## 2. Data and Monetary Policy Shocks Identification

## Source details

**Canonical URL:** [2. Data and Monetary Policy Shocks Identification](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023204-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023204-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023204-print-pdf.pdf.json)

---

### Data
- Primary approach: construct monetary policy surprises as forecast errors in short-term rates using Consensus Forecasts (CF) data (end-of-month short-term interest rate forecasts).
- Sample and realized series:
  - Sample comprises 33 economies (advanced and emerging).
  - Realized short-term interest rates retrieved from country-specific CF reports when available and from HAVER, CEIC and Bloomberg.
  - Realized quarterly inflation, real GDP growth rates, and nominal effective exchange rates from HAVER.
- Forecasts used:
  - CF monthly, one-year ahead, median forecasts for short-term interest rates, inflation, and real GDP growth.
- Frequency alignment and non-overlapping horizons:
  - CF forecasts are 3 months ahead and end-of-month (e.g., December 2021 vintage forecasts March 2022 end-of-month).
  - Quarterly frequency constructed by taking CF vintages from December, March, June, and September for each year-quarter and comparing them with end-of-month (quarter) realized interest rates for March, June, August, and December, respectively, producing quarterly forecast errors without overlapping distortions.
- Auxiliary datasets for heterogeneity analysis:
  - Exchange rate regime classification from Ilzetzki, Reinhart and Rogoff (2019).
  - IMF Financial Development Index Database.
  - Quarterly uncertainty indexes from Ahir, Bloom and Furceri (2022).
  - Central bank transparency measure from Dincer, Eichengreen and Geraats (2022).
  - Annual cyclically adjusted primary balance from IMF.
  - Financial condition indexes from IMF GFSR.
- Euro Area treatment:
  - CF provides the European Central Bank's unified policy rate and country-specific macro forecasts, allowing identification of shocks exogenous to individual EA country macro conditions and inclusion of Austria, Belgium, Portugal, and Ireland from their EA entry dates.

### Monetary policy shocks identification procedure
- Two-step identification:
  1. Construct quarterly interest rate forecast errors FE_{j,t} = realized short-term interest rate − one-quarter-ahead end-of-month CF forecast; interpret FE_{j,t} as monetary policy surprises.
  2. Address endogeneity by regressing FE_{j,t} on contemporaneous and lagged macroeconomic forecast variables central banks plausibly use; residuals ε_{j,t} from these regressions are the identified monetary policy shocks (MP_{j,t}).
- Baseline regression (Taylor rule–type specification):
  - Dependent variable: FE_{j,t}.
  - Controls include:
    - One-year ahead real GDP growth forecast: E^1_t y_{j,t}^{f}.
    - One-year ahead inflation forecast: E^1_t π_{j,t}^{f}.
    - Changes in one-year ahead forecasts: ΔE^1_t y_{j,t}^{f} and ΔE^1_t π_{j,t}^{f}.
    - Two quarters of lags of those forecasts.
    - Two quarters of lags of end-of-quarter interest rates: i_{j,t−q} (q = 1,2).
    - Changes in nominal effective exchange rate: ΔNEER_{j,t−q}.
    - Country fixed effect: α_j.
  - Residuals ε_{j,t} used as identified monetary policy shocks.
- Implementation notes:
  - Timespan of available shocks varies across countries with data availability.
  - Appendix considers three alternative specifications to Equation (2); results robust across alternatives.
  - For countries where the forecasted interest rate variable changes over time, realized interest rate series adjusted accordingly.

### Comparison with existing shock measures
- U.S. constructed shocks correlation with Romer and Romer (2004) shocks: 0.78 over 2000Q1-2007Q4.
- U.K. constructed shocks positively and statistically significantly correlated with Cloyne and Hürtgen (2016) shocks; correlation lower than U.S.
- Robustness: excluding the U.S. and U.K. from baseline analyses does not alter main results.

---

### 3. Monetary Policy Transmission

### Empirical framework
- Estimation: local projections following Jordà (2005) (Equation (3)).
- Dependent variables: log real GDP and log consumer price index.
- Shock interpretation: MP_{j,t-1} interpreted as percent response to a 100-basis point exogenous increase in domestic monetary policy shocks at horizon h (β_h).
- Controls: Z_{j,t} contains 4 lags of dependent variable and monetary policy shocks; country fixed effects α_j; time fixed effects α_t; standard errors clustered at country level.
- Confidence intervals: 68 percent and 90 percent reported.

### Baseline impulse-response results
- Output:
  - A 100-bps monetary policy tightening shock leads to a 0.3 percent decline in real GDP within one quarter.
  - Effect is persistent and remains at 0.3 percent 7 quarters after the shock.
- Inflation:
  - Transmission to headline and core inflation is slower than to output.
  - Effects become significant by the second quarter and peak at 6 quarters after the shock.
- Robustness:
  - Baseline robust to alternative shock constructions: Model 1 (without lags of realized interest rates and change in NEER), Model 2 (only lags of realized interest rates), Model 3 (only change in NEER).
  - Robust to controlling for fiscal policy (change in cyclically adjusted primary balance) and changes in NEER.
  - Instrumental variable local projection (instrumenting Δi with MP) yields baseline-consistent magnitudes; reported first-stage F-statistic = 153.

### Addressing the price puzzle
- Three measures compared in Equation (3): (1) changes in actual interest rates, (2) forecast errors FE_{j,t}, (3) constructed monetary policy shocks MP_{j,t}.
- Findings:
  - Using changes in actual interest rates: an increase in interest rates associated with rises in consumer prices (evidence of endogeneity).
  - Using forecast errors: reduces magnitude but does not resolve price puzzle.
  - Using constructed monetary policy shocks (purged residuals): prices decline following shocks, resolving the price puzzle.

### Heterogeneity in transmission

- Exchange rate regime:
  - Classification: Fixed includes Pre-announced peg or Crawling peg using US dollar anchor; others Floating.
  - Transmission stronger under Floating regimes; Fixed regimes show limited output and price responses.
  - Statistical significance confirmed in Column 1 of Tables A2 and A3.
- Income and regional groups:
  - AE and EM: similar effects on output and prices.
  - Regional variation: strongest output effect with high persistence in WHD, followed by APD and EUR.
  - Price responses: insignificant in WHD; substantial in EUR; significant but relatively weaker in APD.
- Financial development (interaction Equation (5), D_j = 1 if country mean above sample mean):
  - Transmission on real GDP and consumer prices increases with financial development.
  - GDP differences across regimes confirmed in Table A2; consumer price difference significant only during first two quarters (Table A3).
  - Interpreted as consistent with a credit channel: stronger responses where financial development higher.
- Central bank transparency:
  - Using Dincer, Eichengreen and Geraats (2022) index.
  - Real GDP responses not significantly different across transparency levels.
  - Prices: stronger transmission in countries with higher central bank transparency; differences statistically significant (Table A3).
  - Interpretation: transparency improves transmission via credibility and better-anchored inflation expectations.

### Role of cyclical factors, complementary policies, and state of the economy
- Uncertainty (Ahir, Bloom and Furceri (2022) index):
  - Transmission to output stronger when countries experience lower uncertainty; difference statistically significant by the 6th quarter (Table A2).
  - Prices: transmission not different across uncertainty states.
- Financial conditions (IMF GFSR index):
  - Transmission to output and prices stronger when financial conditions are tighter (higher index).
  - Differences more evident and significant for output than prices.
- Business cycle and shock sign:
  - Output gap (HP filter): transmission to output stronger when output gap below country mean; prices more responsive when output gap is high.
  - Shock sign asymmetry: contractionary shocks more powerful than expansionary shocks on GDP (consistent with Tenreyro and Thwaites (2016)).
  - Triple interaction: negative shocks are expansionary when the economy is weak; positive shocks are contractionary when output gap above average.
- Fiscal policy complementarity:
  - Using changes in cyclically adjusted primary balance: monetary tightening shocks have stronger effects on output and prices when fiscal policy is tighter.
  - Interpretation: monetary/fiscal coordination enhances central bank effectiveness.

### Alternative specifications and robustness
- Interaction constructions: median-split dummies, sample-split dummies, and smooth-transition approach (Tenreyro and Thwaites 2016); results similar to baseline.
- Alternative shock purging specifications: Model 1, Model 2, Model 3 produce similar impulse responses (Tables A4, Figures A4–A6).
- LP-IV approach (Δi instrumented with MP) consistent with baseline (Figure A21); first-stage F-statistic = 153.
- Rolling-window time-variation analysis with window size of 60 quarters (e.g., 1995–2009, 1996–2010, …, 2009–2023) for 25 countries (to keep sample constant across windows).

### Time-variation findings
- GDP growth transmission largely remained stable across rolling windows.
- Transmission to inflation weakened in years post-COVID19 pandemic.
- Interpretation: increased supply-driven share of inflation during the pandemic reduces monetary policy effectiveness on inflation.

### Key numerical and sample notes
- Shock interpretation: response to a 100-basis point exogenous increase.
- Short-run GDP response: 0.3 percent decline within one quarter; persists at 0.3 percent at 7 quarters.
- U.S. shock correlation with Romer and Romer (2004): 0.78 (2000Q1-2007Q4).
- LP-IV first-stage F-statistic: 153.
- Sample of country shock availability (selected entries shown in Table 1): e.g., USA 1991q2 2023q1; AUS 1995q2 2022q4; CAN 1991q2 2023q1; GBR 1991q2 2021q4; DEU 1999q2 2023q1; JPN 2010q3 2023q1.
- Number of positive and negative shocks by country (selected entries, Table A5): USA: 70 positive, 58 negative; CAN: 57 positive, 71 negative; AUS: 54 positive, 55 negative.

### Conclusions and policy implications
- Scope: novel monetary policy shocks measure across 33 advanced and emerging market economies spanning three decades.
- Main conclusions:
  - Tightening monetary policy quickly affects economic activity; effects on inflation take substantial time to materialize.
  - Transmission exhibits significant heterogeneity across countries tied to cyclical conditions and structural characteristics.
  - Monetary policy more effective in countries with:
    - More developed financial systems.
    - Credible monetary policy frameworks and higher central bank transparency.
    - Floating exchange rate systems.
  - Transmission is muted during periods of high uncertainty and when fiscal policy operates in the opposite direction.
  - Negative shocks can be expansionary when the economy is weak; positive shocks are contractionary when output gap above average.
- Policy recommendations:
  - Design stabilization policy accounting for conditions under which monetary policy is more effective.
  - Pursue structural reforms to enhance financial development to strengthen the credit channel.
  - Maintain central bank independence and transparency to improve transmission via credibility and anchored expectations.
  - Incorporate non-linearity in models assessing monetary policy effectiveness to reflect structural and cyclical country conditions.

*Source: 2. Data and Monetary Policy Shocks Identification & 3. Monetary Policy Transmission, wpiea2023204-print-pdf*

### 2. Data and Monetary Policy Shocks Identification

### 2. Data and Monetary Policy Shocks Identification

### Data
- Primary approach: construct monetary policy surprises as forecast errors in short-term rates using Consensus Forecasts (CF) data, which provides end-of-month short-term interest rate forecasts.
- Realized short-term interest rates: retrieved from country-specific CF reports when available and from HAVER, CEIC and Bloomberg; sample comprises 33 economies, both advanced and emerging.
- CF forecasts used for inflation and real GDP growth as well: monthly, one-year ahead, median (across individual forecasters) forecasts.
- Realized macro and price series: HAVER database for realized quarterly inflation, real GDP growth rates, and nominal effective exchange rates.
- Frequency alignment and non-overlapping horizons:
  - CF provides short-term interest rate forecasts that are 3 months ahead and end-of-month (e.g., December 2021 vintage forecasts March 2022 end-of-month).
  - To avoid overlapping forecast horizons inherent to monthly data, the study uses quarterly frequency by taking CF vintages from December, March, June, and September for each year-quarter and comparing them with end-of-month (quarter) realized interest rates for March, June, August, and December, respectively. This produces quarterly forecast errors without overlapping distortions.
- Auxiliary datasets used to estimate heterogeneity in transmission:
  - Exchange rate regime classification from Ilzetzki, Reinhart and Rogoff (2019).
  - Level of financial development from the IMF Financial Development Index Database.
  - Quarterly uncertainty indexes from Ahir, Bloom and Furceri (2022).
  - Central bank transparency measure from Dincer, Eichengreen and Geraats (2022).
  - Annual cyclically adjusted primary balance from IMF.
  - Financial condition indexes from IMF GFSR.
- Euro Area (EA) treatment:
  - For EA countries CF provides the European Central Bank's unified policy rate and country-specific macro forecasts, enabling identification of shocks exogenous to individual EA country macro conditions.
  - This permits inclusion of Austria, Belgium, Portugal, and Ireland from the date they joined the EA even when CF does not provide country-specific policy rate forecasts directly.

### Monetary Policy Shocks Identification
- Two-step identification procedure:
  1. Construct quarterly interest rate forecast errors (monetary policy surprises) for each country as the difference between realized short-term interest rates and the one-quarter-ahead end-of-month CF forecast. These forecast errors (FE_{j,t}) capture information available to market participants before the quarter and are interpreted as monetary policy surprises.
  2. Address endogeneity by regressing forecast errors on contemporaneous and lagged macroeconomic forecast variables that monetary authorities would plausibly use when setting policy; residuals from these regressions are taken as identified monetary policy shocks.
- Baseline regression (Taylor rule–type specification, following Romer and Romer (2004) and Furceri, Loungani, and Zdzienicka (2018)):
  - Dependent variable: forecast error in short-term rate for country j at time t (FE_{j,t}).
  - Controls include:
    - One-year ahead real GDP growth forecast: E^1_t y_{j,t}^{f}.
    - One-year ahead inflation forecast: E^1_t π_{j,t}^{f}.
    - Changes in one-year ahead forecasts: ΔE^1_t y_{j,t}^{f} and ΔE^1_t π_{j,t}^{f}.
    - Two quarters of lags of those forecasts.
    - Two quarters of lags of end-of-quarter interest rates: i_{j,t−q} (q = 1,2).
    - Changes in nominal effective exchange rate: ΔNEER_{j,t−q} (included because some central banks consider exchange rate movements).
    - Country fixed effect: α_j.
  - After estimating the country-specific regression (Equation (2)), residuals ε_{j,t} are used as the identified monetary policy shocks.
- Implementation notes:
  - Timespan of available shocks varies across countries depending on data availability for the variables in Equation (2).
  - Appendix and results consider three alternative specifications to Equation (2); results are robust across these alternatives.
  - For some countries where the forecasted interest rate variable changes over time, the realized interest rate series used in forecast error calculations is adjusted accordingly.

### Comparison with Shocks from the Literature
- U.S. comparison: the constructed U.S. monetary policy shocks are highly correlated (0.78) with Romer and Romer (2004) shocks over the sample 2000Q1-2007Q4.
- U.K. comparison: the constructed U.K. shocks are positively and statistically significantly correlated with Cloyne and Hürtgen (2016) shocks; the correlation is lower than for the U.S.
- Robustness check: excluding the U.S. and U.K. from baseline analyses does not alter the main results.

*Source: 2. Data and Monetary Policy Shocks Identification, wpiea2023204-print-pdf*

### 3. Monetary Policy Transmission

### 3. Monetary Policy Transmission

### Empirical framework
- Estimation method: local projections following Jorda (2005) (Equation (3)).
- Dependent variables: log real GDP and log consumer price index.
- Monetary policy shock: MP_{j,t-1}, interpreted as percent response to a 100-basis point exogenous increase in domestic monetary policy shocks at horizon h (β_h).
- Controls: vector Z_{j,t} contains 4 lags of dependent variable and monetary policy shocks; country fixed effects α_j; time fixed effects α_t; standard errors clustered at the country level.
- Confidence intervals reported: 68 percent and 90 percent.

### Baseline results
- Output response:
  - A 100-bps monetary policy tightening shock leads to a 0.3 percent decline in real GDP within one quarter.
  - The effect is persistent and remains at 0.3 percent 7 quarters after the shock (Panel A).
- Inflation response:
  - Transmission to headline and core inflation is slower than to output.
  - Effects become significant by the second quarter and peak at 6 quarters after the shock (Panels B and C).
- Robustness:
  - Baseline robust to alternative specifications generating monetary policy shocks: Model 1 (without lags of realized interest rates and change in NEER), Model 2 (only lags of realized interest rates), Model 3 (only change in NEER).
  - Results robust to controlling for fiscal policy (change in cyclically adjusted primary balance) and changes in NEER.
  - Instrumental variable local projection (instrumenting changes in interest rates with monetary policy shocks) yields baseline-consistent magnitudes (Figure A21).

### Addressing the price puzzle
- Comparison of three measures in Equation (3): (1) changes in actual interest rates, (2) forecast errors (FE_{j,t}), (3) constructed monetary policy shocks (MP_{j,t}).
- Findings:
  - An increase in interest rates is associated with rises in consumer prices (Figure 2 Panel A) — evidence of endogeneity.
  - Using forecast errors reduces the magnitude but does not resolve the price puzzle (Figure 2 Panel B).
  - Prices decline only following the monetary policy shocks that purge current and expected state information from forecast errors (Figure 2 Panel C).

### Heterogeneity in transmission: exchange rate regime, income, and regions
- Exchange rate regime:
  - Estimation via Equation (4) with Fixed dummy from Ilzetzki, Reinhart and Rogoff (2019) classification (Pre-announced peg or Crawling peg using US dollar anchor assigned to Fixed; others as floating).
  - Monetary policy transmission is stronger in floating exchange rate regimes than fixed regimes; fixed regimes show limited output and price responses (Figure 3). Column 1 in Tables A2 and A3 confirm statistical significance.
- Income and regional groups:
  - AE and EM: similar effects on output and prices (Figure A9).
  - Regional variation: strongest output effect with high persistence in WHD, followed by APD and EUR.
  - Price responses: insignificant in WHD; substantial in EUR; significant but relatively weaker in APD (Figure A10).

### Heterogeneity: economic structures and institutions
- Method: Equation (5) interacts MP with D_j (state variable dummy equal to 1 if country mean of state variable above sample mean).
- Financial development:
  - Transmission on both real GDP and consumer prices is increasing with financial development (Figure 4).
  - Table A2 confirms statistical GDP differences across regimes; Table A3 shows consumer price difference significant only during the first two quarters.
  - Interpretation: consistent with credit channel—stronger output and price responses where financial development is higher.
- Central bank transparency:
  - Using Dincer, Eichengreen and Geraats (2022) index: real GDP responses not significantly different across transparency levels (Figure 5 Panel A).
  - Prices: stronger transmission in countries with higher central bank transparency; differences statistically significant (Table A3).
  - Interpretation: transparency improves transmission via credibility and better-anchored inflation expectations.

### Role of cyclical factors, complementary policies, and state of the economy
- Uncertainty (Equation (6)):
  - Using Ahir, Bloom and Furceri (2022) uncertainty index and country-time dummy above country mean.
  - Output: transmission stronger when countries experience lower uncertainty; difference statistically significant by the 6th quarter (Table A2).
  - Prices: transmission not different across uncertainty states (Figure 6).
- Financial conditions:
  - Quarterly financial condition indexes from IMF GFSR; high/low assigned using lagged index relative to country mean.
  - Transmission to output and prices stronger when financial conditions are tighter (higher index) (Figure 7).
  - Differences more evident and significant in output than prices.
- Business cycle and shock sign:
  - Output gap computed using Hodrick-Prescott filter; high/low by country mean.
  - Transmission to output stronger when output gap is below country mean; prices more responsive when output gap is high (Figure 8).
  - Shock sign (Equation (8)): contractionary shocks are more powerful than expansionary shocks on GDP (Figure 9), consistent with Tenreyro and Thwaites (2016).
  - Triple interaction (Equation (9)): negative shocks are expansionary when the economy is weak, and positive shocks are contractionary when output gap is above average (Figure 10).
- Fiscal policy (Equation (7)):
  - Cyclically adjusted primary balance changes; “tight” if change positive, “loose” if negative.
  - Monetary tightening shocks have stronger effects on output and prices when fiscal policy is tighter (Figure 11).
  - Interpretation: monetary/fiscal coordination enhances central bank effectiveness.

### Alternative specifications and robustness
- Alternative interaction constructions:
  - Dummies based on median of state variable and sample.
  - Smooth-transition approach allowing responses to vary smoothly (Tenreyro and Thwaites 2016).
  - Results similar to baseline (Appendix Figures A11-20).
- Instrumental variable local projection (interest rate changes instrumented with monetary policy shocks) consistent with baseline (Figure A21).

### Time-variation in transmission
- Rolling-window methodology: overlapping windows shifting forward by one year (example: first window 1995–2009, then 1996–2010, …, final window 2009–2023).
- Findings:
  - GDP growth transmission largely remained stable (Figure A22).
  - Transmission to inflation weakened in the years post-COVID19 pandemic (Figure A22).
  - Interpretation: increased supply-driven share of inflation during the pandemic makes monetary policy less effective on inflation (Firat and Hao (2023)).

### Conclusions and policy implications
- Scope: novel monetary policy shocks measure across 33 advanced and emerging market economies spanning three decades.
- Key findings:
  - Tightening monetary policy quickly affects economic activity; effects on inflation take substantial time to materialize.
  - Significant heterogeneity in transmission across countries tied to cyclical conditions and structural characteristics.
  - Monetary policy more effective in countries with:
    - More developed financial systems.
    - Credible monetary policy frameworks and higher central bank transparency.
    - Floating exchange rate systems.
  - Transmission muted during periods of high uncertainty and when fiscal policy operates in the opposite direction.
  - Negative shocks are expansionary when the economy is weak; positive shocks are contractionary when output gap is above average.
- Policy recommendations and implications:
  - Design stabilization policy accounting for conditions under which monetary policy is more effective.
  - Pursue structural reforms to enhance financial development to strengthen the credit channel.
  - Maintain central bank independence and transparency to improve monetary transmission via credibility and anchored expectations.
  - Incorporate non-linearity in models assessing monetary policy effectiveness to reflect structural and cyclical country conditions.

*Source: wpiea2023204-print-pdf - 3. Monetary Policy Transmission*

### References

### References

### Cited works
- Ahir, Hites, Nicholas Bloom, and Davide Furceri, 2022. “The World Uncertainty Index,” NBER Working Paper

*Source: wpiea2023204-print-pdf - References*

### 29763. National Bureau of Economic Research: Cambridge, MA.

### Monetary Policy Transmission: Figures, Tables, and References (Appendix extracts)

### References and literature cited
- Albagli, Elias, Mr Francesco Grigoli, and Emiliano Luttini. Inflation Expectations and the Supply Chain. International Monetary Fund, 2022.
- Alvarez, Fernando E., Francesco Lippi, and Luigi Paciello. "Optimal price setting with observation and menu costs." The Quarterly journal of economics 126, no. 4 (2011): 1909-1960.
- Angrist, Joshua D., Òscar Jordà, and Guido M. Kuersteiner. "Semiparametric estimates of monetary policy effects: string theory revisited." Journal of Business & Economic Statistics 36, no. 3 (2018): 371-387.
- Barnichon, Regis, and Christian Matthes. "Functional approximation of impulse responses." Journal of Monetary Economics 99 (2018): 41-55.
- Bauer, Michael D., and Eric T. Swanson. "A Reassessment of Monetary Policy Surprises and High-Frequency Identification." NBER Macroeconomics Annual 2022, Volume 37.
- Bems, Rudolfs, Francesca Caselli, Francesco Grigoli, and Bertrand Gruss. "Gains from anchoring inflation expectations: Evidence from the taper tantrum shock." Economics Letters 188 (2020): 108820.
- Beyer, Robert, Rupa Duttagupta, Alexandra Fotiou, Keiko Honjo, Mark Horton, Zoltan Jakab, Jesper Lindé et al. "Shared Problem, Shared Solution: Benefits from Fiscal-Monetary Interactions in the Euro Area." (2023).
- Brandao-Marques, Mr Luis, Mr RG Gelos, Mr Thomas Harjes, Ms Ratna Sahay, and Yi Xue. Monetary policy transmission in emerging markets and developing economies. International Monetary Fund, 2020.
- Champagne, Julien, and Rodrigo Sekkel. "Changes in monetary regimes and the identification of monetary policy shocks: Narrative evidence from Canada." Journal of Monetary Economics 99 (2018): 72-87.
- Choi, Sangyup, Tim Willems, and Seung Yong Yoo. Revisiting the monetary transmission mechanism through an industry-level differential approach. International Monetary Fund, 2022.
- Cloyne, James, and Patrick Hürtgen. "The macroeconomic effects of monetary policy: a new measure for the United Kingdom." American Economic Journal: Macroeconomics 8, no. 4 (2016): 75-102.
- Debortoli, Davide, Mario Forni, Luca Gambetti, and Luca Sala. "Asymmetric effects of monetary policy easing and tightening." (2020).
- Firat, Melih and Otso Hao. “Demand vs. Supply Decomposition of Inflation: Cross-Country Evidence with Applications.” International Monetary Fund, 2023.
- Furceri, Davide, Prakash Loungani, and Aleksandra Zdzienicka. "The effects of monetary policy shocks on inequality." Journal of International Money and Finance 85 (2018): 168-186.
- Georgiadis, Georgios. "Towards an explanation of cross-country asymmetries in monetary transmission." Journal of Macroeconomics 39 (2014): 66-84.
- Gertler, Mark, and Peter Karadi. "Monetary policy surprises, credit costs, and economic activity." American Economic Journal: Macroeconomics 7, no. 1 (2015): 44-76.
- Gürkaynak, Refet S., Brian Sack, and Eric Swanson. "The sensitivity of long-term interest rates to economic news: Evidence and implications for macroeconomic models." American economic review 95, no. 1 (2005): 425-436.
- Ilzetzki, Ethan, Carmen M. Reinhart, and Kenneth S. Rogoff. “Exchange arrangements entering the twenty-first century: Which anchor will hold?” Quarterly Journal of Economics 134, no. 2 (2019): 599-646.
- Jarociński, Marek, and Peter Karadi. "Deconstructing monetary policy surprises—the role of information shocks." American Economic Journal: Macroeconomics 12, no. 2 (2020): 1-43.
- Miranda-Agrippino, Silvia, and Giovanni Ricco. "The transmission of monetary policy shocks." American Economic Journal: Macroeconomics 13, no. 3 (2021): 74-107.
- Nakamura, Emi, and Jón Steinsson. "Five facts about prices: A reevaluation of menu cost models." The Quarterly Journal of Economics 123, no. 4 (2008): 1415-1464.
- Pellegrino, Giovanni. "Uncertainty and monetary policy in the US: A journey into nonlinear territory." Economic Inquiry 59, no. 3 (2021): 1106-1128.
- Romer, Christina D., and David H. Romer. "A new measure of monetary shocks: Derivation and implications." American economic review 94, no. 4 (2004): 1055-1084.
- Rusnák, Marek, Tomas Havranek, and Roman Horváth. "How to solve the price puzzle? A meta‐analysis." Journal of Money, Credit and Banking 45, no. 1 (2013): 37-70.
- Sims, Christopher A. "Interpreting the macroeconomic time series facts: The effects of monetary policy." European economic review 36, no. 5 (1992): 975-1000.
- Stock, James H., and Mark W. Watson. "Identification and estimation of dynamic causal effects in macroeconomics using external instruments." The Economic Journal 128, no. 610 (2018): 917-948.
- Taylor, John B. "The monetary transmission mechanism: an empirical framework." Journal of economic perspectives 9, no. 4 (1995): 11-26.
- Tenreyro, Silvana, and Gregory Thwaites. "Pushing on a string: US monetary policy is less powerful in recessions." American Economic Journal: Macroeconomics 8, no. 4 (2016): 43-74.

### Table 1 — Availability of monetary policy shocks by country (iso3 code, Start, End)
- AUS 1995q2 2022q4
- AUT 1998q4 2023q1
- BEL 1998q4 2023q1
- BRA 2002q4 2023q1
- CAN 1991q2 2023q1
- CHE 1991q2 2022q1
- CHL 2002q4 2023q1
- CZE 2008q3 2022q4
- DEU 1999q2 2023q1
- ESP 1999q2 2023q1
- FRA 1999q2 2023q1
- GBR 1991q2 2021q4
- GRC 1998q4 2023q1
- HUN 2008q3 2022q4
- IDN 2014q2 2023q1
- IND 2002q3 2023q1
- IRL 1998q4 2023q1
- ITA 1999q2 2023q1
- JPN 2010q3 2023q1
- KOR 1995q2 2023q1
- MEX 2002q4 2023q1
- MYS 1995q2 2023q1
- NLD 1999q2 2023q1
- NOR 2004q2 2023q1
- NZL 1995q4 2023q1
- PHL 2009q3 2023q1
- POL 2008q3 2022q4
- PRT 1998q4 2023q1
- SVK 2008q3 2022q4
- SWE 2004q2 2020q2
- THA 1999q2 2023q1
- TUR 2008q3 2022q4
- USA 1991q2 2023q1

### Key methodology notes from figures and captions
- Impulse response functions are based on local projection methods following Jordà (2005) using an unbalanced panel of quarterly data for 33 countries from 1990 to 2019.
- Regression specification (general form used across figures):
  - y_{j,t+h} − y_{j,t−1} = β_h MP_{j,t−1} + ∑_{ℓ} θ_ℓ Z_{j,t−ℓ} + α_j + α_t + ε_{j,t}
  - Where y_{j,t+h} is log real GDP or log consumer price index (or log core CPI for Figure A3) and MP_{j,t−1} is the lag of monetary policy shocks in country j.
  - The coefficient β_h denotes (percent) response of variables to domestic monetary policy shocks at a horizon of h quarters.
  - Z contains 4 lags of the dependent variable and monetary policy shocks.
  - α_j are country fixed effects; α_t are time fixed effects (e.g., VIX, world energy and food prices).
  - Solid line: point estimate for β_h; dashed and short-dashed lines: 68 percent and 90 percent confidence intervals respectively.
  - Standard errors are clustered at country level (unless otherwise noted).

### Figure themes and heterogeneity dimensions (enumerated)
- Baseline transmission to:
  - GDP, Headline Inflation, Core Inflation (Figure 1; Figure A3 for Core CPI).
- Price puzzle analysis using:
  - Raw Interest Rates; Forecast Errors; Monetary Policy Shocks (Figure 2).
- Heterogeneity by:
  - Exchange rate regime: Floating vs Fixed (Figure 3).
  - Financial development: High vs Low; median split; smooth transition function (Figures 4, A11, A12).
  - Central bank transparency: High vs Low; median split; smooth transition function (Figures 5, A13, A14).
  - Uncertainty: High vs Low; median split; smooth transition function (Figures 6, A15, A16).
  - Financial Conditions Index: High vs Low; median and smooth transition (Figures 7, A17, A18).
  - Business cycle / output gap: High vs Low; median and smooth transition (Figures 8, A19, A20).
  - Shock sign asymmetry: Positive vs Negative shocks; Positive shock * business cycle and Negative shock * business cycle interactions (Figures 9, 10).
  - Complementarity with fiscal policy: Tight vs Loose fiscal stance (Figure 11).
  - Advanced Economies (AEs) vs Emerging Markets (EMs) (Figure A9).
  - Regional variation: Asia-Pacific, Europe, Americas (Figure A10).
  - AE vs EM and regional breakdowns reported separately in figures.
- Robustness and alternative specifications:
  - Alternative shock constructions: Model 1 (baseline excluding purging for lagged interest rates and changes in NEER), Model 2 (purging only for changes in NEER), Model 3 (purging only for lagged interest rates) (Table A4 and Figures A4–A6).
  - Baseline robustness controlling for Fiscal Expenditure (ΔX controls for change in cyclically adjusted primary balance) (Figure A7).
  - Baseline robustness controlling for Nominal Effective Exchange Rate (ΔX controls for changes in NEER) (Figure A8).
  - LP-IV approach similar to Stock and Watson (2018) (Figure A21).
  - Time-variation via rolling window regressions with window size of 60 quarters (Figure A22).

### Table A1 — Interest rates forecasted by CF and realized rates sources (selected country entries)
- AUS: CF Previous CF rate = 90-day Dealer Bill Rate; Realized rates source = CEIC
- AUT: CF Previous CF rate = 3 month Euro Rate; Realized rates source = HAVER
- BEL: CF Previous CF rate = 3 month Euro Rate; Realized rates source = HAVER
- BRA: CF Previous CF rate = Overnight Interbank Interest Rate, SELIC; Realized rates source = HAVER
- CAN: CF Previous CF rate = 3 month Treasury Bill Rate; Realized rates source = HAVER
- CHE: CF Previous CF rate = 3 month Swiss Av. Rate (SAR3M); Realized rates source = Bloomberg (Realized: 3 month Euro-Franc Rate Bloomberg)
- CHL: CF Previous CF rate = Central Bank Monetary Policy Rate; Realized rates source = HAVER (Realized: Nominal Central Bank 90-Day Bill Rate (%) HAVER)
- CZE: CF Previous CF rate = 3 Month PRIBOR Interbank Deposit Rate; Realized rates source = HAVER
- DEU: CF Previous CF rate = 3 month Euro Rate; Realized rates source = HAVER
- ESP: CF Previous CF rate = 3 month Euro Rate; Realized rates source = HAVER
- FRA: CF Previous CF rate = 3 month Euro Rate; Realized rates source = HAVER
- GBR: CF Previous CF rate = 3 month SONIA Rate; Realized rates source = Bloomberg
- JPN: CF Previous CF rate = 3 month Yen TIBOR Rate; Realized rates source = HAVER
- KOR: CF Previous CF rate = 91-day Certificate of Deposit; Realized rates source = HAVER
- MEX: CF Previous CF rate = 28-Day CETES Rate; Realized rates source = Bloomberg
- MYS: CF Previous CF rate = 3-month Interbank Rate; Realized rates source = Bloomberg
- NOR: CF Previous CF rate = 3 month Interbank Rate; Realized rates source = Bloomberg
- NZL: CF Previous CF rate = 90-day Bank Bill Rate; Realized rates source = Bloomberg
- PHL: CF Previous CF rate = 91-Day Treasury Bill Rate; Realized rates source = HAVER (Realized: 3-month PHIBOR Interbank Rate HAVER)
- SWE: CF Previous CF rate = 3 month Interbank Rate; Realized rates source = HAVER
- THA: CF Previous CF rate = 3-month Interbank Rate; Realized rates source = HAVER
- TUR: CF Previous CF rate = Overnight Interbank Interest Rate (%); Realized rates source = HAVER (Realized: Overnight Interbank Interest Rate HAVER)
- USA: CF Previous CF rate = 3 month Treasury Bill Rate; Realized rates source = HAVER

### Table A2 — P values: GDP (p-values by horizon and heterogeneity dimension)
- Horizons 1 to 8 with columns:
  - Exchange Rate System, Financial Development, Central Bank Transparency, Uncertainty, Financial Conditions Index, Asymmetry Positive shock * business cycle, Negative shock * business cycle
- Values by horizon:
  - 1: 0.23 0.26 0.95 0.16 0.76 0.41 0.17 0.00
  - 2: 0.43 0.99 0.05 0.18 0.47 0.76 0.04 0.00
  - 3: 0.00 0.51 0.52 0.67 0.18 0.42 0.01 0.02
  - 4: 0.08 0.10 0.81 0.46 0.16 0.50 0.04 0.03
  - 5: 0.26 0.20 0.70 0.45 0.14 0.42 0.01 0.02
  - 6: 0.10 0.03 0.86 0.11 0.15 0.35 0.01 0.03
  - 7: 0.05 0.23 0.65 0.09 0.11 0.43 0.00 0.04
  - 8: 0.56 0.18 0.46 0.02 0.33 0.37 0.03 0.06
- Note: Numbers represent p-values. Those in italic (in original) indicate ρ ≤ 0.1 signifying rejection of the null hypothesis (β1 = β2).

### Table A3 — P values: CPI (p-values by horizon and heterogeneity dimension)
- Horizons 1 to 8 with columns:
  - Exchange Rate System, Financial Development, Central Bank Transparency, Uncertainty, Financial Conditions Index, Asymmetry Positive shock * business cycle, Negative shock * business cycle
- Values by horizon:
  - 1: 0.87 0.08 0.35 0.04 0.08 0.14 0.13 0.15
  - 2: 0.74 0.05 0.05 0.95 0.34 0.24 0.03 0.15
  - 3: 0.68 0.39 0.00 0.61 0.70 0.36 0.30 0.17
  - 4: 0.03 1.00 0.04 0.92 0.29 0.44 0.26 0.10
  - 5: 0.12 0.57 0.08 0.94 0.35 0.35 0.61 0.12
  - 6: 0.20 0.86 0.05 0.88 0.48 0.45 0.29 0.17
  - 7: 0.06 0.69 0.01 0.66 0.71 0.58 0.65 0.30
  - 8: 0.01 0.35 0.15 0.57 0.98 0.67 0.34 0.22
- Note: Numbers represent p-values. Those in italic (in original) indicate ρ ≤ 0.1 signifying rejection of the null hypothesis (β1 = β2).

### Table A4 — Alternative model specifications for monetary policy shocks (model brief)
- Model 1: Baseline excluding purging for lagged interest rates and changes in NEER
  - FE_{j,t} = α_j + γ1 E^1 y_{j,t}^{t} + γ2 ΔE^1 y_{j,t}^{t} + γ3 E^1 π_{j,t}^{t} + γ4 ΔE^1 π_{j,t}^{t} + ∑ δ_j,ℓ y_{j,t−ℓ} + ∑ δ_j,ℓ π_{j,t−ℓ} + ε_{j,t}
  - E^1 y and E^1 π denote one-year ahead real GDP growth and inflation forecasts.
- Model 2: Purging only for changes in NEER — includes ΔNEER controls.
- Model 3: Purging only for lagged interest rates — includes two-quarter lags of end-of-quarter interest rates i_{j,t−ℓ}.
- All models control for one-year ahead forecasts, their changes, and two-quarter lags of dependent variables where specified.

### Table A5 — Number of positive and negative shocks in the sample by country (selected entries)
- AUS: Number of Positive MP shocks 54; Number of Negative MP shocks 55
- AUT: 42 54
- BEL: 46 50
- BRA: 40 42
- CAN: 57 71
- CHE: 58 64
- CHL: 45 37
- CZE: 31 23
- DEU: 45 49
- ESP: 50 44
- FRA: 44 50
- GBR: 68 55
- GRC: 46 50
- HUN: 27 27
- IDN: 21 13
- IND: 42 41
- IRL: 44 52
- ITA: 50 44
- JPN: 28 21
- KOR: 54 56
- MEX: 39 43
- MYS: 56 54
- NLD: 46 48
- NOR: 35 39
- NZL: 53 55
- PHL: 29 23
- POL: 28 26
- PRT: 48 48
- SVK: 25 26
- SWE: 28 35
- THA: 52 42
- TUR: 22 32
- USA: 70 58

### Figure comparisons and robustness highlights
- Figure A1: Comparison between the study's US shocks and Romer and Romer (2004) — correlation uses Model 1A specification for closeness; main results use specification purged for interest rate lags and NEER changes.
- Figure A2: Comparison between the study's UK shocks and Cloyne and Hürtgen (2016).
- Figures A4–A6: Baseline estimations using shocks from Model 1 (not purging), Model 2 (purging only for NEER), and Model 3 (purging only for lagged interest rates) reported to show robustness of impulse responses.
- Figure A21 (LP-IV): First-stage regression: Δi_{j,t} = β_h MP_{j,t−1} + ∑ θ_ℓ Z_{j,t−ℓ} + α_j + α_t + ε_{j,t}. Second stage uses predicted changes in interest rates. Reported F-statistics = 153, well above threshold value of 10 indicating statistical significance.

### Time-variation and sample notes
- Time-variation analysis via rolling window regressions with window size of 60 quarters, starting window 1995Q1–2009Q4; reported β_h from the 6th quarter of each sample (6th quarter is the horizon where impulse responses reach their peak).
- For rolling window results, focus is on 25 countries (excluded 8 countries with short shock availability) to keep country sample unchanged across windows.

*Content extracted from the working paper appendix, tables, figures, and references as provided.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023204-print-pdf.pdf_