## wpiea2025003-print-pdf

## Source details

**Canonical URL:** [wpiea2025003-print-pdf](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025003-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025003-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025003-print-pdf.pdf.json)

---

### Introduction and regional inflation trajectory
- Inflation in most CCA countries:
  - "almost doubled to 13 percent in 2022 from around 7 percent in 2020"
  - "significantly above the global average of 8.7 percent in 2022"
  - "before dropping to 9.8 percent in 2023 and continuing to trend down in 2024"
- Upswing drivers:
  - Dislocations from the COVID-19 pandemic (supply chain disruptions, large fiscal stimulus, accommodative monetary policies, pent-up consumer demand).
  - Russia’s invasion of Ukraine affecting energy and food prices.
- Purpose: investigate inflation dynamics in the CCA, main drivers, and provide conclusions to inform policymaking; first study focusing on the CCA region to the authors' knowledge.

### Stylized facts and recent developments (Section 3)
- Historical patterns:
  - Hyperinflation in early 1990s after Soviet Union collapse; single-digit inflation returned by end of the decade in Armenia, Azerbaijan, Georgia, Kazakhstan.
  - Uzbekistan: low single-digit inflation for most recent history.
  - Acceleration before the Global Financial Crisis; oil-price slump from 2014 influenced regional inflation heterogeneously.
- Recent shocks (2020–23):
  - COVID-19 in 2020, global fiscal and monetary stimulus, post-pandemic demand recovery, and the war in Ukraine in February of 2022 drove surging energy and food prices.
  - Ruble volatility with three phases: initial depreciation, sharp appreciation, recent depreciation below pre-conflict levels.
- Principal component analysis (PCA) insights:
  - Large global component became a major contributor to the post-Covid inflation surge.
  - Tajikistan and Uzbekistan display more idiosyncratic behavior.
  - Regional component affected oil importers and exporters differently (e.g., regional component contributed to disinflation in Armenia, Georgia, Tajikistan; rose in Azerbaijan and Kazakhstan).
  - Domestic component generally more stable; in Armenia domestic factors were important though smaller than foreign factors.
- Developments since 2023:
  - Inflation pressures subsided due to dissipation of base effects, decline in commodity prices, tighter monetary policies, and strong exchange rate appreciations in several countries.
  - Notable appreciations: Armenia, Georgia, Tajikistan (linked to capital inflows from Russia).
  - Several countries loosened monetary policy as inflation moderated: Armenia, Azerbaijan, Georgia, Kazakhstan, Kyrgyz Republic, Tajikistan.
- Exchange-rate pass-through and commodity exposure:
  - Average pass-through to inflation of a depreciation versus the US dollar: "around 10 percent on impact and about 25 percent after 12 months."
  - This average pass-through is higher than estimates for CIS countries and emerging economies in Central and Eastern Europe.
  - Commodity dependence: in 2022, the Commodity Price Index "increased by around a third," with food as main contributor to CPI inflation over the past three years; energy was second largest until early 2022.
- Wages, fiscal and administrative influences:
  - Positive correlation between wage growth and inflation in most CCA countries.
  - Since early 2022, strong nominal wage growth while inflation decelerated, producing real wage gains.
  - Fiscal response to pandemic: on average CCA countries adopted fiscal measures equivalent to "about 4.7 percent of 2020 GDP (calculated as arithmetic average)"; range from "1.2 percent of GDP in Armenia to over 9 percent of GDP in Kazakhstan."
  - Administrative measures: regulated utility tariffs across all countries; fuel price regulation in Azerbaijan, Kazakhstan, Uzbekistan; export restrictions (Kazakhstan), price controls (Tajikistan), tax reductions for socially sensitive goods (Azerbaijan, Kyrgyz Republic).
  - Gasoline prices deregulated in Armenia, Georgia, Kyrgyz Republic, Uzbekistan; in Kazakhstan only AI-95 fuel price is not regulated.

### Empirical approaches (Section 5 overview)
- Three complementary empirical strategies:
  - Phillips curve (PC) estimation:
    - Baseline determinants: lagged inflation, inflation expectations, output gap.
    - Augmented with external factors: NEER, global oil and food prices, foreign inflation.
  - Panel vector autoregression (pVAR):
    - Similar variable set; used to analyze timing, persistence, and contributions to medium-term price level changes.
    - Extensions evaluate link from wages to inflation and interactions with external factors.
  - Local projections (Jordà 2005), following Brandao-Marques et al. (2020):
    - Study transmission of policy rates to inflation and interactions with NEER, foreign inflation, and global food prices.

### Key empirical findings — Phillips curve estimates (Section 5.1)
- Model specifications (representative):
  - Equation (1): 휋i,t = 훽1 휋i,t−1 + 훽2 휋ei,t + 훽3 Y(g)i,t + 훽4 ei,t + 훾i + 휀i,t
  - Equation (2): 휋i,t = 훽1 휋i,t−1 + 훽2 휋ei,t + 훽3 Y(g)i,t + 훽4 ei,t + Zf 휑t−1′ + 훾i + 휀i,t
  - Variables: year-on-year headline or core inflation, expected inflation (휋ei,t), output gap (Y(g)i,t), annual change in NEER (ei,t), external factors vector (oil prices, food prices, foreign inflation) lagged one quarter.
- Sample and estimation:
  - Period: 2001:Q1-2022:Q4.
  - Panel: seven CCA countries (excluding Turkmenistan).
  - Estimator: least squares with clustered standard errors; variables standardized in robustness checks; country-level bootstrapped confidence intervals.
- Main quantitative results:
  - Persistence: lagged inflation coefficient about "0.7-0.8" in baseline panel estimates; using seasonally adjusted quarterly inflation, lagged coefficient falls to between "0.21 and 0.45."
  - Inflation expectations: positive association; coefficient approximately "0.15" in standard specification.
  - NEER: negative coefficient (currency depreciation drives inflation upwards).
  - Output gap: positive sign, coefficients "between 0.02 and 0.08"; not significant in core inflation regressions; loses significance when foreign factors included.
  - External factors:
    - Global food prices and foreign inflation: significant positive association with headline domestic inflation.
    - Quantitative elasticity: "a one percent increase in international food prices or foreign inflation is, on average, associated with approximately a 0.1 percent growth in domestic CPI."
    - Food prices–core link weaker but still significant.
    - Oil prices: no significant association with domestic inflation in panel estimates (possible role of administered prices, subsidies, long-term contracts).
- Country-level heterogeneity (standardized, bootstrapped):
  - Inflation persistence present in every country, slightly less in Tajikistan.
  - Stronger expectation–inflation relationship in Kyrgyz Republic, Kazakhstan, Armenia; weaker in Tajikistan, Azerbaijan, Georgia.
  - Output gap relationship generally uniform except weaker/not significant in Kyrgyz Republic and Armenia.
  - NEER strongly associated with inflation in most countries; not significant in Kyrgyz Republic.
  - Food prices: strong impact region-wide, less so in Tajikistan.
  - Foreign inflation: significant driver of headline inflation only in Kazakhstan and Azerbaijan.
  - Oil prices: not statistically significant at country level.

### Key empirical findings — pVAR results (Section 5.2)
- Baseline and identification:
  - pVAR: Yi,t = Ai(L) Yi,t−1 + ui + εi,t with Yi,t initially headline inflation, NEER, output gap; augmentations add oil prices, food prices, foreign inflation.
  - Baseline ordering: NEER (most exogenous), output gap, inflation; foreign variables ordered exogenous when included.
  - Estimator: GMM; lag selection by AIC/BIC; confidence intervals via simulation and bootstrap.
- Dynamic impulse-response magnitudes (cumulative, quarter-on-quarter changes):
  - "A one percent increase in foreign inflation can cause a 1.7 percent rise in domestic inflation within the first two quarters."
  - "A one percent increase in global food prices can bring an additional 0.23 percent in domestic inflation in the first four quarters from the shock."
  - "A one percent depreciation in the exchange rate adds 0.20 percent to domestic inflation within the first six quarters."
  - Oil price shocks: negative immediate effect in some specifications, turning positive after a couple of quarters; overall limited direct effect.
  - Output gap changes: small effects, often low statistical significance.
- Forecast error variance decomposition (FEVD), average four- to six-quarters-ahead (standardized variables):
  - Own inflation shock accounts for "approximately 50 percent" of error variance for both headline and core inflation.
  - Food prices cumulatively account for "approximately 35 percent" of FEVD of headline inflation.
  - Food prices contribution to core inflation FEVD: "approximately 0.2 percent."
  - Foreign CPI: next-largest contributor to headline CPI variations after food prices.
  - NEER: share of error variance explained is considerably higher for core CPI than for headline CPI.
  - Oil prices and output gap: considerably lower contributions.

### Wage growth, foreign prices, and inflation dynamics (pVAR extensions)
- Two pVAR models including nominal wage growth:
  - Model A: NEER, domestic inflation, output gap, annualized nominal wage changes.
  - Model B: adds food prices and foreign inflation to Model A.
  - Ordering (example): food prices, foreign inflation, NEER, domestic inflation, wages, output gap; alternative orderings yield similar results.
- Reported effect of wages:
  - When global factors are excluded: a one percent increase in wage growth is cumulatively associated on average with up to "0.4 percent increase in inflation three to four" (text truncated; subsequent clarifying bullet provided in annex).
  - Explicit stated result: "A 0.4 percent increase in inflation is observed three to four quarters after the shock to wage growth."
  - Impact of wage growth on inflation becomes weaker once global food prices and partners’ inflation are included; suggests wages partly reflect external prices and second-round effects.

### Role of monetary policy — local projections and results
- Identification of monetary policy shocks:
  - Country-specific Taylor-rule model includes lagged inflation (π_t−1), inflation expectations (π^e_t), output gap (Y^(g)_t), change in NEER (e_t), lagged change in policy rate (∆i_t−4); monetary policy shock is residual (ε_t̂).
- Local projection specification:
  - π_i,t+h = β1 π_i,t−1 + β2 ε_i,t̂ + Z_f ϕ′ + γ_i + u_i,t with horizon h = 1..6 quarters; Z_f includes lags of food prices and foreign inflation; ε_i,t̂ standardized.
  - Interaction specification adds k_t and θ(ε_i,t̂ × k_i,t) to test conditional effects where k_t is contemporaneous movement in NEER, food prices, or foreign CPI.
- Quantitative transmission results:
  - "A 1 percent increase in the policy rate is associated with approximately 0.3-0.4 percentage point reduction in inflation, five to six quarters after the shock;" effect statistically significant after five quarters.
  - Effectiveness stronger when tightening coincides with nominal NEER appreciation.
  - Monetary policy limits pass-through from foreign factors:
    - Example for food prices: pass-through to domestic inflation estimated at "around 0.02 percent in the first four quarters after a shock, on average, when policy tightens" versus "0.2 percent in case of no policy tightening."
    - For foreign inflation: estimated pass-through is "close to zero" when shock is accompanied by interest rate increase.
    - Effects of foreign prices not statistically significant from two quarters after the shock when policymakers tighten.

### Robustness checks and alternative specifications
- Phillips curve re-estimation with sample starting January 2015: results broadly align with baseline, with somewhat stronger impact from foreign inflation.
- Alternative proxy for inflation expectations:
  - Regression-based one-year ahead expectations (Hamilton et al. 2016; Jordà et al. 2022) has standard deviation "around 2 percent" versus "over 6 percent" for realized year-on-year inflation.
  - When standardized, coefficient sizes considerably smaller at around "0.4"; headline CPI coefficients for expectations larger when untransformed variables used due to limited variation in alternative measure.
- Estimator checks:
  - Addressed bias from fixed effects with lagged variables using random effects and system GMM (Nickell 1981; Arellano and Bond 1991).
  - Findings: slightly greater association between past and current inflation in constrained and median regressions; larger coefficients for foreign inflation under random effects and GMM.
  - Overall signs and sizes remain in line with baseline estimates.

### Summary of empirical findings (Section 5.3 summary)
- Inflation in the CCA is substantially persistent and influenced by both foreign and domestic factors.
- Phillips curve estimates:
  - Inflation expectations and domestic demand (output gap) play modest but statistically significant roles.
  - Strong exchange rate pass-through to inflation.
  - Foreign inflation and global food prices exert strong influence; global oil prices weaker.
- Monetary policy:
  - Central banks have limited influence over local price dynamics given strong global factor roles, but domestic policy matters.
  - Monetary policy operates with a lag of about "4-6 quarters" and is most effective when accompanied by exchange rate appreciation.
  - Monetary policy responses can limit impacts of global shocks (food and energy) on domestic inflation.

### Policy implications and recommendations
- Monetary policy and frameworks:
  - Strengthen monetary policy frameworks in the CCA to manage inflation pressures more effectively.
  - Resolve structural weaknesses (high dollarization, underdeveloped financial markets) to improve transmission (see Poghosyan et al., 2023).
  - Stronger commitment to price stability associated with weaker persistence, lower exchange rate pass-through, and better-anchored expectations (Mishkin and Hebbels 2007; Brandao-Marques, Gelos et al. 2020).
  - Pursue prudent macro policies to reduce dollarization (remains elevated despite recent decline; Cakir et al., 2022).
- Fiscal, wage, and administrative policies:
  - Fiscal and wage policies matter for inflation outcomes.
  - Avoid administrative policies to control inflation (price controls, trade restrictions) as they are "fiscally costly, ineffective, and counterproductive in the longer term."
  - Allow prices to adjust to preserve resource allocation, energy conservation incentives, and private investment (IMF, 2022a).
- Data and measurement caveats:
  - Smaller measured contribution from domestic factors could reflect measurement difficulties for economic slack and inflation expectations given data limitations in the CCA.

### Data coverage, variables, and key statistics (Annex I highlights)
- Country-level observation counts (seven countries; Turkmenistan excluded):
  - Armenia 623
  - Azerbaijan 567
  - Georgia 510
  - Kazakhstan 569
  - Kyrgyz Republic 578
  - Tajikistan 545
  - Uzbekistan 285
- Selected variable means and standard deviations (quarterly, year-on-year percent change unless noted):
  - Headline Inflation: Observations 548; Mean 7.28; Standard Deviation 6.04
  - Core Inflation: Observations 382; Mean 4.98; Standard Deviation 4.99
  - Output Gap (level): Observations 557; Mean 0.03; Standard Deviation 4.30
  - Nominal Effective Exchange Rate: Observations 616; Mean -2.06; Standard Deviation 11.36
  - Food Prices (UN FAO): Observations 616; Mean 4.27; Standard Deviation 13.54
  - Brent Crude Oil Prices (Bloomberg): Observations 616; Mean 11.54; Standard Deviation 34.55
  - Foreign Inflation (CHN, RUS, EU average): Observations 616; Mean 4.71; Standard Deviation 1.97
  - Wages (annualized quarterly growth): Observations 476; Mean 15.67; Standard Deviation 11.49
  - Inflation Expectations (t+3, WEO): Observations 616; Mean 6.70; Standard Deviation 11.74
  - Policy Rate: Observations 482; Mean 8.50; Standard Deviation 4.47
- Note: All variables quarterly; output gap in levels; Turkmenistan excluded due to data limitations.

### Inflation decomposition approach (Annex II)
- Statistical framework for PCA decomposition:
  - π_ijt = λ_ij f_t + μ_ij g_jt + e_ijt where π_ijt is standardized annual CPI change, f_t global factor, g_jt regional factor, e_ijt country-specific.
- Sample and frequency:
  - G20 economies and CCA countries; monthly data from January 2017 to May 2024; focus on headline inflation.
  - Region defined as the CCA and Russia; remaining countries constitute the global economy.
- Estimation steps:
  1. Extract global component via principal components for whole sample.
  2. Extract regional factor from residuals for CCA countries and Russia.
  3. Regress country inflation on global and regional components; residual is country-specific component.
- Note: PCA captures correlations and descriptions of dynamics but does not identify transmission mechanisms.

*Source: IMF Working Paper — wpiea2025003-print-pdf (excerpts from References, Sections 3 and 5, and Annexes).*

### References .............................................................................................................

### wpiea2025003-print-pdf - References .............................................................................................................

### Figures and Tables (listed)
- Figures mentioned include:
  - Figure 1. Inflation in the CCA over the longer run
  - Figure 2. CCA Countries and Russia
  - Figure 3. Inflationary Shocks Affecting the CCA in 2020-23
  - Figure 4. Domestic and External Components of Inflation in the CCA
  - Figure 5. CPI Growth vs. Wage Growth
  - Figure 6. Administered Prices and Inflation
  - Figure 7. Inflation Developments in the CCA during 2020-23
  - Figure 8. Country-level Phillips Curve Estimates, Baseline Model
  - Figure 9. Country-level Phillips Curve Estimates, Augmented Model
  - Figure 10. Impulse Response Functions, pVAR Models of Inflation Correlates, Baseline Specification
  - Figure 11. Forecast Error Variance Decomposition for pVAR Estimates, Including External Factors
  - Figure 12. Estimated Response of Inflation to a Change in Wage Growth
  - Figure 13. Estimated Response of Inflation to Monetary Policy Tightening
  - Figure 14. Estimated Response of Inflation to Shocks in Food Prices and Foreign Inflation
- Tables mentioned include:
  - Table 1. Phillips Curve Estimation, Augmented by External Factors
  - Table 2. Phillips Curve Estimation, Augmented by External Factors, Alternative Sample Period (2015-2022)
  - Table 3. Phillips Curve Estimation, Augmented by External Factors: Alternative Proxy for Inflation Expectations
  - Table 4. Phillips Curve Estimation, Augmented by External Factors, Alternative Estimators

### Introduction: regional inflation trajectory and motivation
- Inflation in most CCA countries:
  - "almost doubled to 13 percent in 2022 from around 7 percent in 2020"
  - "significantly above the global average of 8.7 percent in 2022"
  - "before dropping to 9.8 percent in 2023 and continuing to trend down in 2024"
- Upswing drivers cited:
  - Dislocations from the COVID-19 pandemic (supply chain disruptions, large fiscal stimulus, accommodative monetary policies, pent-up consumer demand)
  - Russia’s invasion of Ukraine affecting energy and food prices
- Purpose: investigate inflation dynamics in the CCA, main drivers, and provide conclusions to inform policymaking; first study focusing on the CCA region to the authors' knowledge.

### Empirical approaches used
- Three complementary empirical strategies:
  - Phillips curve (PC) estimation:
    - Standard determinants: lagged inflation, inflation expectations, measure of economic slack
    - Augmented with foreign factors: nominal effective exchange rates, global oil and food prices, and foreign inflation
  - Panel vector autoregression (pVAR):
    - Similar variable set as PC
    - Used to analyze timing, persistence, and quantify contribution to medium-term price level changes
    - Extension includes evaluating link from wages to inflation and interactions with external factors
  - Local projections (Jordà 2005) to study monetary policy:
    - Following Brandao-Marques et al. (2020)
    - Explore transmission of policy rates to inflation and interactions with NEER, foreign inflation, and global food prices

### Key empirical findings (as stated)
- Role of external vs. domestic factors:
  - "a greater contribution to inflation dynamics from external factors"
  - "domestic factors are found to also contribute to price dynamics even if more modestly"
- Significant and large effects from:
  - Foreign inflation
  - Global food prices
- Inflation persistence and pass-through:
  - "significant persistence of inflation processes in the CCA"
  - "pass-through from exchange rate fluctuations"
- Global oil prices:
  - Mixed findings: PC analysis points to no significant impact on domestic prices, while pVAR suggests a "small lagged effect"
  - Possible explanation: role of administered prices, subsidies, and long-term contracts in several countries
- Monetary policy effectiveness:
  - "an increase in the policy rate is associated with a statistically significant decline in inflation four- to six quarters after the policy decision"
  - Effectiveness stronger when accompanied by a nominal effective exchange rate appreciation
  - "Pass-through from external factors to domestic prices is found to be substantially smaller when a rise in food prices or foreign inflation is met by a policy tightening"

### Policy implications and recommendations (as stated)
- Policies matter for managing inflation outcomes in the region:
  - Monetary policy is important, especially when exchange rates are allowed to adjust
  - Fiscal and wage policies also play important roles
  - Administrative measures (e.g., price controls) "may be counterproductive, and should be avoided"
- Data and measurement caveats:
  - Smaller measured contribution from domestic factors could reflect difficulties in measuring economic slack and inflation expectations given data limitations in the CCA

### Literature context (selected points)
- Related strands:
  - Growing influence of foreign factors on domestic inflation (Auer et al. 2017; Borio and Filardo 2007; Ball 2006; Bems et al. 2022; Calza 2008; Forbes 2020)
  - Phillips curve augmented with global factors: Galí and Gertler (1999); Galí et al. (2001, 2003); Auer et al. (2017); Kamber et al. (2020); IMF (2022c)
  - VAR and structural VAR applications to inflation drivers: IMF (2021); Finck and Tilman (2022); Minasyan et al. (2023)
  - Panel VAR usage noted as less frequent; reviewed in Canova and Cicarelli (2013)
  - Local projections for policy impulse responses: Jordà (2005); Brandao-Marques et al. (2020); Binici et al. (2022); Baba and Lee (2022)
  - Monetary policy transmission in the CCA and regional context: Poghosyan et al. (2023); IMF (2023a); Brandao-Marques et al. (2020); Eklou (2023)
  - Importance of credible and independent central banks highlighted (Dincer et al., 2014; IMF, 2024), with Poghosyan et al. (2023) detailing CCA challenges (weak bank competition, high dollarization, small and illiquid capital markets)

*IMF Working Papers — Drivers of Inflation in the Caucasus and Central Asia (excerpts from References and introductory sections)*

### 3. Stylized facts and recent developments

### 3. Stylized facts and recent developments

### Historical inflation episodes and medium-term patterns
- The CCA region experienced hyperinflation in the early 1990s following the collapse of the Soviet Union.  
- As countries transitioned to market economies, macroeconomic stability was gradually re-established:
  - Inflation returned to single digits by the end of the decade in Armenia, Azerbaijan, Georgia, Kazakhstan.
  - Uzbekistan has had inflation in the low single digits for most of its recent history.  
- Inflation accelerated again in the period preceding the Global Financial Crisis.  
- The slump in oil prices starting in 2014 and ensuing currency depreciations pushed inflation up in the region’s oil exporting countries and contributed to price moderation in oil importers.

### Recent shocks (2020–23) and external drivers
- Inflation in the CCA started accelerating in 2020 driven by multiple shocks:
  - The COVID-19 pandemic in 2020 caused global supply chain disruptions and supply shortages.
  - Large fiscal and monetary stimulus globally supported demand during and after the pandemic.
  - Strong post-pandemic recovery was accompanied by surging energy and food prices.
  - The breakout of the war in Ukraine in February of 2022 reinforced global commodity price trends.
  - Volatility of the Russian ruble was important given the region’s reliance on imports from Russia; the ruble underwent three phases since the war’s commencement: initial depreciation, sharp appreciation, and more recent depreciation to levels below those observed before the onset of the conflict.
- Principal component analysis (PCA) decompositions show:
  - A large global component that was a moderate contributor to inflation up to 2020 and became a major contributor to the post-Covid inflation surge.
  - Tajikistan and Uzbekistan display more idiosyncratic behavior with less dominance of the global component.
  - The war in Ukraine appears to have affected the regional component differently for oil importers and exporters:
    - In Armenia, Georgia, and Tajikistan the regional component contributed to disinflation.
    - In Azerbaijan and Kazakhstan the regional component rose, offsetting a smaller contribution from the global component.
  - The domestic component has been generally more stable; in Armenia domestic factors have been important (smaller than foreign factors) in both upward and downward inflation trends.

### Developments since 2023 and exchange-rate dynamics
- Since 2023 inflation pressures have subsided across the majority of CCA countries due to:
  - Dissipation of base effects.
  - Decline in commodity prices (i.e., food oil and gas).
  - Tighter monetary policies.
  - Strong exchange rate appreciations in several countries that tempered inflationary trends.
- Notable currency appreciations were recorded in Armenia, Georgia, and Tajikistan, spurred by substantial capital inflows from Russia.
- Several countries began loosening monetary policy as inflation moderated (Armenia, Azerbaijan, Georgia, Kazakhstan, Kyrgyz Republic, Tajikistan).

### Exchange rate pass-through, commodity prices, and main contributors to inflation
- Exchange rate pass-through:
  - For the CCA, the average pass-through to inflation of a depreciation of the local currency vis-a-vis the US dollar has typically been estimated at around 10 percent on impact and about 25 percent after 12 months.
  - The pass-through has been found to be broadly similar in fixed versus floating exchange rate regimes.
  - This average pass-through is higher than those estimated for the Commonwealth of Independent State (CIS) countries and for emerging economies in Central and Eastern Europe.
- Global commodity prices:
  - High import dependence exposes CCA countries to fluctuations in global commodity prices.
  - In 2022, the Commodity Price Index increased by around a third, leading to the observed peak in inflation; the subsequent retraction of commodity prices contributed to the disinflation that followed.
  - Food has been the main contributor to CPI inflation dynamics over the past three years; energy prices were the second largest contributor until early 2022 when services and other domestic factors gained strength.

### Wages, fiscal and administrative policy influences
- Wages:
  - Wage growth and inflation have exhibited a positive correlation in most CCA countries.
  - Since early 2022, nominal wages have grown strongly while inflation decelerated, producing significant gains in real wages; this could reflect improved productivity but could also lead to a buildup of inflationary pressures.
- Fiscal and monetary policy:
  - To address the effects of the pandemic, on average CCA countries adopted fiscal measures equivalent to about 4.7 percent of 2020 GDP (calculated as arithmetic average).
  - The range of fiscal support was from 1.2 percent of GDP in Armenia to over 9 percent of GDP in Kazakhstan.
  - Protracted accommodative monetary policies in some CCA countries contributed to inflation; policy rates were reduced in 2020 in Kazakhstan, Armenia and Uzbekistan and kept low even as inflation gathered pace, contributing to fast growth of the money supply.
  - Subsequent policy tightening contributed to disinflation and gradual normalization as inflation declined towards objectives.
- Administrative and regulatory measures:
  - Some countries used administered prices to contain inflation; administered prices are a relatively small share of CPI but in some cases moderated inflation surges at the expense of price distortions and expanded subsidies.
  - All countries have regulated utility tariffs for electricity, natural gas, heating, and water and sewage.
  - Some countries regulate prices for certain fuels (Azerbaijan, Kazakhstan, Uzbekistan) and domestic transportation (Azerbaijan, Georgia).
  - Domestic prices of certain regulated products were adjusted to reflect international price variations in Armenia and Georgia.
  - Contracts to purchase gas at fixed prices helped limit domestic price volatility in Armenia, Kyrgyz Republic and Uzbekistan.
  - Other measures included export restrictions (Kazakhstan), price controls (Tajikistan), and reduced tax rates for socially sensitive goods (Azerbaijan, Kyrgyz Republic).
  - Gasoline prices are deregulated in Armenia, Georgia, Kyrgyz Republic, and Uzbekistan. In Kazakhstan only the price of AI-95 fuel is not regulated.

### Data, empirical approach and key data characteristics
- Coverage:
  - The empirical analysis covers all countries of the CCA region except Turkmenistan (insufficient data).
  - Sample period: quarterly data for 2001:Q1 through 2022:Q4.
- Dependent variables and regressors:
  - Dependent variables in models: headline and core consumer price inflation (core generally excludes volatile components such as food and energy; country-specific definitions in Annex I).
  - Regressors in the PCA include lagged inflation, the output gap, inflation expectations, and a set of external variables: changes in the NEER, global food prices, global oil prices, foreign CPI, and the foreign output gap.
- Variable transformations and robustness:
  - Variables are transformed into year-on-year growth rates to mitigate non-stationarity and seasonal effects; a robustness check uses seasonally adjusted quarterly growth rates.
  - pVAR analysis examines the relationship between wages and inflation.
- Unobservables and estimation:
  - Output gap measure: percent difference between actual annualized output and trend output computed with a Hodrick-Prescott filter with a smoothing parameter of 1600.
  - Inflation expectations (baseline): three-year ahead forecast of CPI inflation from the IMF’s World Economic Outlook (WEO); lagged series of WEO inflation forecasts three years ahead used as a proxy for inflation expectations.
  - Robustness alternative: measure of expectations based on Hamilton (2016) and Jordà et al. (2022), regressing one-year ahead inflation on lags of CPI (quarter-on-quarter annualized change).
- Data availability and variability:
  - Number of available observations ranges from a total of 623 for Armenia to only 285 for Uzbekistan across seven series (see Annex I).
  - Standard deviation of year-on-year changes in the sample: CPI about 6 percent, NEER about 11 percent, and roughly 2 percent for trading partners’ inflation and some proxies for inflation expectations.
  - Given high variability, estimates are also presented after standardizing main variables (zero mean and unit standard deviation) for interpretability.

*Source: IMF Working Paper — Section 3, "Stylized facts and recent developments."*

### 5.  Empirical analysis

### 5.  Empirical analysis

### 5.1. Inflation drivers: an augmented Phillips curve for the CCA — baseline and estimation setup
- Model specifications:
  - Equation (1): 휋i,t = 훽1 휋i,t−1 + 훽2 휋ei,t + 훽3 Y(g)i,t + 훽4 ei,t + 훾i + 휀i,t
  - Equation (2): 휋i,t = 훽1 휋i,t−1 + 훽2 휋ei,t + 훽3 Y(g)i,t + 훽4 ei,t + Zf 휑t−1′ + 훾i + 휀i,t
  - Variables: 휋i,t = year-on-year headline or core inflation; 휋ei,t = expected inflation; Y(g)i,t = output gap; ei,t = annual change in the nominal effective exchange rate (NEER); 휑t−1′ = vector of external factors (oil prices, food prices, foreign inflation) entering with one-quarter lag; 훾i = country fixed effects.
- Sample and estimation:
  - Period: 2001:Q1-2022:Q4.
  - Panel: seven CCA countries (excluding Turkmenistan).
  - Estimators and inference: least squares with standard errors clustered by countries; variables also standardized to zero mean and unit standard deviation for relative-coefficient comparisons; country-level estimates with bootstrapped confidence intervals to address limited observations.
- Methodological caveats noted in the source:
  - Measurement problems (inflation expectations, output gap) may bias estimates.
  - Correlation among regressors, especially global variables, may reduce precision.
  - Robustness checks include alternative proxies for expectations and models extracting autonomous oil price movements.

### 5.1. Estimation results — main findings from the augmented Phillips curve
- Persistence:
  - Lagged inflation coefficient is large and significant, about 0.7-0.8 in baseline panel estimates.
  - Using seasonally adjusted quarterly inflation rates, the lagged inflation coefficient falls to between 0.21 and 0.45 across specifications, indicating continued inertia.
- Inflation expectations and NEER:
  - Inflation expectations have a positive association with inflation; coefficient approximately 0.15 in the standard specification.
  - NEER shows a negative coefficient (i.e., currency depreciation drives inflation upwards).
- Output gap:
  - Sign is positive across specifications; coefficients tend to be small, between 0.02 and 0.08.
  - Estimates for the output gap are not significant in the core inflation regressions.
  - When foreign factors are included, the output gap loses statistical significance (possible correlation between domestic demand proxy and foreign factors).
- External factors:
  - Significant positive association between global food prices and headline domestic inflation.
  - Significant positive association between foreign inflation and headline domestic inflation.
  - Quantitative elasticity: a one percent increase in international food prices or foreign inflation is, on average, associated with approximately a 0.1 percent growth in domestic CPI.
  - Link between food prices and core inflation is weaker but still significant (possible second-round effects).
  - Oil prices: no significant association with domestic inflation in panel estimates (potentially due to administered prices, subsidies, and long-term contracts).
- Country-level heterogeneity (from country-by-country standardized estimates; confidence intervals via bootstrap):
  - Inflation persistence present in every country, slightly less in Tajikistan.
  - Relationship between inflation and inflation expectations:
    - Relatively stronger for the Kyrgyz Republic, Kazakhstan, and Armenia.
    - Weaker for Tajikistan, Azerbaijan, and Georgia.
  - Relationship with the output gap:
    - More uniform across countries, except Kyrgyz Republic (weaker and not statistically significant) and Armenia (not statistically significant).
  - NEER:
    - Strongly associated with inflation in most countries, except the Kyrgyz Republic (not significant).
  - Food prices:
    - Strong impact throughout the region, but less so in Tajikistan.
  - Foreign inflation:
    - Significant driver of headline inflation only in Kazakhstan and Azerbaijan.
  - Oil prices:
    - Not statistically significant at country level, consistent with panel results.

### 5.2. Inflation dynamics: pVAR approach — setup and estimation
- Model: panel VAR along Canova and Ciccarelli (2013) framework:
  - Equation (3): Yi,t = Ai(L) Yi,t−1 + ui + εi,t
  - Yi,t is a vector of endogenous variables. Initial set: headline inflation, NEER, output gap. Augmentations add oil prices, food prices, and foreign inflation one at a time.
- Identification and ordering:
  - Baseline ordering: NEER (most exogenous), then domestic output gap, then inflation.
  - When including foreign variables: foreign variables ordered as exogenous to domestic factors; ordering of remaining variables unchanged.
  - For FEVD and some pVARs, external variables ordered as oil prices, food prices, foreign inflation (most exogenous), followed by NEER, output gap, inflation.
- Estimation details:
  - Quarterly data, same period 2001:Q1-2022:Q4.
  - Estimator: Generalized Method of Moments (GMM).
  - Lag selection: Akaike and Bayesian information criteria.
  - Confidence intervals: simulated standard errors from pVAR implementation (Abrigo and Love 2015).
  - Bootstrap inference for impulse responses where reported.

### 5.2. pVAR results — impulse responses and magnitudes
- General consistency with PC results: substantial inflation persistence; quick pass-through of foreign inflation, global food prices, and exchange rate fluctuations; modest role for domestic demand changes.
- Key quantitative dynamic effects (cumulative impulse responses, quarter-on-quarter changes):
  - A one percent increase in foreign inflation can cause a 1.7 percent rise in domestic inflation within the first two quarters.
  - A one percent increase in global food prices can bring an additional 0.23 percent in domestic inflation in the first four quarters from the shock.
  - A one percent depreciation in the exchange rate adds 0.20 percent to domestic inflation within the first six quarters.
  - Oil price shocks show a negative immediate effect in some specifications, turning positive after a couple of quarters (overall limited and counterintuitive direct effect).
  - Changes in the output gap have a small effect on inflation, often with lower statistical significance.
- Visualization: orthogonalized impulse response functions with 90% simulated confidence intervals were used to display dynamics (figures referenced in source).

### 5.2. pVAR results — forecast error variance decomposition (FEVD)
- FEVD focus: average contributions four to six quarters ahead, using standardized variables and pVAR models including headline and core CPI.
- Main FEVD findings (average four- to six-quarters-ahead):
  - A shock to CPI (own inflation shock) accounts for approximately 50 percent of the error variance for both headline and core inflation, confirming substantial persistence.
  - Food prices cumulatively account for approximately 35 percent of the forecast error variance of headline inflation.
  - Food prices contribution to core inflation FEVD is lower, at approximately 0.2 percent.
  - Foreign CPI is the next-largest contributor to headline CPI variations after food prices.
  - The share of error variance explained by the NEER is considerably higher for core CPI than for headline CPI.
  - Contributions of oil prices and the output gap to the FEVD are considerably lower.

### Wage growth and inflation dynamics (pVAR extension)
- Setup:
  - Two separate pVAR models include nominal wage growth:
    - Model A: NEER, domestic inflation, output gap, and annualized changes in nominal wages.
    - Model B: adds food prices and foreign inflation to Model A.
  - Ordering used in the extended model: food prices (most exogenous), foreign inflation, NEER, domestic inflation, wages, output gap. Alternative orderings yield similar results.
- Reported result preview:
  - Cumulative IRFs with standardized variables are presented (Figure 12 in source).
  - When global factors are excluded (blue line in Figure 12), a one percent increase in wage growth is cumulatively associated on average with up to [text truncated in source before numeric completion].

*Source: IMF Working Paper — chapter 5: "Empirical analysis" (file: wpiea2025003-print-pdf - 5.  Empirical analysis).*

### 0.4 percent increase in inflation three to four

### wpiea2025003-print-pdf - 0.4 percent increase in inflation three to four

### Wage growth, foreign prices, and inflation dynamics
- A 0.4 percent increase in inflation is observed three to four quarters after the shock to wage growth.
- The impact of wage growth on inflation becomes weaker once global food prices and partners’ inflation are included in the analysis (orange line in the plot), increasing somewhat after two quarters.
- The smaller response of inflation to wage growth after controlling for foreign prices suggests wages in part reflect external prices, possibly indicating second-round effects from food prices and foreign inflation to wages.
- Estimates highlight:
  - Strong pass-through to inflation from fluctuations in the exchange rate.
  - Strong influence of inflation in the rest of the world and global food prices.
  - Global oil prices show a weaker association with domestic inflation, likely reflecting widespread regulation of energy prices and long-term contracts across the region.

### Robustness checks and alternative specifications
- Baseline analysis uses data for the period 2001-2020.
- Re-estimating Phillips curve models with data starting in January 2015 produces results broadly in line with baseline estimates, with a somewhat stronger impact from foreign inflation.
- Alternative proxy for inflation expectations:
  - Expectations derived from a regression of one-year ahead inflation against four lags of quarter-on-quarter inflation (Hamilton et al. (2016); Jordà et al. (2022)).
  - The standard deviation of this alternative measure of inflation is around 2 percent, as opposed to over 6 percent for realized year-on-year inflation.
  - When standardized variables are used, the size of the coefficients is considerably smaller at around 0.4.
  - For headline CPI, coefficients for inflation expectations are substantially larger when untransformed variables are used in the regression (columns 1-2), given limited variation in the alternative measure.
- Additional modeling and estimator checks:
  - Addressed potential bias from fixed effects with lagged variables (Nickell 1981; Arellano and Bond 1991) by using random effects and system GMM.
  - Findings: slightly greater association between past and current inflation with constrained and median regression; larger coefficients for foreign inflation using random effects and GMM.
  - Overall, sign and size of key coefficients remain in line with baseline estimates.

### Role of monetary policy — empirical framework
- Three exercises to evaluate transmission from policy interest rates to inflation:
  1. Unconditional response of inflation to monetary policy shocks.
  2. Response considering movements in the exchange rate.
  3. Response considering changes in external conditions (global food prices and foreign inflation).
- Monetary policy shocks identified for each country using a Taylor-rule model including the NEER.
- Quarterly estimated equation for policy rate changes (∆i_t) includes:
  - Lagged year-on-year inflation (π_t−1), inflation expectations (π^e_t), output gap (Y^(g)_t), year-on-year change in the NEER (e_t), and lagged change in the policy rate (∆i_t−4).
  - The monetary policy shock is the residual (ε_t̂) — the fraction of interest rate change unexplained by observed data.
- Responses of inflation to policy shocks estimated using local projections (Jordà 2005):
  - Specification: π_i,t+h = β1 π_i,t−1 + β2 ε_i,t̂ + Z_f ϕ′ + γ_i + u_i,t
  - ε_i,t̂ is the standardized country-specific policy shock; Z_f includes lags on food prices and foreign inflation; γ_i are country fixed effects.
  - Horizon h varies from 1 to 6 quarters; shock assumed to have no contemporaneous impact on inflation; standard errors clustered by country.
- To test external environment effects on transmission, an interaction term is added:
  - π_i,t+h = β1 π_i,t−1 + β2 ε_i,t̂ + β3 k_t + θ(ε_i,t̂ × k_i,t) + Z_f ϕ′ + γ_i + u_i,t
  - k_t corresponds sequentially to contemporaneous movement in the NEER, food prices, or foreign CPI.
  - θ measures the effect of a monetary policy shock conditional on a change in the external variable.

### Role of monetary policy — estimation results
- Monetary policy tightening has a significant effect on inflation, especially through movements in the nominal exchange rate.
- A 1 percent increase in the policy rate is associated with approximately 0.3-0.4 percentage point reduction in inflation, five to six quarters after the shock; the effect is statistically significant after five quarters.
- Impact of interest rate changes is somewhat stronger when tightening is associated with a nominal appreciation.
- Monetary policy limits pass-through from foreign factors to inflation:
  - Response functions for a 1 percent increase in food or foreign CPI show a significant increase in inflation several quarters after the shock when no policy tightening occurs (green line).
  - Effects are considerably smaller when coinciding with monetary policy tightening (blue line).
  - Example for food prices:
    - Pass-through to domestic inflation estimated at around 0.02 percent in the first four quarters after a shock, on average, when policy tightens.
    - Pass-through estimated at 0.2 percent in case of no policy tightening.
  - For foreign inflation, estimated pass-through is close to zero when the shock is accompanied by an increase in interest rates.
  - In both instances, effects of foreign prices are not statistically significant starting from two quarters after the shock when policymakers opt for tightening.
- Findings are consistent with literature asserting prudent monetary policy anchors expectations and limits second-round effects from foreign price pressures to expectations and broad-based domestic price increases (Blanco, Ottonello, and Ranosova 2022; Reis 2022).

### Summary of empirical findings
- Inflation in the CCA region is substantially persistent and influenced by both foreign and domestic factors.
- Phillips curve estimates:
  - Inflation expectations and domestic demand conditions (output gap) play a modest but statistically significant role.
  - Strong exchange rate pass-through to inflation.
  - Foreign inflation and global food prices exert strong influence; global oil prices weaker.
- Central banks have limited influence over local price dynamics due to strong global factor roles, but domestic policy still matters.
- Monetary policy operates with a lag of about 4-6 quarters and is most effective when accompanied by exchange rate appreciation.
- Monetary policy responses can limit impacts of global shocks (food and energy) on domestic inflation.

### Policy recommendations
- Strengthen monetary policy frameworks in the CCA region to more effectively manage inflation pressures.
  - Resolve structural weaknesses such as high dollarization and underdeveloped financial markets to improve monetary policy transmission (see Poghosyan et al., 2023).
  - Stronger commitment to price stability associated with weaker persistence, lower exchange rate pass-through, and better anchored inflation expectations (Mishkin and Hebbels 2007; Brandao-Marques, Gelos et al. 2020).
  - Pursue prudent macroeconomic policies to further reduce dollarization, which despite recent decline remains elevated (Cakir et al., 2022).
- Avoid administrative policies to control inflation (e.g., price controls and trade restrictions) because they are fiscally costly, ineffective, and counterproductive in the longer term.
  - Allow prices to adjust; price signals are critical to allocate resources, promote energy conservation, and encourage private investment (IMF, 2022a).

*Source: wpiea2025003-print-pdf (IMF Working Paper excerpt).*

### References

### wpiea2025003-print-pdf - References

### Key themes in the references
- Empirical methods for panel VAR, local projections, and impulse response analysis.
- Exchange rate pass-through, global spillovers, and the role of oil and commodity prices in inflation.
- Inflation dynamics: Phillips-curve analyses, expectations, anchoring, and wage-price interactions.
- Monetary policy transmission in emerging markets and the Caucasus and Central Asia (CCA).
- Globalization and its changing influence on inflation processes.
- Central bank transparency, independence, and strengthening monetary policy frameworks in the CCA.

### Notable methodological and empirical sources cited
- Panel VAR estimation and specification tests (Abrigo and Love 2016; Canova and Ciccarelli 2013; Arellano and Bond 1991; Holtz-Eakin et al. 1988; Nickell 1981).
- Local projections and impulse-response inference (Jordà 2005; Ventzislav and Lutz 2005).
- Phillips-curve and structural analyses of inflation dynamics (Gali and Gertler 1999; Del Negro et al. 2020; Jasova et al. 2020).
- Studies of global drivers, supply chain disruptions, and oil price effects (Auer et al. 2017; Finck and Tillmann 2022; Ha et al. 2023; Baba and Lee 2022).
- Country- and region-specific work on pass-through, wage dynamics, and policy frameworks in the CCA (Poghosyan 2020; Poghosyan et al. 2023; Boranova et al. 2019; Cakir et al. 2022).

### Data sources, coverage, and key statistics (ANNEX I)
- Country-level observation counts:
  - Armenia 623
  - Azerbaijan 567
  - Georgia 510
  - Kazakhstan 569
  - Kyrgyz Republic 578
  - Tajikistan 545
  - Uzbekistan 285
  - Note: Turkmenistan was excluded from the empirical analysis due to data limitations.
- Variables (in y-o-y percent change), sources, coverage, observations, mean, standard deviation:
  - Headline Inflation
    - Source: IMF WEO, Haver, MCDHFI.dmxe
    - Coverage: ARM, AZE, GEO, KAZ, KGZ, TJK, UZB *
    - Observations: 548
    - Mean: 7.28
    - Standard Deviation: 6.04
  - Core Inflation
    - Source: IMF WEO, MCDHFI.dmxe
    - Coverage: ARM, AZE, GEO, KAZ, KGZ, TJK
    - Observations: 382
    - Mean: 4.98
    - Standard Deviation: 4.99
    - Notes: Inputs from country desks; generally excludes food and energy
  - Output Gap (level)
    - Source: IMF WEO, Haver, MCDHFI.dmxe, Authors' calculations
    - Coverage: ARM, AZE, GEO, KAZ, KGZ, TJK, UZB *
    - Observations: 557
    - Mean: 0.03
    - Standard Deviation: 4.30
    - Notes: Calculated as percent difference between annualized actual output and trend output using HP filter
  - Nominal Effective Exchange Rate
    - Source: IMF INS (STA)
    - Coverage: ARM, AZE, GEO, KAZ, KGZ, TJK, UZB
    - Observations: 616
    - Mean: -2.06
    - Standard Deviation: 11.36
  - Food Prices
    - Source: UN FAO Food Price Index
    - Coverage: Global
    - Observations: 616
    - Mean: 4.27
    - Standard Deviation: 13.54
  - Brent Crude Oil Prices
    - Source: Bloomberg
    - Coverage: Global
    - Observations: 616
    - Mean: 11.54
    - Standard Deviation: 34.55
  - Foreign Inflation
    - Source: IMF WEO, Authors' calculations
    - Coverage: CHN, RUS, EU
    - Observations: 616
    - Mean: 4.71
    - Standard Deviation: 1.97
    - Notes: Constructed as the average CPI of CHN, RUS, & EU
  - Wages
    - Source: Haver
    - Coverage: ARM, AZE, GEO, KAZ, KGZ, TJK, UZB *
    - Observations: 476
    - Mean: 15.67
    - Standard Deviation: 11.49
    - Notes: Annualized quarterly growth rate of nominal average wages in each country
  - Inflation Expectations (t+3)
    - Source: IMF WEO vintages (1999-2019), Authors' calculations
    - Coverage: ARM, AZE, GEO, KAZ, KGZ, TJK, UZB
    - Observations: 616
    - Mean: 6.70
    - Standard Deviation: 11.74
    - Notes: t +3 inflation forecasts from period t WEO are used as proxy for inflation expectations
  - Policy Rate
    - Source: IMF MFS (STA)
    - Coverage: ARM, AZE, GEO, KAZ, KGZ, TJK, UZB *
    - Observations: 482
    - Mean: 8.50
    - Standard Deviation: 4.47
    - Note: * Limited coverage for UZB.
- Note: All variables are gathered in quarterly frequency and, with the exception of output gap, are expressed as year-on-year percent change to mitigate seasonality and non-stationarity issues.
- Repeated note: Turkmenistan was excluded from the empirical analysis due to data limitations.

### Inflation decomposition approach (ANNEX II)
- Statistical framework:
  - Equation estimated:
    - π_ijt = λ_ij f_t + μ_ij g_jt + e_ijt
    - Where:
      - π_ijt is the standardized annual change of the consumer price index in country i and region j (subtracting the mean and dividing by the standard deviation).
      - f_t is the global factor.
      - g_jt is the regional factor in region j.
      - e_ijt is the country-specific component.
      - Parameters λ_ij and μ_ij may differ across countries.
- Sample and data frequency:
  - Sample: G20 economies and countries in the Caucasus and Central Asia (CCA) region.
  - Period: Monthly data since January 2017 until May 2024.
  - Focus: Headline inflation.
  - Regional definition: The region is defined as the CCA and Russia; the remaining countries constitute the global economy.
- Estimation steps:
  1. Obtain principal components for inflation of the whole sample of countries to extract the global component.
  2. Obtain principal components from the residuals of step 1 (i.e., excluding the global component) for the CCA countries and Russia to extract the regional factor.
  3. Regress inflation in each country on the estimated global and regional components; the regression residual yields the country-specific component.
- Methodological note:
  - This principal component decomposition captures correlations between time series and provides a useful description of inflation dynamics, but it does not by itself explain the transmission mechanisms of inflation.

*Source: wpiea2025003-print-pdf - References*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025003-print-pdf.pdf_