## wpiea2023086-print-pdf

## Source details

**Canonical URL:** [wpiea2023086-print-pdf](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023086-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023086-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023086-print-pdf.pdf.json)

---

### Scope, data, and methods
- Sample: 46 advanced and emerging market economies since 1990; larger robustness sample: unbalanced sample of 141 countries.
- Primary empirical approach: common local projections specification (Jordà (2005)) to estimate exchange rate pass-through into consumer prices, import prices, and inflation expectations.
- Additional identification: difference-in-differences combined with instrumental variables to isolate the impact of U.S. monetary policy shocks on domestic prices through the exchange rate channel.
  - Instrument construction uses U.S. monetary policy shocks from Jarociński and Karadi (2020).
  - Instrument: Instrument_{i,t} = Quinn_i × Regime_{it} × USMPShock_t, where Regime_{it} distinguishes pre-announced peg to the US dollar versus other regimes and Quinn_i indicates relatively open capital accounts (Quinn and Toyoda (2008) threshold at 10th percentile).
- Key data and measures:
  - Monthly data January 1990–December 2022 (baseline), dependent variables include monthly inflation (lagged yearly change in log CPI), import price index, and one-year-ahead inflation expectations.
  - State variables: quarterly GDP growth (deviation from country mean), output gap (HP filter, smoothness parameter 1600), Ahir–Bloom–Furceri (2022) uncertainty index, inflation forecast disagreement from Consensus Forecasts, USD invoice share (Boz and others (2022)), dollarization rates (Reinhart, Rogoff, and Savastano (2003)).
  - Observations with dependent variable below 1st percentile or above 99th percentile excluded.
  - Number of countries in IV analysis drops to 41 due to missing Quinn and Toyoda (2008) index for five countries.

### Baseline quantitative findings
- Average pass-through:
  - A depreciation of one percent of the local exchange rate against the US dollar is associated with an increase in domestic prices of 0.16 percent after one year.
  - Advanced economies: 0.08 percent after one year.
  - Emerging market economies: 0.3 percent after one year.
- Pass-through to import prices:
  - Average about 0.7 percent after one month; remains about 0.7 percent one year after the shock.
  - Import-price pass-through materializes more quickly and is more homogeneous across income levels than to consumer prices.
- Pass-through to inflation expectations:
  - Inflation expectations rise by 0.08 percentage points after six months.
  - Emerging markets: 0.12 percentage points.
  - Advanced economies: 0.03 percentage points.
- Example magnitude statement:
  - A 10 percent depreciation is associated with an increase in the level of consumer prices by about 0.5 percent within one month, rising to about 1.6 percent after one year.

### State-dependent and cross-sectional determinants of pass-through
- Inflation level and expectation anchoring:
  - Pass-through increases with the prevailing level of inflation; non-linear increase concentrated in the fourth quartile (initial inflation above 3.8 percent).
  - Pass-through increases with disagreement among professional forecasters (weaker central bank credibility associated with larger pass-through).
- Uncertainty:
  - Pass-through into consumer prices and inflation expectations increases with the level of economic uncertainty (Ahir, Bloom, and Furceri (2022)); median uncertainty approached its pandemic high during 2022.
- Currency of invoicing and dollarization:
  - Countries with higher USD invoice share of imports experienced more significant pass-through into import prices (consistent with global currency pricing theory).
  - Pass-through rate doubles from 0.1 percent for countries below the sample median USD invoice share to over 0.2 percent for those above the sample median; import price pass-through reaches 0.8 percent above median vs 0.5 percent below.
  - In EMDEs, higher dollarization is associated with stronger pass-through; responses rise substantially when dollarization is above a threshold.
- Nonlinearities by shock sign and size:
  - Pass-through rises with the size of the exchange rate fluctuation; quadratic term 휗ℎ positive and significant for horizons 0–10 months.
  - Pass-through materializes faster following depreciations than appreciations over the first six months; asymmetry more pronounced in EMDEs.
  - Example nonlinear magnitudes at 12 months: pass-through rises from 0.16 percent for a depreciation of 1 percent, to 0.175 for a depreciation of 10 percent, and to about 0.2 for a depreciation of 25 percent.

### Shock-dependent pass-through (U.S. monetary policy origin)
- Difference-in-differences IV strategy isolates exchange rate fluctuations driven by U.S. monetary policy shocks.
  - First-stage result: a U.S. monetary policy tightening of 100 basis points causes the exchange rate to depreciate by 4.5 percent more for countries with a flexible exchange rate and open capital account compared to others.
  - Instrument strength: Kleibergen-Paap rk Wald F-statistic varies between 10.4 and 11.1 for horizons between six months and one year.
  - Exclusion restriction test: regressing residuals on the instrument yields OLS estimates for 훽 very close to zero across horizons with p-value = 1.000 (cannot reject 훽=0).
- Second-stage result: pass-through is about three times larger when exchange rate fluctuations are provoked by U.S. monetary policy shocks; pass-through reaches 0.47 after one year for U.S.-policy-driven fluctuations (compare to average 0.16 at same horizon).

### Baseline results, heterogeneity, and robustness
- Advanced economies vs EMDEs:
  - Pass-through to consumer prices about three times larger in EMDEs than in AEs; import-price responses similar across groups.
  - Core prices and inflation expectations responses substantially larger in EMDEs: a 1 percent depreciation associated with increases in core (expected) inflation of 0.5 (0.3) percentage point in AEs versus 1.3 (1.2) percentage points in EMDEs.
- Regional patterns:
  - Pass-through smallest in Asia and Europe, largest in Latin America; regional differences not statistically significant.
- Robustness:
  - Using the nominal effective exchange rate slightly raises pass-through.
  - Larger sample of 141 countries almost doubles the size of the effect (driven by higher pass-through in developing economies).
  - Results robust to alternative trimming (5th–95th percentiles) and alternative lag orders (6 and 24 lags versus baseline 12 lags).

### Policy and modeling implications
- Policy:
  - Exchange rate movements in 2022 likely have stronger impacts on domestic prices than implied by previous linear estimates.
  - Stronger monetary tightening may be needed to address resulting inflation pressures when exchange rate movements occur in high-inflation, high-uncertainty environments.
- Modeling:
  - Macroeconomic models should generate larger responses of domestic prices to exchange rates for high levels of inflation and uncertainty; linear estimates may understate these effects.

### Summary statistics and instrument diagnostics
- Summary statistics (selected):
  - Consumer Prices (y/y % change): N 12,495; Mean 3.08; Std. Dev. 4.30; Median 2.24.
  - Import Prices (y/y % change): N 12,495; Mean 2.78; Std. Dev. 9.39; Median 1.90.
  - Inflation Expectations (% one-year ahead): N 12,495; Mean 3.12; Std. Dev. 4.07; Median 2.32.
  - Exchange Rate (m/m % change): N 12,495; Mean 0.007; Std. Dev. 2.22; Median 0.01.
  - USD Invoice Share (%, country average): N 7,226; Mean 44.01; Std. Dev. 25.73; Median 32.60.
  - Uncertainty Index: N 12,034; Mean 0.19; Std. Dev. 0.19; Median 0.14.
- Instrument diagnostics:
  - First-stage Kleibergen-Paap rk Wald F-statistic between 10.4 and 11.1 for horizons six months to one year.
  - Exclusion restriction regressions report R-squared 0.000 across horizons and p-value = 1.000 for test of 훽=0.

*IMF Working Paper — State-Dependent Exchange Rate Pass-Through (wpiea2023086-print-pdf)*

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

### Introduction

### Scope, data, and methods
- Sample: 46 advanced and emerging market economies since 1990.
- Primary empirical approach: common local projections specification to estimate exchange rate pass-through into consumer prices, import prices, and inflation expectations.
- Additional identification: difference-in-differences combined with instrumental variables to isolate the impact of U.S. monetary policy shocks on domestic prices through the exchange rate channel.
  - Instrument construction uses a series of U.S. monetary policy shocks externally identified by Jarociński and Karadi (2020) (unanticipated change in rates within high-frequency windows around FOMC announcements).
  - Time fixed effects and interactions of U.S. monetary policy shocks with country-specific measures of exchange rate regime and capital account openness are used to satisfy the exclusion restriction and capture the relative impact via the portfolio investment channel.

### Baseline quantitative findings
- Average pass-through: a depreciation of one percent of the local exchange rate against the US dollar is associated with an increase in domestic prices of 0.16 percent after one year.
- Heterogeneity across country groups (after one year):
  - Advanced economies: 0.08 percent.
  - Emerging market economies: 0.3 percent.
- Pass-through to import prices:
  - Average about 0.7 percent after one month.
  - Pass-through to import prices materializes more quickly and is more homogeneous across income levels than to consumer prices.
- Pass-through to inflation expectations:
  - Inflation expectations rise by 0.08 percentage points after six months.
  - Emerging markets: 0.12 percentage points.
  - Advanced economies: 0.03 percentage points.

### State-dependent and cross-sectional determinants of pass-through
- Uncertainty:
  - Pass-through into consumer prices and inflation expectations increases with the level of economic uncertainty.
  - Interpretation: firms may be less willing to adjust mark-ups after cost increases during periods of elevated uncertainty.
- Inflation and expectation anchoring:
  - Pass-through increases with the prevailing level of inflation.
  - Pass-through increases with the level of disagreement among professional forecasters of inflation (weaker central bank credibility associated with larger pass-through).
- Currency of invoicing:
  - Countries with higher USD invoice share of imports experienced more significant pass-through into import prices (consistent with global currency pricing theory).
- Dollarization:
  - Higher pass-through documented in highly dollarized economies in prior literature; balance-sheet and regime interaction effects noted.
- Nonlinearities: magnitude and direction
  - Pass-through rises with the size of the exchange rate fluctuation.
  - Pass-through materializes faster following depreciations than appreciations.
  - Implication: recovery of a local currency after a depreciation does not produce offsetting effects on prices until about a year later, leaving strong transitory impacts on inflation.

### Shock-dependent pass-through (source of exchange rate fluctuations)
- US monetary policy shocks:
  - Pass-through is about three times larger when exchange rate fluctuations are provoked by U.S. monetary policy shocks than when provoked by other drivers.
  - Identification strategy isolates the exchange rate channel by comparing economies with open capital accounts and flexible exchange rates to those with closed capital accounts or pegged regimes.
  - Interpretation in line with prior work: if firms expect contraction in future demand (e.g., from U.S. monetary tightening), they are less willing to reduce mark-ups following an appreciation and thus pass through higher prices.

### Policy implications and modeling implications
- Policy:
  - Exchange rate movements in 2022 may have stronger impacts on domestic prices than previous linear estimates imply.
  - Stronger monetary tightening may be needed to address resulting inflation pressures when exchange rate movements occur in high-inflation, high-uncertainty environments.
- Modeling:
  - Macroeconomic models should be able to generate larger responses of domestic prices to exchange rates for high levels of inflation and uncertainty; linear estimates may understate these effects.

*IMF Working Paper — Introduction*

### Section 3 presents main results and discusses robustness checks. Section 4 concludes.

### wpiea2023086-print-pdf - Section 3 presents main results and discusses robustness checks. Section 4 concludes.

### Data and sample
- Baseline sample: monthly data from January 1990 to December 2022 covering 46 countries, of which 28 are classified as advanced economies and 18 as emerging market economies.
- Larger robustness sample: unbalanced sample of 141 countries for which consumer price data are available.
- Dependent variables and key state variables:
  - Monthly inflation: (lagged) yearly change in log consumer prices.
  - Quarterly GDP growth (in deviation from country mean) and output gap (HP filter with smoothness parameter equal to 1600).
  - Monthly uncertainty indexes from Ahir, Bloom, and Furceri (2022).
  - Monthly indicator of inflation uncertainty: interquartile range of professional forecasters from Consensus Forecasts (one-year ahead synthetic forecasts computed via linear combination of current and next calendar year forecasts).
  - Average share of imports invoiced in US dollars from Boz and others (2022) (country-average).
  - Dollarization rates for EMDEs from Reinhart, Rogoff, and Savastano (2003) (country-average).
- U.S. monetary policy shock series: monthly Jarociński and Karadi (2020) shock series available from February 1990 to June 2019; these separate central bank information shocks from monetary policy shocks.
- Data handling note: observations with the dependent variable below the 1st percentile or above the 99th percentile of the sample’s empirical distribution are excluded (episodes of hyperinflation or economic collapse).
- Additional datasets/measures used: Ilzetzki, Reinhart, and Rogoff (2019) for anchor currencies and exchange rate arrangements; Quinn and Toyoda (2008) capital account openness score (annual, 0–100) through 2014. Robustness: also use Chinn and Ito (2006) updated capital account openness indexes.
- Practical notes:
  - USD invoice share, forecast disagreement, and dollarization use country average observations due to limited time variation.
  - Number of countries in IV analysis drops to 41 because the Quinn and Toyoda (2008) index is not available for Armenia, Cyprus, Estonia, Slovak Republic, and Slovenia.

### Baseline empirical specification (average pass-through)
- Local projections (Jordà (2005)) estimate impulse response functions:
  - Dependent variable: p_{i,t+h} − p_{i,t−1}, where p is log price index (CPI, import price index, or expectations).
  - Key regressor: ΔER_{i,t}, change in log bilateral exchange rate against the US dollar.
  - Coefficient β^h: (percent) response of prices to a one percentage point change in local currency against the US dollar at horizon h months.
  - Controls M_{i,t} include country-specific controls with 12 lags: output gap, lagged inflation, lagged change in exchange rate, and trade-weighted producer price index of export partners.
  - Fixed effects: δ_i (country) and δ_t (time) to control for time-invariant heterogeneity and common time shocks (e.g., VIX, U.S. monetary policy shocks, world energy and food prices).

### Advanced economies vs. EMDEs
- Specification allows differential β^h for advanced economies (AE) and emerging market and developing economies (EMDE):
  - A_E indicator equals 1 for advanced economies (IMF WEO classification) and 0 otherwise.
  - β^h_AE and β^h_EMDE capture pass-through magnitudes at horizons h for average advanced and emerging market countries.

### State-dependent specifications (non-linearity by economic state)
- Four approaches considered to examine pass-through variation with economic state:
  1. Regime dummy (binary) specification:
     - p_{i,t+h} − p_{i,t−1} = β^h_high D_{it} ΔER_{i,t} + β^h_low (1−D_{it}) ΔER_{i,t} + controls + fixed effects.
     - D_{it} = 1 if the state variable (lagged inflation, output gap, uncertainty) is above the sample median or average.
     - Augment controls with D_{it}; β^h_high and β^h_low capture pass-through when state variable is high vs low.
  2. Smooth transition specification (Auerbach and Gorodnichenko (2012); Tenreyro and Thwaites (2016)):
     - p_{i,t+h} − p_{i,t−1} = β^h_low F(z_{it}) ΔER_{i,t} + β^h_high (1−F(z_{it})) ΔER_{i,t} + controls + fixed effects.
     - F(z_{it}) = exp(−γ z_{it}) / (1 + exp(−γ z_{it})), where z_{it} is state variable normalized to zero mean and unit variance (z_{it} = (x_{it} − x̄)/σ_x).
     - Interpretation: F(z_{it}) ≅ 1 when state variable is very low (negative extreme), F(z_{it}) ≅ 0 when state variable is very high (positive extreme). Allows smooth regime weights and a continuum of states.
     - Augment controls to include the smooth transition function.
  3. Non-parametric binning specification:
     - p_{i,t+h} − p_{i,t−1} = β^h_g I[x_{it} ∈ G] ΔER_{i,t} + controls + fixed effects.
     - I is indicator for state variable x_{it} belonging to a specific bin (quartile) G. Does not impose functional form for non-linearity; identifies which specific values of state variables affect pass-through.
  4. (Implicit) Additional augmentations: in all cases set of controls includes 12 lags and fixed effects as above.

### Shock-dependent specifications (sign and size of exchange rate shock)
- Asymmetric pass-through by sign:
  - p_{i,t+h} − p_{i,t−1} = β^h_+ D^+_{it} ΔER_{i,t} + β^h_− (1−D^+_{it}) ΔER_{i,t} + controls + fixed effects.
  - D^+_{it} = 1 for an appreciation of the bilateral exchange rate, 0 otherwise.
- Nonlinearity in shock size (quadratic term):
  - p_{i,t+h} − p_{i,t−1} = β^h ΔER_{i,t} + ϖ^h ΔER_{i,t}^2 + controls + fixed effects.
  - Pass-through elasticity at a given ΔER_{i,t}: ∂(p_{i,t+h} − p_{i,t−1}) / ∂ΔER_{i,t} = β^h + 2 ϖ^h.

### Shock origin: difference-in-differences IV for US monetary policy–driven pass-through
- Goal: isolate pass-through from exchange rate fluctuations caused by US monetary policy tightening using a difference-in-differences instrumental variables approach (Nunn and Qian 2014).
- Instrument construction:
  - Instrument_{i,t} = Quinn_i × Regime_{it} × USMPShock_t.
  - USMPShock_t: externally identified U.S. monetary policy shocks (Jarociński and Karadi (2020)).
  - Regime_{it} = 0 when a country’s exchange rate regime is a pre-announced peg to the US dollar, and 1 otherwise (built using Ilzetzki, Reinhart, and Rogoff (2019) coarse index; euro area countries classified as pre-announced peg due to common currency of Euro are included with additional condition of using US dollar as anchor).
  - Quinn_i = 0 for countries with relatively closed capital accounts, and 1 otherwise; constructed using Quinn and Toyoda (2008) index with threshold at the 10th percentile of cross-country distribution (because sample predominance of highly open capital accounts).
  - Rationale: portfolio investment channel transmits US monetary policy shocks to domestic exchange rates primarily for countries with flexible exchange rates and more open capital accounts.
- Two-stage IV system:
  - First stage: ΔER_{i,t} = β^h_1 Instrument_{i,t} + Σ_{l=0}^{12} θ_l Z^M_{i,t−l} + δ_i + δ_t + ε_{i,t}.
  - Second stage: p_{i,t+h} − p_{i,t−1} = β^h ΔER̂_{i,t} + Σ_{l=0}^{12} θ_l Z^M_{i,t−l} + δ_i + δ_t + ε_{i,t}.
- Identification and controls:
  - Country and time fixed effects imply a difference-in-differences estimator: identification contrasts countries with open capital accounts and flexible exchange rates versus those with relatively closed accounts or a peg to the US dollar.
  - Time fixed effects control for direct effects of US monetary policy shocks on domestic inflation not mediated by exchange rates and other global factors correlated with US monetary policy shocks (including shifts in risk premia).
  - Country fixed effects control for time-invariant unobserved characteristics related to capital account openness.
  - Instrument is constructed to satisfy the exclusion restriction by design and is reported to be strong (F-test reported in the main paper).

*Source: wpiea2023086-print-pdf - Section 3 presents main results and discusses robustness checks. Section 4 concludes.*

### 10.4 and 11.1 for all estimation horizons (Figure 11C).

### 10.4 and 11.1 for all estimation horizons (Figure 11C)

### Baseline results
- A one-percent depreciation of the local currency against the US dollar (equivalent to 0.45 standard deviation of the percent change in the bilateral exchange rate) is associated with sizeable and persistent increases in (log) consumer prices.
- A 10 percent depreciation is associated with:
  - an increase in the level of consumer prices by about 0.5 percent within one month, rising to about 1.6 percent after one year.
- Robustness checks:
  - Using the nominal effective exchange rate instead of the bilateral exchange rate raises pass-through slightly (Figure A4).
  - Estimating with a larger sample of countries almost doubles the size of the effect (Figure A5), driven by higher pass-through in developing economies relative to advanced economies.
  - Results are robust to excluding data outside the 5th and 95th percentiles (versus baseline 1st and 99th) and to alternative lag orders (6 and 24 lags versus baseline 12 lags); similar pass-through coefficients across horizons (Figures A11, A12).
- Transmission channels (Figure 2A–C):
  - Import prices: response reaches 0.7 percent after one month and remains at 0.7 percent one year after the shock.
  - Core prices: pass-through coefficient of 0.08 (about half the level for overall consumer prices).
  - Inflation expectations: peaks at 7 months after the shock; a 10 percent depreciation associated with a 0.7 basis point increase in inflation expectations lasting at least 12 months.

### Heterogeneity across countries
- Advanced economies (AEs) versus emerging market and developing economies (EMDEs):
  - Pass-through to consumer prices is about three times larger in EMs than in AEs (Figure A6); result robust across baseline and large samples.
  - Responses of import prices are similar between AEs and EMDEs, if anything more persistent in AEs (Figure A7).
  - Pass-through to core prices and inflation expectations is markedly different:
    - A 1 percent depreciation is associated with increases in core (expected) inflation of 0.5 (0.3) percentage point in AEs compared to 1.3 (1.2) percentage points in EMDEs (Figures A8, A9).
  - Interpretation: central banks in AEs better anchor inflation expectations and mitigate second-round effects following cost-push shocks.
- Regional differences:
  - Pass-through smallest in Asia and Europe, largest in Latin America, but differences across regions are not statistically significant (Figure A10).

### State-dependent exchange rate pass-through
- Dependence on inflation level (Figure 3, Table 4):
  - Pass-through magnitude increases with the level of inflation.
  - Increase is non-linear: the fourth quartile (initial inflation above 3.8 percent) shows a significantly larger pass-through both economically and statistically.
  - Implication: inflation pressures from current depreciations are larger than previously estimated given recent higher inflation environments.
- Business cycle:
  - Pass-through coefficient not different across rates of GDP growth or different levels of the output gap (Figures 4.1, 4.2; Table 5.1).
  - Pass-through into core prices (Figure A23) and inflation expectations (Figure A31) is stronger when the output gap is larger.
- Uncertainty:
  - Using the Ahir, Bloom, and Furceri (2022) uncertainty index (Figure 5), pass-through is stronger during periods of higher uncertainty; finding robust across specifications.
  - Median uncertainty approached its pandemic high during 2022 after the invasion of Ukraine; uncertainty over 2020–22 approximately double the average over 2000–10 (Figure A3).
- Forecast disagreement:
  - Pass-through significantly higher in countries with higher disagreement on one-year ahead inflation forecasts (Consensus Economics, Figure 6).
  - Largest marginal effects when disagreement exceeds the 75th percentile; associated with periods of high and volatile inflation and low central bank credibility.
- Invoicing currency and dollarization:
  - Pass-through higher in countries with a larger share of imports invoiced in US dollars (Figure 7): rate doubles from 0.1 percent for countries below the sample median to over 0.2 percent for those above the sample median.
  - Import price pass-through reaches 0.8 percent for countries above the median vs 0.5 for those below.
  - In EMDEs, higher dollarization is associated with stronger pass-through (Figure 8); rise is not linear—substantially larger responses when dollarization is above a threshold (supported by Table 9).

### Shock-dependent exchange rate pass-through
- Asymmetry (appreciation vs depreciation):
  - Pass-through materializes faster following depreciations than appreciations over the first six months; impacts converge within about a year (Figure 9).
  - Asymmetry more pronounced in emerging market economies than in advanced economies (Figures 9B vs 9C). Table 10 confirms statistical significance of asymmetry at short horizons.
- Nonlinearity with shock size (Equation (8), Figure 10):
  - Quadratic term coefficient 휗ℎ positive and significant for horizons between 0 and 10 months, implying stronger pass-through for larger exchange rate fluctuations.
  - Economic magnitudes at 12 months:
    - Rate of pass-through rises from 0.16 percent for a depreciation of 1 percent, to 0.175 for a depreciation of 10 percent, and to about 0.2 for a depreciation of 25 percent (Figure 10B).
  - Consistent with prior findings of threshold effects for large depreciations.
- Instrumenting with U.S. monetary policy shocks (difference-in-difference IV approach):
  - First stage: a U.S. monetary policy tightening of 100 basis points causes the exchange rate to depreciate by 4.5 percent more for countries with a flexible exchange rate and open capital account compared to others (Figure 11, Panel A).
  - Interpretation: strong portfolio balance channel—declining interest rate differentials associated with capital outflows and depreciations in more open and flexible countries.
  - Instrument strength: Kleibergen-Paap rk Wald F-statistic varies between 10.4 and 11.1 (Panel C) for horizons between six months and one year.
  - Exclusion restriction test: regressing residuals from equation (1) on the instrument yields OLS estimates for 훽 very close to zero across horizons and not statistically significant (p-value = 1.000), so the null 훽=0 cannot be rejected (Table 11).
  - Second stage: pass-through is much stronger when exchange rate fluctuations are caused by U.S. monetary policy shocks—pass-through is positive and significant by the fifth month, reaching 0.47 after one year (compare to average pass-through coefficient of 0.16 at the same horizon from Figure 1).

### Conclusion and implications
- Recent global environment:
  - Elevated uncertainty (COVID-19 pandemic peak, renewed surge after Russia’s invasion of Ukraine), inflation well above central bank targets in many major economies, and aggressive monetary tightening by the Federal Reserve and European Central Bank have led to strong local currency depreciations against the US dollar in many countries.
- Main findings:
  - Rate of exchange rate pass-through to domestic prices is state-dependent: relatively low on average but significantly larger during periods of high inflation and elevated uncertainty.
  - Pass-through triples when depreciations are driven by U.S. monetary policy tightening.
- Policy implication:
  - Given current global conditions, the magnitude of exchange rate pass-through into prices is likely to be significantly larger than in past decades, implying greater inflationary pressure from recent depreciations.

*IMF Working Paper excerpt: State-Dependent Exchange Rate Pass-Through*

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

### State-Dependent Exchange Rate Pass-Through

### Key findings
- Average effect: The solid blue line in Figure 1 presents the impact of a one percent increase (depreciation) in local currency/USD on consumer prices in percent; confidence bands shown are 90 percent and 95 percent.
- Import prices, core prices, and inflation expectations respond to a one percent depreciation; Figures 2A–2C present the percent change impacts with 90 percent and 95 percent confidence bands.
- Nonlinearity and shock-dependence:
  - Figure 10 shows estimated linear and quadratic coefficients from equation (8), the implied elasticity term 휕Δ휋푡+12/휕Δ퐸푅푡 = 훽12 + v12 ×2×Δ퐸푅푖𝑡, and price response to depreciations between 1 percent and 25 percent.
  - Figure 11 reports shock-dependent pass-through using U.S. monetary policy shocks: Panel A shows first-stage 훽 coefficients (positive coefficient implies a depreciation following a U.S. monetary policy tightening shock); Panel B shows second-stage 훽ℎ coefficients (response of consumer prices following a one percent exchange rate depreciation caused by a shock to U.S. monetary policy); Panel C presents the F-statistic from the first stage.
- Asymmetry: Figure 9 compares pass-through for depreciations vs appreciations in the full sample, advanced economies, and emerging market and developing economies; confidence bands are 68 percent.

### State-dependent heterogeneity (selected dimensions)
- Level of inflation:
  - Figures 3A–3D present pass-through across inflation bins (Median, Mean, Quartiles, Smooth Transition); dark shaded region indicates the 68 percent confidence band.
  - Table 4 reports p-values for tests across horizons 0–12 for four specifications (Median, Mean, Quartiles, Smooth Transition). Selected entries (p-values) at horizons 0–3 (Median): 0 .0474105, 1 .0143056, 2 .0040884, 3 .000735.
- Business cycle (GDP growth and output gap):
  - Figures 4.1 and 4.2 show pass-through across real GDP growth deviations and output gap bins; dark shaded region indicates the 68 percent confidence band.
  - Table 5.1 (GDP growth) and Table 5.2 (output gap) report p-values for horizons 0–12. Example entries (Table 5.1, Median, horizons 0–3): 0 .982985, 1 .4457705, 2 .2865373, 3 .4248433. Example entries (Table 5.2, Median, horizons 0–3): 0 .8934448, 1 .7040491, 2 .3442352, 3 .1578355.
- Uncertainty and inflation disagreement:
  - Figure 5 and Figure 6 display pass-through by monthly uncertainty (Ahir, Bloom, and Furceri (2022)) and disagreement among professional forecasters (Consensus Economics), respectively; 68 percent confidence bands shown.
  - Table 6 (uncertainty) and Table 7 (disagreement) report p-values for horizons 0–12. Example (Table 6, Median horizons 0–3): 0 .5880957, 1 .3843448, 2 .1584718, 3 .1374133. Example (Table 7, Median horizons 0–3): 0 .0630722, 1 .0141429, 2 .0058679, 3 .0030565.
- USD invoice share and dollarization:
  - Figure 7 and Figure 8 show pass-through across bins of USD invoice share (data from Boz and others (2022)) and average dollarization (Reinhart, Rogoff, and Savastano (2003)); 68 percent confidence bands shown.
  - Table 8 (USD invoice share) and Table 9 (dollarization) report p-values for horizons 0–12. Example entries (Table 8, Median horizons 0–3): 0 .1338817, 1 .1017968, 2 .0620568, 3 .0815847. Example entries (Table 9, Median horizons 0–3): 0 .0529531, 1 .0442537, 2 .0487326, 3 .0402635.
- Appreciation vs depreciation:
  - Figure 9 and Table 10 compare pass-through for depreciations vs appreciations; Table 10 reports p-values for horizons 0–12. Selected p-values: horizon 0 .5330642, 1 .3179867, 2 .2368264, 3 .1049418, 4 .0551027, 5 .0322834, 6 .0731297.

### Data, sample, and methodology notes
- Country samples:
  - Baseline advanced economy sample shown in Table 1.1 (examples): Australia Consumer Prices 1990M1-2022M9; Japan Consumer Prices 1990M1-2022M12; United Kingdom Consumer Prices 1990M1-2022M12.
  - Baseline emerging market economy sample shown in Table 1.2 (examples): Argentina Consumer Prices 2012M7-2022M12; China Consumer Prices 1990M1-2022M12; India Consumer Prices 2001M1-2022M12.
- Data sources (Table 2):
  - Bilateral Exchange Rate: Haver Analytics, local currency per US dollar.
  - Consumer Price Index and Import Price Index: Haver Analytics (import prices expressed in local currency).
  - Inflation Expectations: Consensus Economics (forecasts for period average headline CPI inflation).
  - USD Invoice Share: Boz and others (2022).
  - Uncertainty: Ahir, Bloom, and Furceri (2022).
  - Output Gap: World Economic Outlook October 2022 vintage.
  - Dollarization: Reinhart, Rogoff, and Savastano (2003) — index constructed as a weighted average of three indicators.
- Summary statistics (Table 3; N indicates observations):
  - Consumer Prices (y/y % change): N 12,495; Mean 3.08; Std. Dev. 4.30; 25th Percentile 1.09; Median 2.24; 75th Percentile 3.83.
  - Import Prices (y/y % change): N 12,495; Mean 2.78; Std. Dev. 9.39; 25th Percentile -2.31; Median 1.90; 75th Percentile 7.04.
  - Inflation Expectations (% one-year ahead): N 12,495; Mean 3.12; Std. Dev. 4.07; 25th Percentile 1.52; Median 2.32; 75th Percentile 3.52.
  - Exchange Rate (m/m % change): N 12,495; Mean 0.007; Std. Dev. 2.22; 25th Percentile -1.16; Median 0.01; 75th Percentile 1.3.
  - Uncertainty Index: N 12,034; Mean 0.19; Std. Dev. 0.19; 25th Percentile 0.06; Median 0.14; 75th Percentile 0.28.
  - USD Invoice Share (%, country average): N 7,226; Mean 44.01; Std. Dev. 25.73; 25th Percentile 22.90; Median 32.60; 75th Percentile 71.50.
  - Inflation Forecast Disagreement (country average): N 10,992; Mean 1.19; Std. Dev. 4.57; 25th Percentile 0.25; Median 0.38; 75th Percentile 0.51.
  - GDP Growth (%, deviation from country mean): N 12,383; Mean 0.71; Std. Dev. 4.44; 25th Percentile 0.15; Median 0.72; 75th Percentile 1.35.
  - Output Gap (deviation from country mean): N 12,495; Mean -0.01; Std. Dev. 2.56; 25th Percentile -0.96; Median 0.01; 75th Percentile 1.16.
- Instrument validity and exclusion tests:
  - Table 11 reports exclusion restriction test results across horizons h=1 to h=12. Observations per horizon range: 9,658 (h=1) down to 9,597 (h=12). R-squared entries reported as 0.000 across horizons. Instrument coefficient entries and robust standard errors are shown in the table; significance markers indicated (*** p<0.01, ** p<0.05, * p<0.1).

### Robustness and extensions (appendix figures)
- Alternative exchange rate measures and larger samples:
  - Figure A4: nominal effective exchange rate — impact of a one percent depreciation on consumer prices with 90 percent and 95 percent confidence bands.
  - Figure A5: average pass-through to consumer prices using a larger sample of 141 countries.
  - Figure A6: comparisons across Advanced Economies (AEs) and Emerging Market Economies (EMEs) for baseline and large samples.
- Extensive appendices (Figures A7–A36) provide state-dependent pass-through results for import prices, core prices, and inflation expectations across the same dimensions (inflation, business cycle, uncertainty, disagreement, USD invoice share, dollarization), with median/mean/quartiles/smooth-transition presentations and 68 percent confidence bands.
- Additional robustness checks: Figure A11 (alternative trimming of sample) and Figure A12 (alternative lag orders).

*Source: IMF Working Paper — State-Dependent Exchange Rate Pass-Through (figures, tables, and appendix as provided in the supplied PDF).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023086-print-pdf.pdf_