## When Policy Bites: State-Dependent Monetary Policy Transmission in Emerging Markets

## Source details

**Canonical URL:** [When Policy Bites: State-Dependent Monetary Policy Transmission in Emerging Markets](https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026096-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2026/english/wpiea2026096-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2026/english/wpiea2026096-source-pdf.pdf.json)

---

### Overview and structure
- Organized sections and page references include: Introduction (page 5); I. Literature Review (page 7); II. Data and Monetary Policy Shock Construction (page 9); III. Transmission in the Linear Model (page 11); IV. State-Dependent Transmission (page 14); V. Robustness (page 25); VI. Conclusion (page 27); Annex I–IV (pages 29–34).
- Key thematic focus: state-dependent transmission along three dimensions — business cycle position, trend inflation level, and monetary policy stance — using a panel of EMs and analysis of consumption, investment, credit, REER, and net exports as channels.

### Motivation, scope, and sample
- Motivation:
  - Recent post-COVID rebound in output and global inflation raises questions about whether monetary policy effects under high inflation and rapid tightening mirror historical patterns.
  - State-dependence and nonlinearities challenge linear frameworks; existing empirical evidence is concentrated in AEs, leaving EM evidence underdeveloped.
- Scope and objectives:
  - Comprehensive analysis of state-dependent transmission of monetary policy to output and consumer prices in EMs.
  - Focus on three cyclical conditions: business cycle position; level of trend inflation; monetary policy stance.
- Sample and period:
  - Panel of eleven large EM economies with inflation-targeting and freely floating exchange rates: Brazil, Chile, Colombia, Hungary, India, Indonesia, Mexico, Peru, Poland, Russia, Thailand.
  - Sample period: 1996Q1 to 2025Q1, resulting in 763 observations.
  - Crisis indicator includes Laeven and Valencia (2020) crisis years and the four quarters of 2020.

### Identification strategy and methodology
- Shock identification:
  - Monetary policy shocks constructed from private-sector forecast errors (Consensus Economics): residuals from country-level regressions that purge short-term rate forecast errors of systematic components.
  - Regression controls include: one-year-ahead forecasts of y-o-y GDP growth and inflation and their first differences (with two lags); two lags of the short-term rate; lagged quarterly changes in the NEER; contemporaneous and lagged global commodity price changes; two lags of the term spread; crisis dummy including 2020Q1–2020Q4.
  - Residual ε(j,t) is labeled shcoc(j,t) and used as the monetary policy shock.
  - Cross-country pairwise correlations of estimated shocks concentrated mostly between -0.2 and 0.2; average correlation very close to zero.
- Estimation:
  - Panel local projections in long differences over horizons h = 1 to 12 quarters.
  - Responses normalized to the dynamic impact of an unexpected 100-basis-point (bp) increase in the short-term rate.
  - Dependent variables: cumulative log-differences; specification includes current and 3 lags of the monetary policy shock, quarterly changes in the outcome variable, crisis indicator, time fixed effects, country fixed effects.
  - Standard errors clustered at the country level.
- State-dependent specifications:
  - Three binary state indicators: business cycle (5-quarter moving average of output gap > 0 for expansion, < 0 for recession); trend inflation (two-sided HP trend above/below EM-sample median of 4.8 percent); monetary policy stance (interest rate gap relative to 20-quarter moving average: tight if gap > 0, loose if gap < 0).
  - Robustness checks use continuous smooth transition functions with logistic F(z) and κ parameters (κ = 3 for business cycle; κ = 5 for trend inflation; κ = 2 for policy stance).

### Construction of monetary policy shocks (specification highlights)
- For each country j and quarter t:
  - Interest-rate forecast error (realized end-of-quarter short-term rate minus forecast made at end of previous quarter) regressed on the controls listed under Identification.
  - Residual ε(j,t) = shcoc(j,t) is used as the monetary policy shock.
- Cross-country correlations:
  - Average pairwise correlations are very close to zero; most bilateral correlations lie between -0.2 and 0.2.
  - Selected cross-country correlation entries (row BRA shown): BRA–CHL 0.14; BRA–COL 0.10; BRA–HUN -0.19; BRA–IDN 0.33; BRA–IND -0.11; BRA–RUS 0.24.

### Key empirical findings — linear (state-independent) model
- A 100-bp tightening shock persistently reduces real GDP; impact after 12 quarters reaches approximately 0.6 percent.
- In the linear specification, transmission to prices is limited; core CPI and headline CPI evidence of significant price responses is weak (analysis focuses on core CPI in subsequent state-dependent analysis).

### Key empirical findings — state-dependent transmission (summary)
- General:
  - Considerable state-dependence across all three cyclical dimensions.
  - GDP responses are echoed by consumption, investment, and credit responses.
  - REER appreciation following tightening is more persistent in states with larger negative output responses; net exports responses are often muted.
- Reported magnitudes and statistical notes:
  - Core inflation: a tightening shock reduces core inflation by 0.3 percent over the same horizon, statistically significant only at the 68 percent confidence level; headline inflation impact not significant.
  - Real credit: peak contraction about -1.6 percent toward the end of the 12-quarter horizon (baseline statement); in the loose-stance state, real credit declines by 3 percent at the end of the projection horizon.
  - Comparison with AEs: transmission to output and prices in the EM sample is very close to a benchmark AE sample.
- Recession versus Expansion:
  - Monetary policy has strong effects on GDP during recessions but no statistically significant effects during expansions in some specifications.
  - Peak output effects are twice as large during recessions compared to expansions.
  - In detailed results: monetary policy shocks reach a peak impact of about -0.8 percent on output after 12 quarters during recessions.
  - Differences across states become statistically significant from the sixth quarter.
  - Core CPI: no significant evidence of transmission during recessions; only mild evidence during expansions (68-percent confidence level).
- Low versus High Trend Inflation:
  - Output declines in both states but relatively more in the low-inflation state, reaching a peak decline of -1 percent in quarter 12.
  - Core CPI falls strongly in high trend inflation states, peaking at about -1.2 percent in quarter 11.
  - In the low-inflation state, a small price puzzle lasts 6 quarters (prices increase significantly following the tightening shock).
- Loose versus Tight Monetary Policy Stance:
  - Output reacts much more strongly after a period of loose monetary policy in the medium term; output barely reacts following a tight stance.
  - Difference in output responses becomes significant from the sixth quarter.
  - Peak impact on output is about twice as strong in the loose-stance state compared with the linear model.
  - Real credit declines by 3 percent at the end of the projection horizon in the loose-stance state; response is insignificant in the tight-stance state.
- Exchange rate and trade:
  - REER initially appreciates, peaking at 100 bp in the second quarter after the shock, before normalizing.
  - Net exports barely respond on average; trade channel limited on average in EMs.
  - State-specific REER dynamics noted (e.g., real depreciation reaching -2 percent until quarter 10 in the high-inflation state in relative comparisons).

### Transmission channels and dynamics
- Domestic demand:
  - Consumption and investment decline almost monotonically after a tightening shock; investment shows a sharper and more delayed contraction.
  - In low-inflation states, investment eventually falls twice as much as consumption.
- Credit channel:
  - Real credit contracts more than output; baseline discussion notes credit drop nearly three times as large as the drop in output, indicating credit supply effects.
  - Credit declines faster and stronger during recessions than in expansions (differences not always statistically significant).
- Exchange rate and trade:
  - REER appreciates by ~100 bp on impact, consistent with UIP and portfolio rebalancing channels.
  - Initial REER appreciation yields a small temporary drop in exports; later import reductions from domestic demand decline occur in year 3.
  - Net exports average response limited; state-dependent patterns can alter net export behavior, sometimes significant by quarter 12.

### Theoretical interpretation and literature alignment
- Mechanisms consistent with empirical results:
  - Financial frictions and occasionally binding borrowing constraints amplify output effects in downturns.
  - Nominal rigidities, state-dependent pricing, and loss-aversion preferences can amplify output responses during recessions.
  - Higher trend inflation steepens the Phillips Curve, amplifying inflation responses to monetary policy.
  - Loose prior policy stance can raise risk-taking and balance-sheet vulnerabilities, increasing the potency of subsequent tightening.
- Empirical comparisons:
  - Stronger output effects during recessions align with recent U.S. studies; AE cross-country studies show mixed findings, with some reporting stronger transmission during expansions and no price state dependence.

### Robustness and sensitivity
- Continuous-state estimation using smooth transition functions confirms main results and can amplify contrasts (e.g., recession peak effect on output of -1.7 percent under a smooth transition specification).
- Changing κ values for the smooth transition does not materially affect main conclusions; sensitivity checks reported (κ = 1 or 3 for policy stance discussed).

### Policy-relevant implications
- Monetary policy effectiveness in EMs is state-dependent across business cycle position, trend inflation level, and prior policy stance.
- Policymakers should account for:
  - Stronger output amplification of policy in recessions and following prolonged accommodation (loose stance) when calibrating tightening or easing.
  - Larger nominal price responses in high trend inflation environments, implying different trade-offs across inflation regimes.
  - Financial vulnerabilities (leverage, credit conditions) that affect transmission; macroprudential and financial-stability tools can influence monetary policy potency.

### Annex I — Data sources, availability, and summary statistics (selected)
- Availability of monetary policy shocks by country (Start Date to End Date) for the 11 EMs:
  - Brazil — 2001q3 to 2024q1
  - Chile — 2005q1 to 2024q1
  - Colombia — 2016q2 to 2024q1
  - Hungary — 1999q3 to 2024q1
  - India — 2002q3 to 2024q1
  - Indonesia — 2014q1 to 2024q1
  - Mexico — 2002q1 to 2024q1
  - Peru — 2016q2 to 2024q1
  - Poland — 1999q4 to 2024q1
  - Russia — 2016q3 to 2024q1
  - Thailand — 2002q4 to 2024q1
- Key summary statistics (sample-level moments; selected variables):
  - Real GDP (yoy change, percent): Mean 2.4; Standard Deviation 3.6; 10th Percentile -0.9; 50th Percentile 2.4; 90th Percentile 5.5
  - CPI (yoy change, percent): Mean 2.9; Standard Deviation 2.8; 10th Percentile 0.2; 50th Percentile 2.3; 90th Percentile 6.3
  - Core CPI (yoy change, percent): Mean 2.4; Standard Deviation 2.3; 10th Percentile 0.4; 50th Percentile 1.8; 90th Percentile 5.1
  - Short-term rate (percent): Mean 3.3; Standard Deviation 3.4; 10th Percentile 0.0; 50th Percentile 2.6; 90th Percentile 7.4
  - Commodity prices (yoy change, percent): Mean 7.5; Standard Deviation 25.3; 10th Percentile -23.5; 50th Percentile 4.4; 90th Percentile 45.9
  - Real credit (yoy change, percent): Mean 3.8; Standard Deviation 6.1; 10th Percentile -2.9; 50th Percentile 3.4; 90th Percentile 10.7
  - Monetary policy shock (percent): Mean 0.0; Standard Deviation 0.4; 10th Percentile -0.4; 50th Percentile 0.0; 90th Percentile 0.4

*IMF Working Paper — When Policy Bites: State-Dependent Monetary Policy Transmission in Emerging Markets (Introduction; Annex I).*

### Introduction ...........................................................................................................

### Introduction

### Overview and Structure
- The document is organized into major sections with page references:
  - Introduction (page 5)
  - I. Literature Review (page 7)
  - II. Data and Monetary Policy Shock Construction (page 9)
  - III. Transmission in the Linear Model (page 11)
  - IV. State-Dependent Transmission (page 14)
  - V. Robustness (page 25)
  - VI. Conclusion (page 27)
  - Annex I. Data Sources and Transformations (page 29)
  - Annex II. Computation of 1-year ahead forecasts (page 32)
  - Annex III. Summary Statistics (page 33)
  - Annex IV. Cross-Country Correlations of Monetary Policy Shocks (page 34)

### Key Thematic Areas (section-level)
- I. Literature Review
  - Theoretical Foundations of State-Dependent Monetary Policy Transmission
  - Empirical Evidence: State-Dependent Output and Price Responses
  - Monetary Policy Transmission in Emerging Markets
- II. Data and Monetary Policy Shock Construction
  - Sample and Data Sources
  - Construction of Monetary Policy Shocks
- III. Transmission in the Linear Model
- IV. State-Dependent Transmission
  - Definitions of States and Distribution of Monetary Policy Shocks across States
  - Estimation of State-dependence in Transmission
  - Recession versus Expansion
  - Low versus High Trend Inflation
  - Loose versus Tight Monetary Policy Stance
- V. Robustness
- VI. Conclusion

### Figures (enumerated)
- 1. Impulse Responses to a 100-bp Monetary Policy Shock of Output and Prices
- 2. Impulse Responses to a 100-bp Monetary Policy Shock of Consumption, Investment, Real Credit, REER and Net Exports in EMs
- 3. Distribution of Monetary Policy Shocks across States
- 4. Impact of Monetary Policy Shocks on Output and Core CPI depending on the State of Business Cycle
- 5. Impact of Monetary Policy Shocks on Consumption, Investment, Real Credit, REER, and Net Exports depending on the State of the Business Cycle
- 6. Impact of Monetary Policy Shocks on Output and Core CPI depending on the State of Trend Inflation
- 7. Impact of Monetary Policy Shocks on Consumption, Investment, Real Credit, REER, and Net Exports depending on the State of Trend Inflation
- 8. Impact of Monetary Policy Shocks on Output and Core CPI depending on the Initial Monetary Policy Stance
- 9. Impact of Monetary Policy Shocks on Consumption, Investment, Net Exports, Real Credit, REER and Net Exports depending on the Initial Monetary Policy Stance
- 10. Impact of Monetary Policy Shocks on Output and Core CPI depending on the State of Business Cycle based on a Smooth Transition Function
- 11. Impact of Monetary Policy Shocks on Output and Core CPI depending on the State of Trend Inflation based on a Smooth Transition Function
- 12. Impact of Monetary Policy Shocks on Output and Core CPI depending on the Initial Monetary Policy Stance based on a Smooth Transition Function
- Annex Figure IV 1. Distribution of the Cross-Country Correlations of the Monetary Policy Shocks

### Tables (enumerated)
- Table 1. Correlations between Types of States (High vs. Low)
- Annex Table I.1. Availability of Monetary Policy Shocks by Country in the Sample
- Annex Table I.2. Data Sources
- Annex Table I.3. Definition of Short-term Rates by Country in the Sample
- Annex Table III.1. Summary Statistics
- Annex Table IV.1. Cross-Country Correlation Matrix of Monetary Policy Shocks

### Acronyms included in the unit
- AE  Advanced Economy
- CPI Consumer Price Index
- EM Emerging Market
- EMDE Emerging Market and Developing Economy
- GDP Gross Domestic Product
- NEER Nominal Effective Exchange Rate
- REER Real Effective Exchange Rate
- VAR  Vector Autoregression

*When Policy Bites: State-Dependent Monetary Policy Transmission in Emerging Markets — IMF Working Papers (Introduction section).*

### Introduction

### Introduction

### Motivation and research gap
- The sharp rebound in output and the global surge in inflation after the COVID-19 pandemic have raised questions about whether monetary policy effects in the recent high-inflation, fast-tightening environment mirror those observed historically.
- Potential state-dependence and other nonlinearities in inflation and policy transmission challenge traditional linear frameworks and motivate a more nuanced understanding of monetary policy transmission under different economic conditions.
- Empirical literature on state-dependence has mostly focused on advanced economies (AEs); evidence for emerging market economies (EMs) remains underdeveloped and concentrated on structural characteristics, leaving a gap for policy-relevant studies in EMs.

### Scope and objectives
- The paper provides a comprehensive analysis of state-dependent transmission of monetary policy to output and consumer prices in EMs.
- Focus on three cyclical conditions: the business cycle position, the level of trend inflation, and the monetary policy stance.
- Panel of eleven large EM economies with an inflation-targeting mandate and a freely floating exchange rate regime is used to minimize heterogeneity in monetary policy and exchange rate frameworks and to benchmark against AE studies.
- Additional real macroeconomic variables analyzed as channels: consumption, investment, credit, real effective exchange rate (REER), and net exports.

### Identification strategy and methodology
- High-frequency identification is limited in EMs; instead the paper exploits private sector macroeconomic forecasts compiled by Consensus Economics to construct orthogonalized forecast errors as exogenous variation in monetary policy for the eleven EMs.
- Monetary policy shocks are constructed as residuals from country-level regressions that purge short-term rate forecast errors of systematic components (including GDP growth and inflation forecasts and their revisions, lagged macro indicators, term spread, nominal effective exchange rate changes, global commodity prices, and crisis dummies).
- Panel local projections in long differences are implemented to estimate impulse responses over a 12-quarter horizon in both linear (state-independent) and three state-dependent settings (business cycle position, trend inflation level, monetary policy stance).
- Dependent variables in local projections are defined as cumulative log-differences; specification includes current and 3 lags of the monetary policy shock, quarterly changes in the outcome variable, crisis indicator, time fixed effects, and country fixed effects. Standard errors clustered at the country level.

### Data and sample
- Sample restricted to 11 large EMs meeting two criteria: explicit inflation target as primary policy objective and a floating exchange rate arrangement according to the 2024 vintage of the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions (IMF 2024); and availability of private sector macro forecasts to construct shocks for at least 32 quarters.
- Sample period: 1996Q1 to 2025Q1, resulting in 763 observations.
- Macroeconomic data sources: national statistical agencies or central banks via Haver Analytics, IMF International Financial Statistics, and the BIS. Private-sector forecasts from Consensus Economics.
- The 11 emerging market economies in the sample: Brazil, Chile, Colombia, Hungary, India, Indonesia, Mexico, Peru, Poland, Russia and Thailand.
- Crisis indicator includes years of banking, currency, and sovereign debt crises from Laeven and Valencia (2020) and the four quarters of 2020 (the first year of the COVID-19 pandemic).

### Construction of monetary policy shocks (specification highlights)
- For each country j and quarter t, the interest rate forecast error (difference between realized short-term rate at end of quarter t and forecast made at end of previous quarter) is regressed on:
  - One-year-ahead forecasts of y-o-y GDP growth and inflation and their first differences, with two lags.
  - Two lags of the short-term interest rate to capture persistence.
  - Lagged quarterly changes in the nominal effective exchange rate.
  - Contemporaneous and lagged y-o-y changes in global commodity prices.
  - Two lags of the term spread (difference between 10-year government bond yield and short-term policy rate).
  - Crisis dummy capturing financial crises and 2020Q1–2020Q4.
- The residual from this country-level regression, ε(j,t), is labeled the monetary policy shock, shcoc(j,t), and used in subsequent analysis.
- Cross-country pairwise correlations of estimated shocks show an average correlation very close to zero and most bilateral correlations concentrated between -0.2 and 0.2, supporting the interpretation that shocks are idiosyncratic rather than global.

### Key empirical findings — linear model (state-independent)
- Local projections estimate impulse responses for horizons h = 1 to 12 quarters.
- Results normalized to the dynamic impact of an unexpected 100-basis-point (bp) increase in the short-term rate.
- A 100-bp tightening shock persistently reduces real GDP, with the impact after 12 quarters reaching approximately 0.6 percent.
- Core CPI and headline CPI evidence in the linear model: statistical significance of transmission to prices is limited in the linear specification (the paper focuses on core CPI in subsequent state-dependent analysis).

### Key empirical findings — state-dependent transmission
- Considerable state-dependence of monetary policy transmission in EMs along all three cyclical dimensions studied:
  - Business cycle position:
    - Monetary policy has strong effects on gross domestic product (GDP) during recessions but no statistically significant effects during expansions.
    - Peak output effects are twice as large during recessions compared to expansions.
    - Wald tests indicate the decline of core consumer prices is initially stronger during expansions following a tightening shock, but neither price response is significant at the 90 percent confidence level.
  - Trend inflation:
    - Monetary policy has a stronger impact on output during low trend inflation periods, particularly in the short term.
    - Impact on prices is much more pronounced during high trend inflation periods at all horizons.
  - Monetary policy stance:
    - Output reacts much more strongly to monetary policy shocks after a period of loose monetary policy in the medium term.
    - Price responses remain largely insensitive to the stance.
- Across all types of state dependence, GDP responses are echoed by consumption, investment, and credit responses.
- Real exchange rate (REER) appreciation following a tightening is more persistent in states associated with larger negative output responses (recessions, low trend inflation, loose stance), potentially reflecting expected demand compression effects on the trade balance.
- Net exports responses are often muted and, in state-dependent models, their dynamics are sometimes difficult to reconcile with REER and domestic demand responses.

### Theoretical interpretation and alignment with literature
- Results broadly support theories emphasizing:
  - Financial frictions such as occasionally binding borrowing constraints.
  - Nominal rigidities, state-dependent pricing, and loss-aversion preferences that amplify output responses in downturns.
  - The role of trend inflation in price-setting behavior, where higher trend inflation steepens the Phillips Curve and amplifies inflation responses to monetary policy.
  - The monetary policy stance affecting risk-taking, balance sheets, and the likelihood that tightening pushes borrowers into constrained regions after prolonged loose policy.
- Empirical comparisons:
  - Stronger output effects during recessions align with recent U.S. studies (De Santis and Tornese 2024; Piger and Stockwell 2025b).
  - Cross-country AE studies have mixed findings; some earlier AE cross-country studies find stronger transmission during expansions and no price state dependence (Jordà et al. 2020; Alpanda et al. 2021), differing from the EM results here.
  - Evidence on trend inflation and stance state dependence is scarce; some AE studies find different patterns (e.g., Jordà et al. 2020, Ascari and Haber 2022), highlighting differences between AE and EM contexts.

### Contribution and implications
- The paper fills a gap by providing a comprehensive, cross-country EM analysis of state-dependent monetary policy transmission using a homogeneous sample (inflation targeting, floating exchange rates).
- Findings call for attention to state-dependence in macroeconomic modeling and monetary policymaking in EMs.
- The analysis informs policymakers that the effectiveness of monetary tightening on output and inflation depends markedly on cyclical conditions (business cycle, trend inflation, prior stance) and that channels—consumption, investment, credit, and REER—play important roles while net exports often provide limited offsetting dynamics.

*IMF Working Paper — Introduction section*

### 0.3 percent over the same horizon, although this effect is statistically significant only at the 68 percent

### When Policy Bites: State-Dependent Monetary Policy Transmission in Emerging Markets

### Key Empirical Findings
- A tightening monetary policy shock reduces core inflation by 0.3 percent over the same horizon, although this effect is statistically significant only at the 68 percent confidence level. The impact on headline inflation is not significant.
- Real credit to the private sector contracts with a peak effect of about -1.6 percent towards the end of the projection horizon.
- Monetary policy transmission to output and prices in the EM sample is very close to that in a benchmark AE sample.
- State-dependent results:
  - Recession versus Expansion:
    - Monetary policy shocks have a stronger impact on output during recessions, reaching a peak impact of about -0.8 percent after 12 quarters.
    - Transmission to output is muted during expansions but remains significant; differences across states become statistically significant from the sixth quarter.
    - Core CPI: no significant evidence of transmission during recessions; only mild evidence (at the 68-percent confidence level) during expansions.
  - Low versus High Trend Inflation:
    - Output declines in both states but relatively more in the low-inflation state, reaching a peak decline of -1 percent in quarter 12.
    - Core CPI falls strongly in high trend inflation states, peaking at about -1.2 percent in quarter 11.
    - In the low-inflation state, a small price puzzle lasts 6 quarters (prices increase significantly following the tightening shock).
  - Loose versus Tight Monetary Policy Stance:
    - Output contracts when the shock follows a period of loose monetary policy but barely reacts when following a tight stance; difference becomes significant from the sixth quarter.
    - Peak impact on output is about twice as strong in the loose-stance state compared with the linear model.
    - Real credit declines by 3 percent at the end of the projection horizon in the loose-stance state while its response is insignificant in the tight-stance state.
- Exchange rate and trade:
  - The REER initially appreciates, peaking at 100 bp in the second quarter after the shock, before normalizing.
  - Net exports barely respond on average; trade channel is limited on average in EMs.
  - REER dynamics differ across states: e.g., real depreciation reaching -2 percent until quarter 10 in the high-inflation state (relative behavior described).

### Definitions, Identification, and Methodology
- Monetary policy shocks are constructed from forecast errors made by professional forecasters; shocks are the regression residuals from Equation (1).
- Baseline impulse responses are to a 100-bp monetary policy tightening shock over 12 quarters, with 68 and 90 percent confidence intervals, estimated via panel local projections with robust standard errors (Equation (2)).
- State definitions (binary indicators):
  - Business cycle: expansion if 5-quarter moving average of the output gap > 0; recession if < 0. Output gap computed using one-sided HP filter with 4-quarter-ahead forecasts and 5-quarter moving average smoothing.
  - Trend inflation: high (low) when two-sided HP-filtered trend inflation is above (below) the EM-sample median trend inflation value of 4.8 percent.
  - Monetary policy stance: interest rate gap = ex-ante real short-term rate (nominal short-term rate minus one-year-ahead inflation expectations) minus its 20-quarter moving average; stance is tight if gap > 0 and loose if gap < 0.
- Correlations between the three state indicators are low (Table 1):
  - Business cycle vs. Trend inflation: -0.004
  - Business cycle vs. Monetary policy stance: 0.12
  - Trend inflation vs. Monetary policy stance: 0.02
- Robustness: alternative continuous state transition functions using logistic transformation F(z) with κ parameters (κ = 3 for business cycle; κ = 5 for trend inflation; κ = 2 for policy stance). Results confirm state-dependence; continuous approach yields an especially large recession peak effect on output of -1.7 percent.

### Transmission Channels and Dynamics
- Domestic demand channels:
  - Consumption and investment decline almost monotonically after a tightening shock; investment shows a sharper contraction and a more delayed response.
  - In low-inflation states, investment eventually falls twice as much as consumption.
- Credit channel:
  - Real credit contracts more than output (credit drop nearly three times as large as the drop in output in baseline discussion), suggesting credit supply effects and presence of a credit channel in EMs.
  - Credit declines faster and stronger during recessions than in expansions (difference not always statistically significant).
- Exchange rate and trade channel:
  - REER appreciates ~100 bp on impact, consistent with uncovered interest parity and portfolio rebalancing channels.
  - Initial REER appreciation leads to a small temporary drop in exports; later domestic demand decline reduces imports in year 3.
  - Net exports show limited average response; state-specific dynamics can alter net export patterns (differences sometimes significant by quarter 12).

### Interpretation and Theoretical Consistency
- The stronger transmission to output during recessions and following loose monetary stance is consistent with financial frictions and occasionally binding borrowing constraints (amplification via investment and credit channels).
- Stronger price responses in high-inflation states and muted output responses in those states align with state-dependent pricing and rational inattention frameworks.
- The muted core CPI response during recessions despite larger real effects may reflect convexities in the Phillips curve and liquidity pressures keeping prices elevated (Gilchrist et al. 2017).
- Findings are also consistent with loss-averse preference models (Santoro et al., 2014) and other studies documenting persistent real effects of monetary policy.

### Robustness and Sensitivity
- Alternative continuous-state estimation (smooth transition) confirms main results and in some cases amplifies contrasts (e.g., recession peak effect on output at -1.7 percent).
- Changing κ values for the smooth transition does not materially affect results (sensitivity checks with κ = 1 or 3 for policy stance discussed).

### Policy-Relevant Implications
- Monetary policy effectiveness in EMs is state-dependent along three dimensions: business cycle position, trend inflation level, and prior policy stance.
- Policymakers should account for stronger output amplification of policy in recessions and following prolonged accommodation (loose stance) when calibrating policy tightening or easing.
- High trend inflation environments may yield larger nominal price responses to policy changes, suggesting different policy trade-offs across inflation regimes.
- Financial vulnerabilities (leverage, credit conditions) matter for transmission; macroprudential and financial-stability tools can influence monetary policy potency.

*Source: IMF Working Paper — When Policy Bites: State-Dependent Monetary Policy Transmission in Emerging Markets*

### Annex I.  Data Sources and Transformations

### Annex I.  Data Sources and Transformations

### Availability of Monetary Policy Shocks by Country in the Sample
- Country classification uses AE (Advanced Economy) and EM (Emerging Market).
- Availability (Start Date to End Date) of monetary policy shocks by country:
  - Brazil — EM — 2001q3 to 2024q1
  - Chile — EM — 2005q1 to 2024q1
  - Colombia — EM — 2016q2 to 2024q1
  - Hungary — EM — 1999q3 to 2024q1
  - India — EM — 2002q3 to 2024q1
  - Indonesia — EM — 2014q1 to 2024q1
  - Mexico — EM — 2002q1 to 2024q1
  - Peru — EM — 2016q2 to 2024q1
  - Poland — EM — 1999q4 to 2024q1
  - Russia — EM — 2016q3 to 2024q1
  - Thailand — EM — 2002q4 to 2024q1
  - Australia — AE — 1996q1 to 2024q1
  - Canada — AE — 1996q1 to 2024q1
  - Czech Republic — AE — 2001q1 to 2024q1
  - France — AE — 2000q1 to 2024q1
  - Germany — AE — 2000q1 to 2024q1
  - Italy — AE — 2000q1 to 2024q1
  - Japan — AE — 1996q1 to 2024q1
  - Korea — AE — 2001q2 to 2024q1
  - Netherlands — AE — 2000q1 to 2024q1
  - New Zealand — AE — 1996q1 to 2024q1
  - Norway — AE — 1998q3 to 2024q1
  - Slovak Republic — AE — 2009q1 to 2024q1
  - Spain — AE — 2000q1 to 2024q1
  - Sweden — AE — 1996q1 to 2024q1
  - Switzerland — AE — 2000q1 to 2024q1
  - United Kingdom — AE — 1997q3 to 2024q1
  - United States — AE — 1996q1 to 2024q1

### Data Sources (variables and original sources)
- GDP: Real gross domestic product — Haver Analytics; national statistical agencies
- CPI: Consumer price index — Haver Analytics; national statistical agencies
- Core CPI: Consumer price index excluding food and energy — Haver Analytics; national statistical agencies
- Short-term rate: Short-term interest rate corresponding to definitions of consensus forecasts (Annex Table I.3) — Haver Analytics, LSEG Datastream, Bloomberg; national central banks
- 10-year government bond yield: 10-year government bond yield — LSEG Datastream; national central banks
- Term spread: 10-year government bond yield minus short-term rate — Haver Analytics, LSEG Datastream, Bloomberg; national central banks
- NEER: Nominal effective exchange rate — Haver Analytics; IMF International Financial Statistics Database, BIS
- REER: Real effective exchange rate — Haver Analytics; IMF International Financial Statistics Database, BIS
- Commodity prices: S&P GSCI commodity index — Haver Analytics; S&P
- Bilateral exchange rate: Local currency per USD — Haver Analytics; BIS
- Crisis dummy: Dummy variable indicating a systemic banking, currency or sovereign debt crisis or the year 2020 (first year of the COVID-19 pandemic) — Laeven and Valencia (2020) database
- Consumption: Real domestic consumption — Haver Analytics; IMF International Financial Statistics Database and national statistical agencies
- Investment: Real investment — Haver Analytics; IMF International Financial Statistics Database and national statistical agencies
- Net exports: Real net exports — Haver Analytics; IMF International Financial Statistics Database and national statistical agencies
- Real credit: Credit to private non-financial sector from all sectors deflated by CPI — Haver Analytics; BIS

### Definition of Short-term Rates by Country in the Sample
- Short-term rate series used for each country:
  - Brazil — Overnight Interbank Interest Rate, SELIC (%)
  - Chile — Central Bank Monetary Policy Rate (%)
  - Colombia — Central Bank Policy Rate (%)
  - Hungary — 3-month Treasury Bill Rate (%)
  - India — 91-day Treasury Bill Rate
  - Indonesia — 3-month Deposit Rate (%)
  - Mexico — 28-day CETES Rate (%)
  - Peru — Monetary Policy Interest Rate (%)
  - Poland — 3-month Interbank Deposit Rate (%)
  - Russia — Central Bank Monetary Policy Rate (%)
  - Thailand — 3-month Interbank Rate (%)
  - Australia — 90-day Dealer Bill Rate (%)
  - Canada — 3-month Treasury Bill Rate (%)
  - Czech Republic — 3-month PRIBOR Interbank Deposit Rate (%)
  - Euro Area 1/ — 3-month Euro Rate (%)
  - France — 3-month Euro Rate (%)
  - Germany — 3-month Euro Rate (%)
  - Italy — 3-month Euro Rate (%)
  - Japan — 3-month Yen Certificate of Deposit
  - Korea — 91-day Certificate of Deposit (%)
  - Netherlands — 3-month Euro Rate (%)
  - New Zealand — 90-day Bank Bill Rate (%)
  - Norway — 3-month Interbank Rate (%)
  - Slovak Republic — 3-month Euro Rate (%)
  - Spain — 3-month Euro Rate (%)
  - Sweden — 3-month Interbank Rate (%)
  - Switzerland — 3-month Swiss Av. Rate (SAR3M) (%)
  - United Kingdom — 3-month Interbank Rate (%)
  - United States — 3-month Treasury Bill Rate (%)
- Note: Euro-area short-term rates are used to compute monetary policy shocks for euro-area countries (France, Germany, Italy, The Netherlands, Slovak Republic, Spain).

### Computation of 1-year ahead forecasts (method)
- Data source: Consensus Economics monthly forecasts for the current calendar year (y) and following calendar year (y+1) for real GDP growth and inflation.
- The 1-year-ahead forecast for cumulative real GDP over the previous 12 months F_{y,C}(GDP_{y+1,C}) and the 1-year ahead forecast for CPI F_{y,C}(CPI_{y+1,C}) are computed as weighted averages:
  - F_{y,C}(GDP_{y+1,C}) = w_C × (GDP forecast for calendar year y) + (1 − w_C) × (GDP forecast for calendar year y+1)
  - F_{y,C}(CPI_{y+1,C}) = w_C × (CPI forecast for calendar year y) + (1 − w_C) × (CPI forecast for calendar year y+1)
- The weight w_m is determined by the month m (Jan = 1, Feb = 2, etc.) in which the forecast is made:
  - w_m = (13 − m)/12
- These values are then used to compute forecasts of year-on-year GDP growth and CPI inflation for each quarter.

### Summary Statistics (sample-level moments)
- Variables and summary statistics (Mean, Standard Deviation, 10th Percentile, 50th Percentile, 90th Percentile):
  - Real GDP (yoy change, percent): Mean 2.4; Standard Deviation 3.6; 10th Percentile -0.9; 50th Percentile 2.4; 90th Percentile 5.5
  - CPI (yoy change, percent): Mean 2.9; Standard Deviation 2.8; 10th Percentile 0.2; 50th Percentile 2.3; 90th Percentile 6.3
  - Core CPI (yoy change, percent): Mean 2.4; Standard Deviation 2.3; 10th Percentile 0.4; 50th Percentile 1.8; 90th Percentile 5.1
  - Short-term rate (percent): Mean 3.3; Standard Deviation 3.4; 10th Percentile 0.0; 50th Percentile 2.6; 90th Percentile 7.4
  - 10-year government bond yield (percent): Mean 4.0; Standard Deviation 2.4; 10th Percentile 0.8; 50th Percentile 3.9; 90th Percentile 7.2
  - Term spread (percent): Mean 0.7; Standard Deviation 1.9; 10th Percentile -0.9; 50th Percentile 0.9; 90th Percentile 2.6
  - NEER (yoy change, percent): Mean 0.3; Standard Deviation 6.9; 10th Percentile -7.5; 50th Percentile 0.6; 90th Percentile 7.2
  - REER (yoy change, percent): Mean 0.3; Standard Deviation 6.9; 10th Percentile -7.3; 50th Percentile 0.3; 90th Percentile 7.6
  - Commodity prices (yoy change, percent): Mean 7.5; Standard Deviation 25.3; 10th Percentile -23.5; 50th Percentile 4.4; 90th Percentile 45.9
  - Consumption (yoy change, percent): Mean 2.3; Standard Deviation 3.3; 10th Percentile -0.5; 50th Percentile 2.2; 90th Percentile 5.4
  - Investment (yoy change, percent): Mean 2.3; Standard Deviation 9.5; 10th Percentile -8.6; 50th Percentile 2.7; 90th Percentile 11.8
  - Imports (yoy change, percent): Mean 4.4; Standard Deviation 9.0; 10th Percentile -5.5; 50th Percentile 4.5; 90th Percentile 13.8
  - Exports (yoy change, percent): Mean 4.1; Standard Deviation 8.2; 10th Percentile -3.7; 50th Percentile 3.9; 90th Percentile 12.4
  - Net exports (yoy change, percent of t-1 GDP): Mean 0.04; Standard Deviation 1.8; 10th Percentile -1.7; 50th Percentile 0.0; 90th Percentile 1.8
  - Real credit (yoy change, percent): Mean 3.8; Standard Deviation 6.1; 10th Percentile -2.9; 50th Percentile 3.4; 90th Percentile 10.7
  - Monetary policy shock (percent): Mean 0.0; Standard Deviation 0.4; 10th Percentile -0.4; 50th Percentile 0.0; 90th Percentile 0.4

### Cross-Country Correlations of Monetary Policy Shocks
- Cross-country correlation matrix of monetary policy shocks (selected entries):
  - Row and column order: BRA CHL COL HUN IDN IND MEX PER POL RUS THA
  - BRA correlations: BRA 1.00; CHL 0.14; COL 0.10; HUN -0.19; IDN 0.33; IND -0.11; MEX -0.05; PER 0.09; POL 0.00; RUS 0.24; THA 0.10
  - CHL correlations: CHL 1.00; COL 0.05; HUN -0.15; IDN -0.04; IND -0.08; MEX -0.03; PER 0.20; POL 0.27; RUS 0.42; THA 0.12
  - COL correlations: COL 1.00; HUN 0.23; IDN 0.05; IND -0.04; MEX -0.08; PER -0.36; POL 0.01; RUS -0.04; THA -0.07
  - HUN correlations: HUN 1.00; IDN -0.08; IND -0.08; MEX -0.30; PER -0.16; POL 0.04; RUS -0.35; THA -0.04
  - IDN correlations: IDN 1.00; IND -0.04; MEX 0.30; PER 0.22; POL 0.11; RUS -0.09; THA -0.02
  - IND correlations: IND 1.00; MEX 0.12; PER -0.01; POL -0.24; RUS -0.29; THA 0.17
  - MEX correlations: MEX 1.00; PER 0.24; POL -0.10; RUS 0.14; THA -0.01
  - PER correlations: PER 1.00; POL 0.04; RUS 0.17; THA -0.03
  - POL correlations: POL 1.00; RUS 0.06; THA -0.11
  - RUS correlations: RUS 1.00; THA -0.17
  - THA correlations: THA 1.00

*When Policy Bites: State-Dependent Monetary Policy Transmission in Emerging Markets — Annex I. Data Sources and Transformations*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026096-source-pdf.pdf_