixed effects) shows stronger explanatory power for 2Q-ahead than for 4Q-ahead core inflation.
  - Possible reason: strong price control mechanisms, administered food prices, and energy subsidies limit short-term pass-through (IMF, 2022).
- Panel OLS coefficients (Table 1) — exact estimates and standard errors:
  - Core Inflation:
    - 2Q-ahead Core Inflation coefficient = 0.705*** (0.0391)
    - 4Q-ahead Core Inflation coefficient = 0.276*** (0.0518)
  - Inflation Expectations:
    - 2Q-ahead = 0.157** (0.0715)
    - 4Q-ahead = 0.337*** (0.0939)
  - Output Gap:
    - 2Q-ahead = −0.0364** (0.0175)
    - 4Q-ahead = 0.00103 (0.0236)
  - Exchange Rate:
    - 2Q-ahead = 0.000138*** (5.33e-05)
    - 4Q-ahead = 0.000412*** (7.10e-05)
  - Commodities:
    - 2Q-ahead = 0.303* (0.180)
    - 4Q-ahead = −0.0690 (0.257)
  - Constant:
    - 2Q-ahead = 1.057*** (0.340)
    - 4Q-ahead = 1.588*** (0.463)
  - Number of Observations: 2Q-ahead = 563; 4Q-ahead = 543
  - R-squared: 2Q-ahead = 0.530; 4Q-ahead = 0.209
  - Number of countries = 12
  - Standard errors in parentheses; significance: *** p<0.01, ** p<0.05, * p<0.1
- Interpretation of panel results:
  - Current core inflation has the largest coefficient and is statistically significant at the 1 percent level for 2Q-ahead, indicating substantial near-term inertia in price setting.
  - Inflation expectations have lower coefficient and statistical significance at 2Q-ahead but stronger and more frequently significant at 4Q-ahead, consistent with greater forward-looking influence over longer horizons.
  - Exchange rate is statistically significant at the 1 percent level at both horizons, with a larger coefficient for 4Q-ahead core inflation.
  - Output gap and commodity prices have lower statistical significance overall.
- Country-by-country OLS (HAC standard errors) findings:
  - Core inflation generally displays the largest effect on future 2Q-ahead core inflation across most countries; one standard deviation change in current core inflation shows positive coefficients on future conditional prices and results are statistically significant at the 5 percent level for most countries.
  - Inflation expectations effects are more muted; coefficients in OEs show expected sign and are larger and more statistically significant (except Algeria and UAE) than in OIs.
  - Commodity prices are positively associated with future core inflation across most countries (not statistically significant for Morocco).
  - Excluding hard-peg countries (Bahrain, Oman, UAE), exchange rate depreciation leads to inflationary pressures in most countries with results statistically significant at the 5 percent level for most countries.
  - Output gap displays limited explanatory power and statistical significance, implying a limited short-run trade-off between employment and inflation.
- Consistency with prior IMF results:
  - Results align with IMF (2022) where output gap and domestic settings are not statistically significant determinants in Phillips curve estimates; this may reflect endogeneity issues and an empirical disconnect between inflation and the output gap (McLeay and Tenreyro, 2019).

*Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022168-print-pdf.pdf*

### Appendix 2 presents the results of the country-by-country regression for 4Q-ahead core inflation. Commodities

### Appendix 2 — Results: Country-by-Country Regressions and Quantile Analysis for 2Q- and 4Q-Ahead Core Inflation

### Quantile regression methodology and general patterns
- Quantile regressions were used to identify nonlinearities, with the impact of regressors varying across the quantiles of the distribution of the dependent variable.
- Overall, variables have more explanatory power in the right tail of the distribution – i.e for high inflation – except in five countries: Armenia, Iran, Jordan, Morocco, and Pakistan.
- Except for the output gap, variables are standardized for 2Q-ahead analysis; the coefficient equals the size of the standard deviation movement in 2Q-ahead core inflation associated with a one standard deviation movement in the regressor.

### Key empirical findings — regressors’ effects across horizons and quantiles
- Current core inflation
  - Displays the highest coefficients and significant nonlinearities across quantiles.
  - Strong impact at the center and right tail of the distribution for 2Q-ahead core inflation.
  - Four countries show larger effects when inflation is high, especially among oil exporters: Algeria, Kazakhstan, and Oman, and Egypt.
  - For 4Q-ahead core inflation, coefficients for current core inflation are weaker across quantiles relative to 2Q-ahead; the effect on the right tail is also less pronounced.
- Inflation expectations
  - Have more impact on the right tail of the distribution for most countries in the sample, especially strong in Armenia and Iran.
  - For 4Q-ahead, inflation expectations have larger effects in the middle quantiles (from 20 to 70 quantiles), consistent with longer-run inflation being more forward-looking.
- Commodity prices
  - Affect more strongly the right tail of the distribution in both oil exporters (OEs) and oil importers (OIs); effects particularly large in Iran and UAE, and to a lesser extent in Bahrain.
  - Transmission of commodity prices to core inflation takes time; pass-through to core inflation is stronger when inflation levels are high.
  - For 4Q-ahead, commodity prices display on average larger coefficients at the right tail, pointing to larger pass-through in high-inflation contexts.
- Exchange rate
  - Has marginally larger effects than the output gap for 4Q-ahead in OLS results; for 2Q-ahead, exchange rate shows marginally larger effect than the output gap.
  - More effect on the right tail of the distribution in Algeria, Egypt, Jordan, and Pakistan for 2Q-ahead; in Armenia impact is strong from the mean to the right tail.
  - For 4Q-ahead, the impact of the exchange rate appears larger than for 2Q-ahead, especially in the right tail, revealing long-lasting effects of depreciation on prices.
- Output gap (and industrial production where used)
  - Displays very small coefficients in OLS; the domestic output gap has no impact across quantiles for both 2Q-ahead and 4Q-ahead core inflation.
  - Industrial production is used instead of GDP for some countries (noted in figures).

### Distributions of future inflation (2Q-ahead and 4Q-ahead)
- Time comparisons presented for distributions in 2014, 2019, and 2021 for OEs and OIs.
- From 2014–2019, the distributions of conditional 4Q-ahead core inflation moved left in most countries.
- Since the pandemic, by end-2021 higher headline inflation had not yet fully transmitted into the core inflation measures.
  - For 2Q-ahead core inflation since end-2019:
    - Four countries experienced a leftward shift: two OEs (Algeria and UAE) and two OIs (Pakistan and Tunisia).
    - Three OIs experienced a rightward shift: Armenia, Egypt, and Jordan.
    - Morocco and Oman maintained distributions centered around low inflation throughout the sample.
  - For 4Q-ahead core inflation, more countries experienced a rightward shift, reflecting increased risks of higher inflation levels.
- Higher-order moments:
  - Skewness: Distributions have become positively skewed over time in most countries for both horizons, indicating more prominent upside risks.
  - Kurtosis: Increased across countries over time, especially among OEs, showing fatter tails and greater tail risk.
  - Variance: Dispersion widened in most OEs and OIs, especially between 2014 and 2019; over the recent period, variance increased more for 2Q-ahead than for 4Q-ahead core inflation.
- Overall: Future core inflation outcomes have become more volatile and tilted to the right tail of the distribution throughout the period, especially in OEs.

### Conclusions and policy implications (summary of recommendations)
- Four main conclusions:
  - The impact of key drivers varies over time: current core inflation matters more for 2Q-ahead; inflation expectations matter more for 4Q-ahead.
  - Determinants of future core inflation have nonlinear effects across inflation levels: current core inflation and inflation expectations matter most at the center and right tail; exchange rate depreciation and commodity prices also matter more in the right tail (but less than backward- and forward-looking inflation variables).
  - Commodity price transmission to core inflation is delayed in the region (where price controls are pervasive) and stronger when inflation is high; exchange rate effects are long-lasting, especially at the right tail.
  - Limited pass-through of higher headline inflation to core inflation has masked prominent upside risks and more volatile future inflation outcomes; point forecasts alone are insufficient to spot risks and design policy.
- Policy recommendations for central banks in ME&CA:
  - Give greater attention to inflation risks, especially upside risks, rather than relying solely on point forecasts.
  - Incorporate nonlinear effects into forecasting processes to reduce likelihood of persistent deviations from target by acting early on upside risks.
  - Use the VaR approach from the paper to decide on inflation risk tolerance and to guide policy responses (for example, reducing tail risk by increasing the policy rate when significant upside risk is projected).
    - Central banks could set thresholds as a value-at-risk of conditional inflation distributions (e.g., 90 percent, 95 percent).
    - In some cases, central banks might choose to "overshoot" policy rates to firmly anchor expectations.
  - Improve communication of risks:
    - Communicate full distributions or risk balances, not only point forecasts, to avoid giving a misleading impression and to preserve credibility.
    - Reveal whether the balance of risks is tilted to the downside or upside and identify main factors driving the baseline scenario using the full conditional distribution.
    - Examples of existing practices: Morocco’s central bank (BAM) publishes fan charts showing probability bands for CPI (and real GDP) over 8 quarters; Pakistan’s committee provides an inflation range rather than a point forecast.
  - Appropriately constructed alternative scenarios and VaR-driven communications can make monetary policy more predictable and help markets internalize central bank guidance; communication should remain clear and transparent to avoid misunderstanding.

*IMF Working Paper — Appendix 2 results and related figures as presented in the source document.*

### Appendix I.  Data Sources

### Appendix I. Data Sources

### Variable definitions and data sources
- Core Inflation: Z-score of year over year change in core inflation — IMF MCD Regional Studies Division
- Output gap: Percent difference between nominal and potential GDP growth. Potential GDP growth calculated using a Hodrick-Prescott Filter — CEIC Data
- Industrial Production: Substitute for output gap. Percent difference between nominal and potential industrial production. Potential calculated using a Hodrick-Prescott Filter — Haver Analytics
- Nominal Effective Exchange Rate: Z-score of year over year change in nominal effective exchange rate — IMF International Financial Statistics
- Commodity Price Index: Z-score of year over year change in Goldman Sachs Global Commodity Price Index — Bloomberg
- Inflation Expectations: Z-score of 5-year ahead forecast of IMF CPI forecast — IMF World Economic Outlook

### Notes
- Industrial production is used instead of GDP for Jordan and Pakistan in some regressions (noted in appendices).
- For exchange rate coefficients, an increase in the exchange rate translates to a depreciation. Bahrain, Oman, and the United Arab Emirates are excluded from exchange rate panels because they have fixed exchange rates.
- Except for the output gap, variables are standardized in the quantile analysis; coefficients equal the size of the standard deviation movement in 2Q-ahead core inflation associated with a one standard deviation movement in the variable shown.

### OLS results for four-quarter-ahead core inflation — selected coefficients and model statistics
- Core Inflation coefficients (listed in the order presented for countries: Algeria, Armenia, Bahrain, Egypt, Iran, Jordan, Kazakhstan, Lebanon, Morocco, Oman, Pakistan, Tunisia, UAE):
  - -0.0922579, -0.0994558, 0.4845504***, 0.4195608***, -0.009004, -0.4040957**, 0.1745353, -0.3891954, -0.2948147*, 0.9016229***, -0.0346878, 0.4983559***, -0.4592576*
- Inflation Expectations coefficients (same country order):
  - -0.3246123***, -0.0935936, 0.3322314***, 0.1202242, 0.9124813***, 0.3558176***, 0.4445018***, -0.0972418, -0.0539157, -0.5619749***, 0.0601302, 0.1316716, 0.3272545*
- Output Gap coefficients:
  - 0.0002228, -0.0050338***, -0.0019300***, -0.0086566**, -0.0180458***, 0.0025377, 0.0004665, -0.0025730***, -0.0000332*, -0.0005709*, -0.000542
- Exchange Rate coefficients:
  - 1.0049073***, 0.3271818, 0, 0.2727087***, -0.3041564, 0.1328773***, -0.4653001***, -0.3184770**, 0.3877597***, 0.2252192***
- Commodities coefficients:
  - 0.5154859**, 0.6636534***, 0.8250012***, -0.2015048*, 1.3518269***, 0.1827964, -0.6409641***, -0.0881649, -0.1353215, 0.0037902, 0.6388697***, 0.5100771***, 1.5477106*
- Constant terms (same country order):
  - 0.0375197, 0.050977, 0.0444004, -0.2470369**, 0.1369073, -0.1026065, -0.1208117*, -0.3227842*, -0.0339935, 0.3056562**, 0.1038705, 0.2268401, 0.6642141
- Number of observations (No. of Obs.) by country:
  - 39, 40, 39, 48, 35, 56, 68, 31, 44, 57, 46, 40, 30
- R-Squared by country:
  - 0.70, 0.26, 0.40, 0.35, 0.77, 0.17, 0.33, 0.03, 0.30, 0.60, 0.32, 0.53, 0.16
- BIC by country:
  - 88.93, 112.58, 105.66, 141.66, 71.35, 175.05, 189.97, 88.18, 134.34, 129.93, 138.12, 95.28, 108.40
- Significance notation used: * p<0.2  ** p<0.1  *** p<0.05. HAC Standard Errors used.

### Quantile regression highlights (four-quarter-ahead and two-quarter-ahead panels)
- Current Core Inflation — selected country quantile coefficients (quantiles 10,20,30,40,50,60,70,80,90):
  - Algeria: -0.489, -0.003, -0.025, -0.071, -0.021, -0.112, -0.127, -0.066, -0.062
  - Bahrain: 0.547, 0.748, 0.650, 0.513, 0.295, 0.331, 0.248, 0.307, 0.462
  - Oman: 0.773, 0.730, 0.718, 0.701, 0.744, 0.750, 0.740, 1.043, 1.176
  - United Arab Emirates: -0.402, -0.418, -0.451, -0.536, -0.591, -0.544, -0.419, -1.199, -1.246
  - Country Average (Current Core Inflation): 0.059, 0.134, 0.112, 0.103, 0.070, 0.058, 0.079, 0.099, 0.128
- Output Gap — selected country quantile coefficients (quantiles 10–90):
  - Algeria: 0.001, 0.001, 0.001, 0.000, 0.000, 0.001, 0.001, 0.001, 0.001
  - Iran: -0.009, -0.009, -0.013, -0.018, -0.018, -0.020, -0.020, -0.024, -0.015
  - United Arab Emirates: -0.005, -0.005, -0.004, -0.004, -0.005, -0.006, -0.006, -0.008, -0.009
  - Country Aveage: -0.002, -0.001, -0.002, -0.003, -0.003, -0.003, -0.003, -0.004, -0.005
  - Note: 1/ Industrial production used instead of GDP for Jordan and Pakistan in the output gap panels.
- Exchange Rate — selected country quantile coefficients (quantiles 10–90):
  - Algeria: 1.0153, 0.7835, 0.7651, 0.8355, 0.7238, 0.8726, 0.9156, 0.9504, 1.1547
  - Pakistan: 0.3021, 0.3324, 0.3783, 0.3875, 0.5933, 0.5741, 0.4397, 0.3563, 0.2732
  - Country Aveage: 0.096, 0.080, 0.143, 0.087, 0.106, 0.100, 0.077, 0.161, 0.158
- Commodity Prices — selected country quantile coefficients (quantiles 10–90):
  - Algeria: -0.181, -0.146, -0.137, 0.2679, 0.1537, 0.0832, 0.3242, 0.4384, 0.7827
  - Iran: 0.6468, 0.7827, 0.949, 1.267, 1.4096, 1.5157, 1.568, 1.7077, 1.3118
  - United Arab Emirates: 0.4994, 0.5564, 0.7622, 0.8543, 0.7465, 0.8742, 0.9639, 1.4406, 1.3237
  - Country Average: 0.239, 0.201, 0.229, 0.296, 0.304, 0.301, 0.337, 0.360, 0.387

### Methodological/interpretation points from appendices
- Heat maps (Appendix IV) display coefficients for each country at different quantiles; for standardized variables (except output gap) coefficients are interpretable as SD-to-SD effects on 2Q-ahead core inflation.
- Quantile regressions are presented separately for two-quarter-ahead and four-quarter-ahead horizons and by country (Appendices III and V).
- OLS and quantile regressions use HAC standard errors where noted.

*Source: Appendix I–V (Data Sources, OLS tables, Quantile regression coefficients, and associated notes) from the PDF content unit.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022168-print-pdf.pdf_