## 3. CREdIbILITy, COMMUNICATION,  ANd MONETARy pOLICy pROCyCLICALITy IN LATIN AMERICA

## Source details

**Canonical URL:** [3. CREdIbILITy, COMMUNICATION,  ANd MONETARy pOLICy pROCyCLICALITy IN LATIN AMERICA](https://www.imf.org/-/media/files/publications/reo/whd/2018/may/wreo0518-chp3.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/reo/whd/2018/may/wreo0518-chp3.pdf.md)
- [Structured JSON version](/-/media/files/publications/reo/whd/2018/may/wreo0518-chp3.pdf.json)

---

### Procyclicality of monetary policy and role of credibility
- LA5 central banks, with the notable exception of Chile, increased policy interest rates despite weak growth, amounting to some degree of procyclicality (Végh and others 2017).
- Possible contributors to monetary policy procyclicality:
  - Central bank credibility—reflected by the degree of anchoring in inflation expectations.
  - Degree of financial dollarization.
  - Central bank independence, governance, and institutional quality.
  - Delays in fiscal consolidation amid widening external imbalances in some LA5 economies.
- Survey-based medium-term inflation expectations in LA5 (except Chile) remained above the midpoint of central bank targets for a prolonged period.
- Market-based inflation expectations show similar dynamics.

### Evidence: credibility, inflation expectations, and policy responses (IPVAR results)
- Framework:
  - Sample: 20 inflation-targeting economies exposed to protracted terms-of-trade shocks.
  - Method: panel vector autoregression with interaction terms (IPVAR).
- Key quantified findings:
  - Shock considered: a 20 percent decline in the terms of trade.
  - Economies with initial medium-term inflation expectation gaps above the 75th percentile saw a widening of the expectation gap of 30 basis points following the shock.
  - Economies with inflation gaps below the 50th percentile saw no significant widening of inflation expectation gaps following the same shock.
- Policy-rate responses conditional on credibility:
  - Where deterioration of inflation expectations is greater, policy rates rose more aggressively.
  - Economies with better initial anchoring lowered their policy rates despite higher observed inflation.
- Real effects:
  - Higher procyclicality in countries with lower central bank credibility exacerbated the effect of the terms-of-trade shock on domestic demand.

### Central bank transparency and link to credibility
- Transparency is associated with greater central bank credibility and lower monetary policy procyclicality.
- Essential transparency elements:
  1. A formal policy objective such as price stability (including explicit quantification).
  2. An assessment of the current state of the economy.
  3. An explanation of policy decisions.
  4. A forward-looking analysis.
  5. Publication of the economic data and forecasts used in the central bank’s assessment.
- LA5 transparency features and deficiencies:
  - Policy rate decisions accompanied by press releases explaining the decision and assessing the balance of risks for inflation.
  - Baseline scenarios and balance of risks delineated in quarterly monetary policy reports.
  - Release of data used for policy (output gaps, inflation, inflation expectations, wages, employment, and GDP).
  - All LA5 central banks, except the Central Reserve Bank of Peru, release minutes of policy meetings before the subsequent meeting.
  - Names not assigned to transcribed comments; votes attributed only in Chile and Brazil.
  - Central bank governors summoned to periodic parliamentary hearings in Brazil, Chile, Colombia, Mexico.
  - Some central banks publish transcripts of monetary policy decision meetings with a long lag (Chile).
  - Deficiencies: gaps in operational transparency, notably lack of assessments of forecasting and operational performance.
- Empirical results using the Dincer and Eichengreen (2014) transparency index:
  - Conditional IPVAR responses show lower transparency is associated with inflation expectation gaps that widen significantly after inflationary shocks.
  - Gains from increasing central bank transparency display diminishing returns; largest benefits accrue at low levels of transparency.
  - Many Latin America central banks, characterized by lower transparency levels, stand to benefit from expanding transparency frameworks.

### Quantified transparency–policy links and macroeconomic consequences
- Policy-rate procyclicality by transparency level (illustrative findings):
  - Central banks with the current average transparency score of LA5 central banks increased the policy rate by 50 basis points for each 100 basis point increase in inflation following the decline in the terms of trade.
  - In contrast, a country with Australia’s current level of transparency kept the policy rate unchanged following the same shock.
- Macroeconomic consequences:
  - Tighter policy responses in less-transparent settings increase macroeconomic volatility, since inflation and output move in opposite directions after a terms-of-trade decline.
- Note: Transparency scores cited refer to 2014 (Dincer and Eichengreen 2014).

### Communication quality, predictability, and surprises
- Short-term predictability:
  - Short-term predictability of interest rate decisions is low in Latin America, with the notable exception of Chile.
  - Forecast errors for Colombia and Brazil are the largest in a sample of 18 inflation-targeting economies.
  - Monetary policy surprises are most frequent in Latin America; since 2010 the Central Bank of Colombia “surprised” markets once every five meetings.
  - The frequency of monetary policy surprises has increased in recent years for all LA5 central banks except that of Brazil.
- Measurement of predictability:
  - Short-term central bank predictability measured using surveys of financial market analysts gathered the day before each monetary policy decision; the difference between expectation and outcome treated as a forecast error or monetary policy surprise.
- Communication quality principle:
  - Emphasize quality over quantity: strengthen the signal-to-noise ratio; clearer, unambiguous communication improves predictability and credibility (Blinder 2009 framing).

### Transparency, text length, readability, tone, and links to market outcomes
- Press-release length and content:
  - Word count is a quantifiable indicator of communication effort and procedural transparency; Mexico and Brazil exhibit big increases in press-release word counts in recent years.
  - Longer statements do not necessarily lead to clearer communication; verbosity can coexist with low information content.
  - In Brazil, prior to 2016, explanations for Monetary Policy Committee decisions were presented in minutes published with a two-week lag rather than in contemporaneous statements.
- Readability measures:
  - Flesch reading ease (RE) for Brazil (English) and Flesch-Szigriszt index for Spanish press releases (Chile, Colombia, Mexico, Peru).
  - Formulae:
    - Flesch (English): RE = 206.835 – (1.015 x ASL) – (84.6 x ASW)
    - Flesch-Szigriszt (Spanish): RE = 206.835 – (ASL) – (84.6 x ASW)
  - Improvements in clarity occurred in Chile and, to a lesser extent, in Colombia.
  - Press statements in Chile, Colombia, and Peru use similar language complexity as the business section of local newspapers; Brazil and Mexico use more complicated language.
  - Longer central bank statements are associated with lower readability (negative correlation).
  - More readable press statements are associated with lower monetary policy forecast errors; cross-sample adjusted R 2 = 0.38.
- Tone and market reactions:
  - Computational linguistic measures summarize tone as “hawkish” or “dovish.”
  - LA5 central banks tend to back up words with deeds: more “hawkish” (“dovish”) tones tend to result in a tightening (loosening) of the policy rate in subsequent meetings.
  - Tone in documents from Colombia and Mexico are reliable predictors of future policy changes.
  - Market rates respond to tone: in Chile, tone in press releases explains a significant share of variation in market rates (particularly short-term rates); tone in minutes has a negligible effect.
  - Markets show some sensitivity to tone in minutes in Brazil and Mexico, and to a lesser extent in Colombia.
  - Inclusion of tone indices in regressions of daily changes in market rates increases explanatory power relative to regressions using only unexpected monetary policy changes.

### Forward guidance and transmission to inflation expectations
- Use of forward-looking language:
  - Forward-looking language (explicit guidance on likely future direction of monetary policy, e.g., an easing or tightening “bias”) was used infrequently by Latin American central banks during 2011–17, with the notable exception of the Central Bank of Chile.
  - The Central Bank of Chile included a policy “bias” at a rate of once every two meetings.
- Effect on transmission:
  - Announcements that contain an explicit policy “bias” appear associated with higher transmission from policy rates to long-term inflation compensation measures.
  - Empirical identification: effect of an unanticipated increase in policy rate on breakeven inflation (one-day change in the difference between yield on 10-year nominal and inflation-linked government bonds) shows a larger total effect when guidance is present versus the effect of the policy change alone.

### Policy takeaways and recommendations
- Credibility (measured by the absolute difference between medium-term inflation expectations and the midpoint of central bank targets) materially affects policy responses to short-term inflationary shocks.
- Credibility is strongly related to transparency and communication quality.
- Scope to increase central bank transparency in the region to enhance credibility, better anticipate decisions, align medium-term inflation expectations with central bank objectives, strengthen policy transmission, and affect procyclicality/countercyclicality.
- Practical steps recommended for LA5 and the region:
  - Fill data gaps, for example by increasing the horizon of survey-based expectations.
  - Consider publishing votes and comments of individual committee members.
  - Central banks that do not publish minutes of policy meetings should consider publishing them.
  - Reduce the lag in the publication of minutes.
- Communication strategy matters: clear, unambiguous communication and explicit conditional forward guidance improve transmission to long-term inflation compensation measures.
- Caveat: Transparency and communication are not a panacea; central banks should tailor strategies aligned to policy objectives.

### Box 3.1 — Measuring inflation expectations from financial instruments and forecast accuracy
- Limitations of surveys:
  - Surveys of professional forecasters are infrequent, subject to time lag, limited horizons, and documented bias and inertia.
- Affine bond-yield framework (Gimeno and Marqués 2012):
  - Uses standard nominal bonds with factors that include observed inflation and zero-coupon yield curve parameters.
  - Decomposes nominal rates into real risk-free rates, expected inflation, and risk premium.
  - Applied to government bond data from Brazil, Chile, Colombia, Mexico to compute inflation expectations at one-year, five-year, and ten-year horizons.
- Model findings:
  - Brazil and Colombia show similar patterns across horizons.
  - Chile and Mexico show more volatility at the one-year horizon and little change at longer horizons.
  - Most countries show long-horizon expectations within central bank target windows, though Brazil and Colombia had periods with expectations above these limits before a large decrease since the beginning of 2016.
  - Mexico shows long-term expectations slightly above the window limit of 4 percent, mainly related to peso movements and uncertainty about trade relations with the United States.
  - For Brazil, the 2015–16 recession affected expectations, which decreased since the beginning of 2016 and then began increasing for longer horizons at the end of 2016.
  - In Colombia, policy hikes in 2016 appear to have anchored inflation expectations closer to the central bank target.
  - Chile experienced decreasing short-term expectations implicit in debt markets since mid-2014; long-term expectations broadly aligned with the medium-term target.
- Forecast accuracy (Table 3.1.1: ratio of MSE of expected inflation from surveys and model with respect to MSE of using current inflation):
  - Brazil; Sample: February 2007–October 2016; Survey: 0.5833; Model: 0.8812; Survey-Model: 0.6178
  - Chile; Sample: July 2012–December 2016; Survey: 0.7813; Model: 0.6344; Survey-Model: 0.6187
  - Colombia; Sample: February 2005–November 2016; Survey: 0.7956; Model: 0.9356; Survey-Model: 0.7898
  - Mexico; Sample: May 2011–November 2016; Survey: 0.6350; Model: 0.7078; Survey-Model: 0.6349
  - Note on horizons: Expected inflation is 12 months for Brazil, Colombia, and Mexico; 11 months for Chile.
- Combining forecasts:
  - A simple average of survey and model expectations provides lower MSE for all countries except Brazil.
- Methodological note:
  - The affine-model-based expectations can be compared against survey-based expectations by evaluating forecasting accuracy (ratio of mean square errors); the model allows obtaining inflation expectations net of risk premiums by decomposing nominal interest rates.

### The Central Bank of Chile: communication, credibility, and practices (Box excerpt)
- Institutional context:
  - Chile adopted a free-floating exchange rate and an inflation-targeting regime in September 1999.
  - Upon adopting inflation targeting in 2000, the CBC adopted a band of 2 to 4 percent for CPI inflation over a 12- to 24-month horizon; since 2007 the CBC stressed keeping annual CPI inflation at around 3 percent most of the time, within a range of plus or minus 1 percentage point, with a policy horizon of around two years.
- Outcomes:
  - Inflation since 2000 has averaged 3.2 percent.
  - Medium-term expectations have been well anchored around the central bank’s target.
- Recent communication reforms (implemented September 2017):
  - Changes: reducing the frequency of monetary policy meetings; aligning meetings with the release of the Monetary Policy Report; including additional information in statements (vote tally, main arguments of Board members, and macro context); introducing publication of full meeting transcripts with a 10-year lag.
  - Net effect: reduce the number of major communication events from 16 to 8 a year and increase informational content of each event.
  - Estimated CBC transparency score following reforms: 11.5 on the Dincer and Eichengreen (2014) scale.
  - The statement issued February 1, 2018, was 3.5 times longer than the average statement in 2017.
- Key outlets:
  - Monetary Policy Report: released in Spanish and English each quarter, or at every other policy meeting; states policy objective to return inflation to target within the policy horizon of two years; communicates key forecasts and fan charts and (infrequently) estimates of unobservables such as potential growth.
  - Statements: published in Spanish following each monetary policy meeting; almost half the statements issued since 2011 have included explicit conditional guidance about the likely future path of monetary policy.
  - Minutes: released with a two-week delay since 2006; include additional detail about meeting discussion, outlook, policy alternatives, board arguments, and vote balance; minutes have usually had only a very small market impact.
- Consideration on publishing a forward path:
  - Publishing a forward path could offer additional information and precision but must avoid confusion about the conditional nature of the path and the central bank’s commitment; it can expose the central bank to credibility concerns if market expectations do not align with the published forward path.

*Source: Chapter 2 of the April 2016 Regional Economic Outlook: Western Hemisphere (excerpt provided).*

### Chapter 2 of the April 2016 Regional Economic

### Chapter 2 of the April 2016 Regional Economic Outlook: Western Hemisphere

### Procyclicality of Monetary Policy in LA5 and Role of Credibility
- During the period covered, LA5 central banks, with the notable exception of Chile, opted to increase policy interest rates despite weak growth.
- Raising rates when growth is weak amounted to some degree of procyclicality in the monetary policy response (Végh and others 2017).
- Several factors might have contributed to monetary policy procyclicality, including:
  - Central bank credibility—reflected by the degree of anchoring in inflation expectations.
  - Degree of financial dollarization.
  - Central bank independence, governance, and institutional quality.
  - Delays in fiscal consolidation amid widening external imbalances in some LA5 economies.
- Survey-based medium-term inflation expectations in LA5 (except Chile) remained above the midpoint of central bank targets for a prolonged period.
- Market-based inflation expectations show similar dynamics (see Box 3.1 in the source).

### Evidence: Credibility, Inflation Expectations, and Policy Responses
- The chapter compares 20 inflation-targeting economies exposed to protracted terms-of-trade shocks.
- Methodology: panel vector autoregression with interaction terms (IPVAR) to estimate responses to shocks and how they vary with initial inflation-expectation gaps (a measure of central bank credibility).
- Key quantified findings:
  - Scenario: a 20 percent decline in the terms of trade.
  - Economies with initial medium-term inflation expectation gaps above the 75th percentile saw a widening of the expectation gap of 30 basis points following the shock.
  - Economies with inflation gaps below the 50th percentile saw no significant widening of inflation expectation gaps following the same shock.
- Policy-rate responses conditional on credibility:
  - Where deterioration of inflation expectations is greater, policy rates rose more aggressively.
  - Economies with better initial anchoring lowered their policy rates despite higher observed inflation.
- Real effects:
  - Higher procyclicality in countries with lower central bank credibility exacerbated the effect of the terms-of-trade shock on domestic demand.

### Central Bank Transparency and Its Link to Credibility
- Transparency is associated with greater central bank credibility and lower monetary policy procyclicality.
- Essential transparency elements identified:
  1. A formal policy objective such as price stability (including explicit quantification).
  2. An assessment of the current state of the economy.
  3. An explanation of policy decisions.
  4. A forward-looking analysis.
  5. Publication of the economic data and forecasts used in the central bank’s assessment.
- LA5 transparency features:
  - Policy rate decisions accompanied by press releases explaining the decision and assessing the balance of risks for inflation.
  - Baseline scenarios and balance of risks delineated in quarterly monetary policy reports.
  - Release of data used for policy (output gaps, inflation, inflation expectations, wages, employment, and GDP).
  - All LA5 central banks, except the Central Reserve Bank of Peru, release minutes of policy meetings before the subsequent meeting.
  - Names not assigned to transcribed comments; votes attributed only in Chile and Brazil.
  - Central bank governors summoned to periodic parliamentary hearings in Brazil, Chile, Colombia, Mexico.
  - Some central banks publish transcripts of monetary policy decision meetings with a long lag (Chile).
  - Deficiencies: gaps in operational transparency, notably lack of assessments of forecasting and operational performance.
- Empirical results using the Dincer and Eichengreen (2014) transparency index:
  - Conditional IPVAR responses show lower transparency is associated with inflation expectation gaps that widen significantly after inflationary shocks.
  - Gains from increasing central bank transparency display diminishing returns; largest benefits accrue at low levels of transparency.
  - Many Latin America central banks, characterized by lower transparency levels, stand to benefit from expanding transparency frameworks.

### Quantified Transparency–Policy Links and Macroeconomic Effects
- Policy-rate procyclicality by transparency level (illustrative findings):
  - Central banks with the current average transparency score of LA5 central banks increased the policy rate by 50 basis points for each 100 basis point increase in inflation following the decline in the terms of trade.
  - In contrast, a country with Australia’s current level of transparency kept the policy rate unchanged following the same shock.
- Macroeconomic consequences:
  - Tighter policy responses in less-transparent settings increase macroeconomic volatility, since inflation and output move in opposite directions after a terms-of-trade decline.
- Note on data: Transparency scores cited refer to 2014 (Dincer and Eichengreen 2014).

### Communication: Emphasize Quality over Quantity to Anchor Expectations
- Effective communication should strengthen the signal-to-noise ratio; it is the quality, not the quantity, of information that matters.
- Blinder (2009) framing: successful central bank communication should make policy more predictable and make market expectations about future short-term rates more accurate.
- Measurement approach used:
  - Short-term central bank predictability measured using surveys of financial market analysts gathered the day before each monetary policy decision; the difference between expectation and outcome is treated as a forecast error or monetary policy surprise.
- The chapter analyzes structure and content of press releases and minutes over the last eight years to assess effects on predictability and credibility.

*Source: Chapter 2 of the April 2016 Regional Economic Outlook: Western Hemisphere (excerpt provided).*

### 3. CREdIbILITy, COMMUNICATION,  ANd MONETARy pOLICy pROCyCLICALITy IN LATIN AMERICA

### 3. CREdIbILITy, COMMUNICATION, ANd MONETARy pOLICy pROCyCLICALITy IN LATIN AMERICA

### Monetary policy predictability and surprises
- Short-term predictability of interest rate decisions is low in Latin America, with the notable exception of Chile.
- Forecast errors for Colombia and Brazil are the largest in a sample of 18 inflation-targeting economies.
- Monetary policy surprises are most frequent in Latin America; since 2010 the Central Bank of Colombia “surprised” markets once every five meetings.
- The frequency of monetary policy surprises has increased in recent years for all LA5 central banks except that of Brazil, despite previous inflationary shocks having mostly dissipated.
- The analysis suggests deficiencies in the communication framework could contribute to low policy predictability for some central banks.

### Transparency, text length, and content of communications
- The length (word count) of central bank press releases is a quantifiable indicator of communication effort and procedural transparency; Mexico and Brazil exhibit big increases in press-release word counts in recent years as part of pushes to improve policy transparency.
- Longer statements do not necessarily lead to clearer communication: verbosity can coexist with low information content.
- Publishing of explanations varied historically: in Brazil, prior to 2016, explanations for Monetary Policy Committee decisions were presented in minutes published with a two-week lag rather than in contemporaneous statements.

### Readability, tone, and links to predictability
- Readability is measured with the Flesch reading ease (RE) index for Brazil (English) and the Flesch-Szigriszt index for Spanish press releases (Chile, Colombia, Mexico, Peru); the index formulae used are:
  - Flesch (English): RE = 206.835 – (1.015 x ASL) – (84.6 x ASW)
  - Flesch-Szigriszt (Spanish): RE = 206.835 – (ASL) – (84.6 x ASW)
- Improvements in clarity occurred in Chile and, to a lesser extent, in Colombia.
- Press statements in Chile, Colombia, and Peru use similar language complexity as the business section of local newspapers; press statements from Brazil and Mexico use more complicated language.
- A longer central bank statement is associated with lower readability (negative correlation between text length and readability).
- More readable press statements are associated with lower monetary policy forecast errors; the cross-sample adjusted R-squared of the relation reported is R 2 = 0.38.

### Forward guidance and transmission to inflation expectations
- Forward-looking language in communications (explicit guidance on the likely future direction of monetary policy, e.g., an easing or tightening “bias”) was used infrequently by Latin American central banks during 2011–17, with the notable exception of the Central Bank of Chile.
- The Central Bank of Chile included a policy “bias” at a rate of once every two meetings.
- Announcements that contain an explicit policy “bias” appear associated with higher transmission from policy rates to long-term inflation compensation measures: monetary policy decisions have a larger effect on breakeven inflation when accompanied by explicit guidance.
- Empirical identification: effect of an unanticipated increase in policy rate on breakeven inflation (one-day change in the difference between yield on 10-year nominal and inflation-linked government bonds) shows a larger total effect when guidance is present versus the effect of the policy change alone.

### Tone consistency, minutes, and market reaction
- Computational linguistic measures summarize tone in policy discussions as “hawkish” or “dovish” depending on word choice and context.
- LA5 central banks tend to back up words with deeds: more “hawkish” (“dovish”) tones tend to result in a tightening (loosening) of the policy rate in subsequent meetings.
- The tone in documents from Colombia and Mexico are reliable predictors of future policy changes.
- Market rates respond to the tone of press releases and minutes; in Chile, tone in press releases explains a significant share of variation in market rates (particularly short-term rates), while tone in minutes has a negligible effect—attributed to minutes being closely aligned with statements and published with a lag.
- Markets show some sensitivity to the tone in minutes in Brazil and Mexico, and to a lesser extent in Colombia.
- Inclusion of tone indices in regressions of daily changes in market rates increases explanatory power relative to regressions using only unexpected monetary policy changes.

### Policy takeaways and recommendations
- Credibility (measured by the absolute difference between medium-term inflation expectations and the midpoint of central bank targets) materially affects policy responses to short-term inflationary shocks.
- Credibility is strongly related to transparency and communication quality.
- There is scope to increase central bank transparency in the region to enhance credibility, helping the public better anticipate decisions and aligning medium-term inflation expectations with central bank objectives—thereby strengthening policy transmission and affecting the procyclicality/countercyclicality of monetary policy.
- Practical steps for strengthening frameworks in LA5 and the region include:
  - Filling data gaps, for example by increasing the horizon of survey-based expectations.
  - Considering publishing votes and comments of individual committee members.
  - Central banks that do not publish minutes of policy meetings should consider publishing them.
  - Reducing the lag in the publication of minutes.
- Communication strategy matters: what is published and how it is communicated to the general public both affect predictability and credibility. Clear, unambiguous communication and explicit conditional forward guidance improve transmission to long-term inflation compensation measures.
- Transparency and communication are not a panacea; central banks should tailor strategies aligned to policy objectives.

### Measuring inflation expectations using an affine bond-yield framework (box summary)
- Surveys of professional forecasters are one method to obtain inflation expectations but have drawbacks: infrequent release, time lag, limited horizons, and documented bias and inertia.
- In Latin America only a few countries issue inflation-linked bonds and there is no market for inflation options; liquidity of inflation-linked securities is relatively low (notable exception: Chile).
- An alternative approach (Gimeno and Marqués 2012) uses standard nominal bonds in an affine framework with factors that include observed inflation and zero-coupon yield curve parameters, allowing decomposition of nominal rates into real risk-free rates, expected inflation, and risk premium.
- Government bond data from four countries—Brazil, Chile, Colombia, Mexico—are used to estimate the affine model and compute inflation expectations at one-year, five-year, and ten-year horizons.
- Findings from the model:
  - Brazil and Colombia show similar patterns of expectations across horizons.
  - Chile and Mexico show more volatility at the one-year horizon and little change at longer horizons.
  - Most countries show long-horizon expectations within central bank target windows, though Brazil and Colombia had periods with expectations above these limits before a large decrease since the beginning of 2016.
  - Mexico shows long-term expectations slightly above the window limit of 4 percent, mainly related to peso movements and uncertainty about trade relations with the United States.
  - For Brazil, the 2015–16 recession affected expectations, which decreased since the beginning of 2016 and then began increasing for longer horizons at the end of 2016.
  - In Colombia, policy hikes in 2016 appear to have anchored inflation expectations closer to the central bank target.
  - Chile experienced decreasing short-term expectations implicit in debt markets since mid-2014; long-term expectations broadly aligned with the medium-term target.
- The affine-model-based expectations can be compared against survey-based expectations by evaluating forecasting accuracy (ratio of mean square errors), with the model providing an approach to obtain inflation expectations free of risk premiums because the model decomposes nominal interest rates into components.

*International Monetary Fund | April 2018*

### Box 3.1. Inflation Expectations from Financial Instruments in Latin America

### Box 3.1. Inflation Expectations from Financial Instruments in Latin America

### Forecast accuracy of inflation expectations
- Survey- and model-based expected inflation provide lower mean square error (MSE) than a unit-root prediction (current inflation as predicted value).
- A simple average of survey and model expectations provides lower MSE for all countries except Brazil.
- Table 3.1.1. Expected Inflation Forecast Errors (ratio of MSE of expected inflation from surveys and our model with respect to the MSE of prediction using current inflation as the predicted value):
  - Brazil; Sample: February 2007–October 2016; Survey: 0.5833; Model: 0.8812; Survey-Model: 0.6178
  - Chile; Sample: July 2012–December 2016; Survey: 0.7813; Model: 0.6344; Survey-Model: 0.6187
  - Colombia; Sample: February 2005–November 2016; Survey: 0.7956; Model: 0.9356; Survey-Model: 0.7898
  - Mexico; Sample: May 2011–November 2016; Survey: 0.6350; Model: 0.7078; Survey-Model: 0.6349
- Note on horizons in the table: Expected inflation is 12 months for Brazil, Colombia, and Mexico; 11 months for Chile.

### Inflation expectations from financial instruments (horizons)
- Figure 3.1.1 in the source presents inflation expectations derived from financial instruments at different horizons: one year, five years, ten years, and the inflation target, across Brazil, Chile, Colombia, and Mexico (visual time series covering July 2012–July 2017).

### The Central Bank of Chile: communication, credibility, and practices
- Institutional and historical context:
  - Chile adopted a free-floating exchange rate and an inflation-targeting regime in September 1999.
  - Upon adopting inflation targeting in 2000, the CBC adopted a band of 2 to 4 percent for CPI inflation over a 12- to 24-month horizon; since 2007 the CBC stressed keeping annual CPI inflation at around 3 percent most of the time, within a range of plus or minus 1 percentage point, with a policy horizon of around two years.
- Outcomes:
  - Inflation since 2000 has averaged 3.2 percent, almost exactly in line with the CBC’s point target.
  - Medium-term expectations have been well anchored around the central bank’s target, rarely deviating by more than a few basis points even when actual inflation spent long periods above target.
- Recent communication reforms (implemented September 2017):
  - Changes include reducing the frequency of monetary policy meetings; aligning meetings with the release of the Monetary Policy Report; including additional information in statements (vote tally, main arguments of Board members, and macro context); and introducing publication of full meeting transcripts with a 10-year lag.
  - Net effect: reduce the number of major communication events (meeting statements and report releases) from 16 to 8 a year and increase the informational content of each event.
  - Estimated CBC transparency score following reforms: 11.5 on the Dincer and Eichengreen (2014) scale.
  - The statement issued February 1, 2018, was 3.5 times longer than the average statement in 2017.
- Key outlets in the CBC’s transparency framework:
  - Monetary Policy Report: released in Spanish and English each quarter, or at every other policy meeting; states policy objective to return inflation to target within the policy horizon of two years; communicates key forecasts and fan charts and (infrequently) estimates of unobservables such as potential growth.
  - Statements: published in Spanish following each monetary policy meeting, reporting decisions and explaining recent developments and risks; almost half the statements issued since 2011 have included explicit conditional guidance about the likely future path of monetary policy.
  - Minutes: released with a two-week delay since 2006; include additional detail about meeting discussion, outlook, policy alternatives, board arguments, and vote balance; minutes have usually had only a very small market impact.
- Considerations on forward guidance:
  - Several highly transparent central banks publish the forward path for the policy interest rate under the board’s baseline scenario, often with a fan chart illustrating uncertainty.
  - Publishing a forward path could offer additional information and precision but must avoid confusion about the conditional nature of the path and the central bank’s commitment; it can expose the central bank to credibility concerns if market expectations do not align with the published forward path.

### Empirical strategy: Panel Vector Autoregression (PVAR) for terms-of-trade shocks (Annex 3.1)
- Objective: estimate effect of terms-of-trade shocks on inflation expectation gaps and monetary policy procyclicality using a panel vector autoregression (PVAR) framework.
- Identification and assumptions:
  - Terms of trade are treated as exogenously given (small open economy assumption).
  - Coefficients of domestic variables vary deterministically with structural country characteristics: degree of anchoring when the shock hits and levels of central bank transparency. Both characteristics are lagged by one year.
  - Block-zero restriction imposed a priori; external shocks identified via Cholesky factorization of estimated covariance matrix of reduced-form VAR residuals.
- Estimation details:
  - IPVAR estimated using ordinary least squares and allows for country fixed effects.
  - Two lags chosen following the Schwartz criterion.
  - Dynamic response illustrated using cumulative, conditional impulse-response functions at a four-quarter horizon.
  - Strength of transmission / procyclicality captured by: cumulative impulse response of the policy rate divided by the cumulative response of CPI inflation, both at the four-quarter horizon.
- Variable definitions:
  - External vector y*_t = (ToT_i,t), where ToT_i,t denotes the log first difference of terms of trade (relative price of exports in terms of imports).
  - Domestic vector y_t = (MPR_i,t, DD_i,t, NEER_i,t, CPI_i,t, Gap_i,t)
    - MPR: first difference of the monetary policy rate.
    - DD: log first differences of real final domestic demand.
    - NEER: log first differences of the nominal effective exchange rate.
    - CPI: log first differences of headline CPI.
    - Gap: first difference of the absolute difference between the two-year-ahead inflation expectations gap and the central bank’s target.
- Data sources and sample:
  - National accounts and financial data from Haver Statistics.
  - NEER measure from the IMF’s Information Notice System.
  - Inflation expectation forecasts from Consensus Economics long-term forecasts.
  - Central bank transparency measured using the Dincer and Eichengreen (2014) augmented transparency index.
  - Panel contains 20 economies under inflation-targeting: Australia, Brazil, Canada, Chile, Colombia, Czech Republic, Hungary, India, Indonesia, Korea, Mexico, New Zealand, Norway, Peru, Philippines, Poland, Romania, Russia, Thailand, and Turkey.
  - Panel covers period 2000–17 at a quarterly frequency.
  - Sweden was excluded because its policy rate has been negative since 2014.
- Robustness note: results are robust to the use of forecasts at a five-year horizon from Consensus Economics.

### Construction of the “tone” index (Annex 3.2)
- Methodology follows Hansen and McMahon (2016) and uses Latent Dirichlet Allocation (LDA) to identify relevant topics and sentences in central bank press releases and minutes.
- Preprocessing:
  - Remove “stop words” and reduce words to a common linguistic root (stemming).
  - LDA forms eight topics (probability distributions over words) and document distributions capturing the fraction of words devoted to each topic.
  - Only sentences identified as related to economic situation topics are used to construct the tone series.
- Hawkish and dovish word lists (English stems used in analysis; Spanish press releases and minutes analyzed except for Brazil, where English versions were analyzed):
  - Hawkish words include: increase*, accelerat*, fast*, strong*, high*, gain*, expand*
  - Dovish words include: decreas*, decelerat*, slow*, weak*, low*, loss*, contract*
- Tone index definition:
  - Tone = (#hawkish − #dovish) / TotalWords
  - Words in each sentence are scored and aggregated for the document; the normalized series of this score is the index used in analysis.

*Box 3.1 prepared by Yan Carrière-Swallow and Andrea Pescatori.*

---


_Source: https://www.imf.org/-/media/files/publications/reo/whd/2018/may/wreo0518-chp3.pdf_