## 1.1    Taper Tantrum Episode:  A Natural Experiment

## Source details

**Canonical URL:** [1.1    Taper Tantrum Episode:  A Natural Experiment](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019075.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wpiea2019075.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019075.pdf.json)

---

### Introduction and Motivation
- Monetary policy normalization in advanced economies tests emerging markets' resilience to external shocks.
- Objective: Investigate whether improvements in macroeconomic policy frameworks—manifested as better-anchored long-term inflation expectations—made emerging markets more resilient to external shocks.
- Empirical strategy: Exploit the exogenous shock of the "taper tantrum" following Federal Reserve Chairman Ben Bernanke’s May 22, 2013 testimony, and a survey-based anchoring index from Bems et al. (2018) for a sample of 18 emerging markets.

### Taper Tantrum: Event and Identification
- Event: Bernanke’s testimony on May 22, 2013 raised the possibility of tapering the Fed’s quantitative easing program, shifting expectations toward earlier U.S. monetary policy normalization.
- Transmission: Triggered global portfolio rebalancing away from emerging-market assets, generating capital outflows and currency depreciations—a sudden stop in capital flows.
- Identification: Treats the taper tantrum as exogenous to emerging economies; exploits cross-sectional variation in the anchoring index to estimate heterogeneous effects.

### Measurement of Inflation Expectations’ Anchoring
- Operational characteristics defining anchoring:
  - Long-horizon forecasts centered around the target and unresponsive to transitory fluctuations.
  - Minor revisions of long-term forecasts in response to news.
  - Low dispersion across individual long-term forecasts.
- Anchoring index construction:
  - From survey-based long-term inflation forecasts (three-years ahead and above) of professional forecasters.
  - Four complementary metrics standardized (mean zero, standard deviation one) and averaged; higher values denote better anchoring.
  - Sample: 18 emerging markets reporting long-term expectations in Consensus Economics (three-year-ahead and longer), excluding Ukraine.
  - Sample period for index averages shown: 2002–17.
  - Cross-country heterogeneity: Hungary, Malaysia, Poland, and Chile above the 75th percentile; Turkey, Romania, Bulgaria, Russia, and Argentina below the 25th percentile.

### Empirical Framework and Estimation
- Local projection specification (Jordá, 2005):
  - y_{i,t+h} = α_{ih} + φ_h ∑_{k=1}^3 y_{i,t−k} + β_h taper_t + θ_h anchoring_{it} + γ_h anchoring_i × taper_t + ε_{i,t+h}
  - y_{i,t+h}: monthly changes in output growth forecasts, consumer prices (CPI), or nominal bilateral exchange rate vs. US dollar.
  - taper_t: dummy = 1 in May 2013, 0 otherwise.
  - anchoring_{it}: anchoring index (three-year-ahead focus).
  - γ_h measures differential effect across anchoring levels.
- Group definitions:
  - Strongly anchored: anchoring index above the 75th percentile.
  - Poorly anchored: anchoring index below the 25th percentile.
- Dependent variable transformations:
  - For exchange rate and CPI: y_{i,t+h} = (y_{i,t+h} − y_{i,t}) / y_{i,t}.
  - For growth forecasts: y_{i,t+h} = y_{i,t+h} − y_{i,t}.
- Exchange rate pass-through: ratio of the impulse response of prices to that of the exchange rate at each horizon.
- Estimation details:
  - Monthly frequency over 2002–17.
  - Three lags of the annualized change in the dependent variable included.
  - Country fixed effects included.
  - Standard errors corrected for serial correlation, heteroskedasticity, and cross-sectional dependence using Driscoll and Kraay (1998).
  - Data sources: growth forecasts from Consensus Economics; CPI from Haver Analytics; nominal bilateral exchange rate vs. US dollar from IMF IFS.

### Main Results — High-level Findings
- Growth forecasts:
  - The taper tantrum announcement led to a deterioration of growth prospects in emerging markets.
  - Output forecasts declined sharply and to comparable degrees in both strongly-anchored and weakly-anchored countries on impact.
- Exchange rate response:
  - Currencies depreciated on impact in both groups.
  - Initial three-month depreciation:
    - Weakly-anchored group: 3.2 percent after three months.
    - Strongly-anchored group: 4.7 percent after three months.
  - After the third month, the extent of depreciation converged to the same level across groups.
  - Interpretation: Smaller initial depreciation in less-anchored countries consistent with “fear of floating” due to currency mismatches and high pass-through.
- Consumer price inflation response:
  - Weakly-anchored economies experienced a very persistent and statistically significant increase in CPI.
  - Strongly-anchored economies experienced a significantly smaller CPI response.
  - The difference between groups is statistically significant at all horizons.
- Exchange rate pass-through:
  - Implied pass-through during the taper tantrum:
    - Non-anchored (weakly-anchored) group: almost 20 percent.
    - Anchored (strongly-anchored) group: less than 10 percent.
  - Consistent with the view that improvements in monetary performance and better-anchored expectations reduce exchange rate pass-through.
- Robustness:
  - Results robust to inclusion of the policy rate as a control variable.

### Appendix A — Key Estimation Results (selected coefficients and diagnostics)
- Response of Output Growth (Table A.1):
  - Sample: Countries: 18 (for horizons h = 1 to 6); Observations: 2,688; 2,670; 2,652; 2,634; 2,616; 2,598.
  - R-squared: 0.287; 0.312; 0.316; 0.317; 0.320; 0.328.
  - Taper dummy coefficients (h = 1 to 6):
    - h= 1: -3.636** (1.408)
    - h= 2: -9.602*** (3.319)
    - h= 3: -12.854** (5.566)
    - h= 4: -23.550*** (8.030)
    - h= 5: -22.755** (10.633)
    - h= 6: -23.115* (13.114)
  - Taper dummy × anchoring index (h = 1 to 6):
    - h= 1: -2.941 (1.900)
    - h= 2: -6.562 (4.002)
    - h= 3: -18.053*** (6.459)
    - h= 4: -15.119* (8.910)
    - h= 5: -11.105 (11.082)
    - h= 6: 6.077 (13.284)
  - Anchoring index (h = 1 to 6):
    - h= 1: -0.693 (1.986)
    - h= 2: -2.467 (3.643)
    - h= 3: -4.689 (5.408)
    - h= 4: -6.683 (7.420)
    - h= 5: -8.938 (9.717)
    - h= 6: -11.565 (11.992)
- Response of Consumer Price Inflation (Table A.2):
  - Sample: Countries: 18; Observations: 2,910; 2,892; 2,874; 2,856; 2,838; 2,820.
  - R-squared: 0.457; 0.548; 0.600; 0.633; 0.658; 0.678.
  - Taper dummy coefficients (h = 1 to 6):
    - h= 1: 0.170*** (0.015)
    - h= 2: 0.372*** (0.031)
    - h= 3: 0.405*** (0.049)
    - h= 4: 0.490*** (0.069)
    - h= 5: 0.503*** (0.092)
    - h= 6: 0.567*** (0.120)
  - Taper dummy × anchoring index (h = 1 to 6):
    - h= 1: -0.037 (0.038)
    - h= 2: -0.409*** (0.090)
    - h= 3: -0.532*** (0.157)
    - h= 4: -0.549** (0.231)
    - h= 5: -0.838*** (0.309)
    - h= 6: -1.062*** (0.388)
  - Anchoring index (h = 1 to 6):
    - h= 1: -0.106*** (0.035)
    - h= 2: -0.209*** (0.075)
    - h= 3: -0.324*** (0.123)
    - h= 4: -0.458** (0.182)
    - h= 5: -0.601** (0.251)
    - h= 6: -0.754** (0.328)
- Response of Exchange Rate Depreciation (Table A.3):
  - Sample: Countries: 18; Observations: 2,910; 2,892; 2,874; 2,856; 2,838; 2,820.
  - R-squared: 0.021; 0.025; 0.032; 0.041; 0.051; 0.062.
  - Taper dummy coefficients (h = 1 to 6):
    - h= 1: 2.270*** (0.166)
    - h= 2: 2.959*** (0.346)
    - h= 3: 4.090*** (0.518)
    - h= 4: 4.311*** (0.688)
    - h= 5: 2.704*** (0.852)
    - h= 6: 4.136*** (1.054)
  - Taper dummy × anchoring index (h = 1 to 6):
    - h= 1: 2.028*** (0.177)
    - h= 2: 1.821*** (0.366)
    - h= 3: 2.135*** (0.562)
    - h= 4: 0.595 (0.772)
    - h= 5: -1.212 (0.983)
    - h= 6: -1.163 (1.198)
  - Anchoring index (h = 1 to 6):
    - h= 1: 0.385 (0.307)
    - h= 2: 0.807 (0.636)
    - h= 3: 1.256 (0.971)
    - h= 4: 1.846 (1.259)
    - h= 5: 2.604* (1.488)
    - h= 6: 3.335* (1.693)

### Conclusions and Policy Implications
- Main conclusion: Better-anchored long-term inflation expectations make emerging economies more resilient to external shocks that trigger sudden stops in capital flows (such as the taper tantrum).
- Key implications:
  - The taper tantrum was an adverse external shock that deteriorated growth prospects and induced currency depreciation for emerging markets irrespective of anchoring strength.
  - Macroeconomic trade-offs differed by anchoring level:
    - Strongly-anchored countries: smaller inflation response, lower exchange rate pass-through, lower inflation persistence—easier monetary policy trade-offs.
    - Weakly-anchored countries: larger inflation response despite sometimes smaller initial depreciation—starker dilemma between fighting recession and containing higher inflation.
- Policy message: Strengthening credibility and the anchoring of long-term inflation expectations can reduce exchange rate pass-through and inflation persistence, improving resilience to sudden external shocks.

*Source: IMF working paper section "1.1    Taper Tantrum Episode:  A Natural Experiment" (wpiea2019075).*

### 1.1    Taper Tantrum Episode:  A Natural Experiment    .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .  .4

### 1.1    Taper Tantrum Episode:  A Natural Experiment

### Introduction
- Context: Monetary policy normalization in advanced economies tests emerging markets' resilience to external shocks.
- Motivation: Investigate whether improvements in macroeconomic policy frameworks—manifested as better-anchored long-term inflation expectations—made emerging markets more resilient to external shocks.
- Empirical strategy: Exploit the exogenous shock of the "taper tantrum" following Federal Reserve Chairman Ben Bernanke’s May 22, 2013 testimony, and a survey-based anchoring index from Bems et al. (2018) for a sample of 18 emerging markets.

### Taper Tantrum Episode: Description and Identification
- Event: Bernanke’s testimony on May 22, 2013 raised the possibility of tapering the Fed’s quantitative easing program, shifting expectations toward earlier U.S. monetary policy normalization.
- Transmission: Triggered global portfolio rebalancing away from emerging-market assets, generating capital outflows and currency depreciations—a sudden stop in capital flows.
- Identification: The taper tantrum is treated as exogenous to emerging economies; cross-sectional variation in the anchoring index is exploited to estimate heterogeneous effects.

### Inflation Expectations’ Anchoring: Measurement
- Operational characteristics used to define anchoring:
  - Long-horizon forecasts centered around the target and unresponsive to transitory fluctuations.
  - Minor revisions of long-term forecasts in response to news.
  - Low dispersion across individual long-term forecasts.
- Anchoring index:
  - Constructed by Bems et al. (2018) from survey-based long-term inflation forecasts (three-years ahead and above) of professional forecasters.
  - For each country, four complementary metrics are standardized (mean zero, standard deviation one) and averaged to form the anchoring index; higher values denote better anchoring.
  - Sample: 18 emerging markets reporting long-term expectations in Consensus Economics (three-year-ahead and longer), excluding Ukraine.
  - Sample period for index averages shown: 2002–17.
  - Cross-country heterogeneity: Hungary, Malaysia, Poland, and Chile above the 75th percentile; Turkey, Romania, Bulgaria, Russia, and Argentina below the 25th percentile.

### Empirical Framework
- Local projection specification (Jordá, 2005):
  - y_{i,t+h} = α_{ih} + φ_h ∑_{k=1}^3 y_{i,t−k} + β_h taper_t + θ_h anchoring_{it} + γ_h anchoring_i × taper_t + ε_{i,t+h}
  - y_{i,t+h}: monthly changes in output growth forecasts, consumer prices (CPI), or nominal bilateral exchange rate vs. US dollar.
  - taper_t: dummy = 1 in May 2013, 0 otherwise.
  - anchoring_{it}: anchoring index (three-year-ahead focus).
  - γ_h on interaction measures differential effect across anchoring levels.
- Group definitions:
  - Strongly anchored: anchoring index above the 75th percentile.
  - Poorly anchored: anchoring index below the 25th percentile.
- Dependent variable transformations:
  - For exchange rate and CPI: y_{i,t+h} = (y_{i,t+h} − y_{i,t}) / y_{i,t}.
  - For growth forecasts: y_{i,t+h} = y_{i,t+h} − y_{i,t}.
- Exchange rate pass-through: ratio of the impulse response of prices to that of the exchange rate at each horizon.
- Estimation details:
  - Monthly frequency over 2002–17.
  - Three lags of the annualized change in the dependent variable included.
  - Country fixed effects included.
  - Standard errors corrected for serial correlation, heteroskedasticity, and cross-sectional dependence using Driscoll and Kraay (1998) procedure.
  - Data sources: growth forecasts from Consensus Economics; CPI from Haver Analytics; nominal bilateral exchange rate vs. US dollar from IMF IFS.

### Results
- Growth forecasts:
  - The taper tantrum announcement led to a deterioration of growth prospects in emerging markets.
  - Output forecasts declined sharply and to comparable degrees in both strongly-anchored and weakly-anchored countries on impact (Figure 3).
- Exchange rate response:
  - Currencies depreciated on impact in both groups.
  - Initial three-month depreciation:
    - Weakly-anchored group: 3.2 percent after three months.
    - Strongly-anchored group: 4.7 percent after three months.
  - After the third month, the extent of depreciation converged to the same level across groups.
  - Interpretation: Smaller initial depreciation in less-anchored countries consistent with “fear of floating” (Calvo and Reinhart, 2002) due to currency mismatches and high pass-through.
- Consumer price inflation response:
  - Weakly-anchored economies experienced a very persistent and statistically significant increase in CPI.
  - Strongly-anchored economies experienced a significantly smaller CPI response.
  - The difference between groups is statistically significant at all horizons.
- Exchange rate pass-through:
  - Implied pass-through during the taper tantrum:
    - Non-anchored (weakly-anchored) group: almost 20 percent.
    - Anchored (strongly-anchored) group: less than 10 percent.
  - Method: country-level clustered standard errors computed simultaneously estimating the (co)variance of the exchange rate and CPI models; ratio computed and 95 percent confidence intervals built.
  - Consistent with Taylor (2000): improvements in monetary performance and better-anchored expectations reduce exchange rate pass-through.
- Robustness:
  - Results robust to inclusion of the policy rate as a control variable.

### Conclusions
- Main conclusion: Better-anchored long-term inflation expectations make emerging economies more resilient to external shocks that trigger sudden stops in capital flows (such as the taper tantrum).
- Key implications:
  - The taper tantrum was an adverse external shock that deteriorated growth prospects and induced currency depreciation for emerging markets irrespective of anchoring strength.
  - The macroeconomic trade-off faced by central banks differed by anchoring level:
    - Strongly-anchored countries: smaller inflation response, lower exchange rate pass-through, lower inflation persistence—easier monetary policy trade-offs.
    - Weakly-anchored countries: larger inflation response despite sometimes smaller initial depreciation—starker dilemma between fighting recession and containing higher inflation.
- Policy message: Strengthening credibility and the anchoring of long-term inflation expectations can reduce exchange rate pass-through and inflation persistence, improving resilience to sudden external shocks.

*Source: IMF working paper section "1.1    Taper Tantrum Episode:  A Natural Experiment" (wpiea2019075).*

### Appendix A.  Estimation Results

### Appendix A. Estimation Results

### Response of Output Growth (Table A.1)
- Sample and model:
  - Countries: 18 18 18 18 18 18
  - Observations: 2,688 2,670 2,652 2,634 2,616 2,598
  - R-squared: 0.287 0.312 0.316 0.317 0.320 0.328
  - Notes: all regressions include three lags of the change in consumer prices and country fixed effects. Driscoll-Kraay standard errors in parentheses. ***p <0.01, **p <0.05, *p <0.1.
- Coefficient estimates by horizon (h = 1 to 6):
  - Taper dummy:
    - h= 1: -3.636** (1.408)
    - h= 2: -9.602*** (3.319)
    - h= 3: -12.854** (5.566)
    - h= 4: -23.550*** (8.030)
    - h= 5: -22.755** (10.633)
    - h= 6: -23.115* (13.114)
  - Taper dummy × anchoring index:
    - h= 1: -2.941 (1.900)
    - h= 2: -6.562 (4.002)
    - h= 3: -18.053*** (6.459)
    - h= 4: -15.119* (8.910)
    - h= 5: -11.105 (11.082)
    - h= 6: 6.077 (13.284)
  - Anchoring index:
    - h= 1: -0.693 (1.986)
    - h= 2: -2.467 (3.643)
    - h= 3: -4.689 (5.408)
    - h= 4: -6.683 (7.420)
    - h= 5: -8.938 (9.717)
    - h= 6: -11.565 (11.992)

### Response of Consumer Price Inflation (Table A.2)
- Sample and model:
  - Countries: 18 18 18 18 18 18
  - Observations: 2,910 2,892 2,874 2,856 2,838 2,820
  - R-squared: 0.457 0.548 0.600 0.633 0.658 0.678
  - Notes: all regressions include three lags of the change in consumer prices and country fixed effects. Driscoll-Kraay standard errors in parentheses. ***p <0.01, **p <0.05, *p <0.1.
- Coefficient estimates by horizon (h = 1 to 6):
  - Taper dummy:
    - h= 1: 0.170*** (0.015)
    - h= 2: 0.372*** (0.031)
    - h= 3: 0.405*** (0.049)
    - h= 4: 0.490*** (0.069)
    - h= 5: 0.503*** (0.092)
    - h= 6: 0.567*** (0.120)
  - Taper dummy × anchoring index:
    - h= 1: -0.037 (0.038)
    - h= 2: -0.409*** (0.090)
    - h= 3: -0.532*** (0.157)
    - h= 4: -0.549** (0.231)
    - h= 5: -0.838*** (0.309)
    - h= 6: -1.062*** (0.388)
  - Anchoring index:
    - h= 1: -0.106*** (0.035)
    - h= 2: -0.209*** (0.075)
    - h= 3: -0.324*** (0.123)
    - h= 4: -0.458** (0.182)
    - h= 5: -0.601** (0.251)
    - h= 6: -0.754** (0.328)

### Response of Exchange Rate Depreciation (Table A.3)
- Sample and model:
  - Countries: 18 18 18 18 18 18
  - Observations: 2,910 2,892 2,874 2,856 2,838 2,820
  - R-squared: 0.021 0.025 0.032 0.041 0.051 0.062
  - Notes: all regressions include three lags of the change in consumer prices and country fixed effects. Driscoll-Kraay standard errors in parentheses. ***p <0.01, **p <0.05, *p <0.1.
- Coefficient estimates by horizon (h = 1 to 6):
  - Taper dummy:
    - h= 1: 2.270*** (0.166)
    - h= 2: 2.959*** (0.346)
    - h= 3: 4.090*** (0.518)
    - h= 4: 4.311*** (0.688)
    - h= 5: 2.704*** (0.852)
    - h= 6: 4.136*** (1.054)
  - Taper dummy × anchoring index:
    - h= 1: 2.028*** (0.177)
    - h= 2: 1.821*** (0.366)
    - h= 3: 2.135*** (0.562)
    - h= 4: 0.595 (0.772)
    - h= 5: -1.212 (0.983)
    - h= 6: -1.163 (1.198)
  - Anchoring index:
    - h= 1: 0.385 (0.307)
    - h= 2: 0.807 (0.636)
    - h= 3: 1.256 (0.971)
    - h= 4: 1.846 (1.259)
    - h= 5: 2.604* (1.488)
    - h= 6: 3.335* (1.693)

*Appendix A. Estimation Results — Tables A.1–A.3 (regression coefficients, standard errors, sample sizes, and notes).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019075.pdf_