## _wp14175

## Source details

**Canonical URL:** [_wp14175](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14175.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14175.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14175.pdf.json)

---

### Abstract and central result summary
- Study question: whether greater clarity of central bank inflation reports leads to lower return volatility in financial markets, and whether the global financial crisis affected this relationship.
- Clarity measure: Flesch-Kincaid grade level (FK).
- Sample: inflation reports by the Czech National Bank, the European Central Bank, the Bank of England, and Sveriges Riksbank.
- Main findings:
  - Some evidence, mainly for the euro area, of a negative relationship between clarity and market volatility prior to and during the early stage of the global financial crisis.
  - As the crisis unfolded, there is no longer robust evidence of a negative connection between clarity and volatility.
  - Estimated empirical effects of clarity are small when found, and mostly beneficial (more clarity associated with lower return volatility).
  - Only in a few instances did greater clarity go hand in hand with higher volatility during the global financial crisis.
  - Conclusion: reducing noise using clear reports is possible but challenging, especially in times of crisis.

### Motivation and literature context
- Central banks increased communication to enhance transparency and guide expectations.
- Quality of communication (clarity) may raise signal-to-noise ratios and reduce uncertainty; counterarguments note potential trade-offs and occasions when transparency can increase volatility.

### Empirical strategy and intuition
- Intuition: clearer formulation reduces uncertainty over policy and enables more informed price formation, potentially lowering return volatility.
- Focus: return volatility as a short-term measure; executive summaries of inflation reports are used rather than full texts.
- Readability approach relies on objective text characteristics (word lengths, sentence lengths).

### Readability measure: Flesch-Kincaid (FK)
- FK formula: FK = 0.39 * (words / sentences) + 11.8 * (syllables / words) − 15.59
- Interpretation: higher words/sentences or syllables/words raises FK, indicating lower clarity (more education required).
- Stylized FK examples from the source:
  - “We think inflation will be below two percent next year” → FK = 4.8.
  - Replace “think” with “expect” → FK = 6.0.
  - Add “over the next twelve months” → FK = 6.7.

### Data and markets analyzed
- Central banks/texts: Czech National Bank, European Central Bank (editorial of the Monthly Bulletin), Bank of England, Sveriges Riksbank; mainly executive summaries.
- Market instruments:
  - Treasury bills and government bond yields across maturities ranging from overnight up to 5 year rates.
  - Stock market returns for main indices: PX (Czech Republic), Eurostoxx50 (euro area), OMX30 (Sweden), FTSE100 (United Kingdom).
- Event-window design: ten-day event windows around publication; natural logarithm of standard deviations used for dependent variables.

### Volatility computation and econometric specifications
- Post-event volatility measure:
  - ln(σ_post_t) = ln( sqrt( ∑_{i=1}^{10} (r_{t+i} − μ_r)^2 / 9 ) )
  - r_t denotes yield changes or returns on the day of communication; μ_r denotes averages for r_t over the ten-day post-event window.
- Main regression (estimated for each central bank):
  - ln(σ_post_t) = α + β_crisis * CRISIS_t + β_FK * FK_t + β_FKcrisis * (FK_t * CRISIS_t) + β_pre * ln(σ_pre_t) + β_pol * (Σ_{k=0}^{30} Δi^p_{t−k}) + Y′_t γ + ε_t
  - CRISIS dummy: equals 1 after 14 September 2008, and zero for earlier periods.
  - Interaction term FK_t * CRISIS_t captures changes in the clarity–volatility relationship during the crisis.
  - Controls: ln(σ_pre_t); Σ_{k=0}^{30} Δi^p_{t−k} (average change in policy rate in the 30-day period prior to report release); year dummy vector Y_t.
  - Standard errors: White (1980).
- Ten-day horizon chosen to study longer-run effects beyond immediate hourly/daily windows.

### Implementation choices and limitations
- Objective FK measure avoids subjective content analysis but focuses on textual form.
- Executive summaries prioritized due to higher likelihood of being read.
- Only two control variables included because textual clarity was poorly explained by fundamentals and sample size was limited.
- Crisis period includes unconventional policies and forward guidance; analysis of clarity of such policies left for future work.
- Authors note high-frequency methods (e.g., GARCH, realized volatility) could be used in future research.

### Conceptual framework details
- Lower FK grade levels indicate higher clarity.
- β_FK measures the percentage change in volatility related to unit changes in the FK grade level.
- A positive β_FK implies that lower FK grade levels (greater clarity) coincide with lower volatility.
- Crisis interaction β_FKcrisis: β_FK + β_FKcrisis reflects the net effect during crisis years.
- Neither expected inflation, expected deviations from inflation targets, nor voting records robustly explain variations in clarity (source footnote).

### Summary statistics (cut-off date 31/8/2013)
- Czech Republic
  - FK level (reports): mean 14.10, sd 0.91, p10 13.1, p50 14.0, p90 14.95, count 53
  - O/N rate volatility: mean -2.9, sd 1.6, p10 -5.0, p50 -2.9, p90 -0.8, count 399
  - 3 months rate volatility: mean -3.1, sd 0.6, p10 -3.8, p50 -3.2, p90 -2.3, count 036
  - 1 y rate volatility: mean -3.5, sd 0.6, p10 -4.2, p50 -3.5, p90 -2.8, count 3036
  - 2 y rate volatility: mean -3.6, sd 0.6, p10 -4.3, p50 -3.6, p90 -2.8, count 3036
  - 5 y rate volatility: mean -3.4, sd 0.6, p10 -4.2, p50 -3.4, p90 -2.6, count 3014
  - Stocks volatility: mean 0.1, sd 0.5, p10 -0.5, p50 0.1, p90 0.7, count 3822
- Euro area
  - FK level (reports): mean 16.00, sd 0.81, p10 14.9, p50 16.0, p90 16.9, count 175
  - O/N rate volatility: mean -3.2, sd 1.3, p10 -5.2, p50 -3.0, p90 -1.6, count 3779
  - 3 months rate volatility: mean -4.1, sd 0.9, p10 -5.2, p50 -4.1, p90 -3.0, count 2341
  - 1 y rate volatility: mean -3.6, sd 0.7, p10 -4.4, p50 -3.7, p90 -2.7, count 2341
  - 2 y rate volatility: mean -3.3, sd 0.5, p10 -4.0, p50 -3.4, p90 -2.6, count 2341
  - 5 y rate volatility: mean -3.3, sd 0.4, p10 -3.8, p50 -3.3, p90 -2.7, count 2341
  - Stocks volatility: mean 0.2, sd 0.5, p10 -0.5, p50 0.2, p90 0.8, count 3822
- Sweden
  - FK level (reports): mean 12.40, sd 0.91, p10 11.2, p50 12.4, p90 13.6, count 58
  - O/N rate volatility: mean -4.0, sd 1.4, p10 -5.9, p50 -4.2, p90 -2.4, count 2157
  - 3 months rate volatility: mean -4.0, sd 0.7, p10 -5.0, p50 -3.9, p90 -3.1, count 3809
  - 1 y rate volatility: mean -3.7, sd 0.7, p10 -4.6, p50 -3.7, p90 -2.9, count 2923
  - 2 y rate volatility: mean -3.3, sd 0.5, p10 -3.9, p50 -3.3, p90 -2.7, count 3821
  - 5 y rate volatility: mean -3.2, sd 0.4, p10 -3.7, p50 -3.2, p90 -2.7, count 3821
  - Stocks volatility: mean 0.2, sd 0.5, p10 -0.4, p50 0.2, p90 0.9, count 4345
- United Kingdom
  - FK level (reports): mean 12.60, sd 0.81, p10 11.5, p50 12.6, p90 13.5, count 67
  - O/N rate volatility: mean -3.6, sd 2.7, p10 -7.4, p50 -2.4, p90 -0.6, count 3164
  - 3 months rate volatility: mean -4.3, sd 0.9, p10 -5.5, p50 -4.3, p90 -3.2, count 4130
  - 1 y rate volatility: mean -3.6, sd 0.6, p10 -4.4, p50 -3.6, p90 -2.9, count 4337
  - 2 y rate volatility: mean -3.3, sd 0.5, p10 -3.9, p50 -3.3, p90 -2.7, count 4337
  - 5 y rate volatility: mean -3.2, sd 0.4, p10 -3.7, p50 -3.1, p90 -2.7, count 4337
  - Stocks volatility: mean -0.1, sd 0.5, p10 -0.7, p50 -0.1, p90 0.6, count 4345

### Main estimation results — pre-crisis
- Evidence that greater clarity coincided with lower volatility prior to the crisis is limited and statistically significant mainly for the ECB.
- ECB: point estimate for β_FK roughly 0.20 for some maturities.
  - Interpretation given in source: A point estimate of 0.20 indicates that volatility declines by 20% if the FK grade level of an inflation report decreases by one unit.

### Main estimation results — crisis and early-crisis dynamics
- During the financial crisis as a whole, evidence of a negative effect largely disappears; in some cases the relationship turns positive.
- Bank of England:
  - For BoE communications and volatility of FTSE100 returns, estimated parameter for clarity in crisis years equals 0.44 (p=0.02).
- ECB:
  - β_FKcrisis is smaller than zero but not statistically significant; unable to reject null that β_FK + β_FKcrisis = 0.
- Rolling-window regressions for the euro area (window length: four years; shift: six months) show:
  1. When samples do not exclusively include crisis observations, point estimates for β_FK fluctuate around 0.20.
  2. Once only observations after September 2008 are included, estimates for β_FK decline steadily towards zero.
  3. In the middle period (samples include some crisis observations), there is an increase in point estimates for β_FK for two-year and three-year rates; these were significantly different from zero at the 5% level.
- Interpretation: Government bond yield volatility became more responsive to clarity during early stages of the crisis — more (less) clarity coincided with lower (higher) return volatility.

### Evidence of positive clarity–volatility association in some crisis contexts
- Some indications that during crisis years clearer communications have coincided with higher volatility (positive relationship) in certain contexts and maturities:
  - Sweden: sum β_FK + β_FKcrisis is negative for the one-year and the five-year maturity (heterogeneous across maturities and countries).
  - Euro area: rolling-window point estimates become negative (but not significantly different from zero) as more crisis observations are included.
- The results illustrate that increased transparency can, at times, create news rather than reduce noise.

### Selected regression coefficients and significance (representative results from Tables)
- Table 3, Panel B — euro area (selected):
  - 2 year: FK level (reports) 0.20 (0.10) ∗
  - 3 year: FK level (reports) 0.17 (0.09) ∗∗
  - 5 year: FK level (reports) 0.15 (0.09) ∗
  - Stock returns: FK level (reports) -0.00 (0.06); FK*Crisis -0.13 (0.14)
  - Significance notation: ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01
- Table 4, Panel B — United Kingdom (selected):
  - Stock returns: FK level (reports) -0.14 (0.12); FK*Crisis 0.58 (0.21) ∗∗∗ (Observations 66; Adjusted R2 0.48; F-statistic 5.98; p-value 0.02)
- Table 4, Panel A — Sweden (selected):
  - 1 year: FK level (reports) -0.07 (0.14); FK*Crisis -0.26 (0.15) (Observations 39; Adjusted R2 0.38; F-statistic 10.34; p-value 0.00)
  - 5 year: FK level (reports) -0.00 (0.08); FK*Crisis -0.21 (0.11) ∗ (Observations 50; Adjusted R2 0.43; F-statistic 7.95; p-value 0.01)

### Conclusions and policy implications
- Long-run effects of clear central bank communication on volatility are small, but evidence points to beneficial effects in some periods.
- Findings:
  - Prior to and during the early stages of the financial crisis, clarity of reports and asset return volatility were negatively related.
  - During the financial crisis as a whole, the negative relationship largely disappeared and may have, to a limited extent, become positive.
- Two main implications:
  1. Greater clarity of inflation reports will not always coincide with reduced return volatility, but clear communication can sometimes reduce market noise.
  2. Reducing noise by publishing clear inflation reports is challenging during crises; other communication channels (press conferences, speeches, interviews) may have different effects and merit further research.

*Italic source: WP/14/175 — Does the Clarity of Inflation Reports Affect Volatility in Financial Markets? Aleš Bulíř, Martin Čihák, and David-Jan Jansen (September 2014), Sections 1–3 (cut-off date 31/8/2013).*

### Section 1

### Does the Clarity of Inflation Reports Affect Volatility in Financial Markets? — Section 1

### Abstract and central result summary
- Study question: whether greater clarity of central bank inflation reports leads to lower return volatility in financial markets, and whether the global financial crisis affected this relationship.
- Clarity measure: Flesch-Kincaid grade level (FK), a standard readability measure.
- Sample: inflation reports by the Czech National Bank, the European Central Bank, the Bank of England, and Sveriges Riksbank.
- Main findings:
  - Some evidence, mainly for the euro area, of a negative relationship between clarity and market volatility prior to and during the early stage of the global financial crisis.
  - As the crisis unfolded, there is no longer robust evidence of a negative connection between clarity and volatility.
  - Estimated empirical effects of clarity are small when found, and mostly beneficial (more clarity associated with lower return volatility).
  - Only in a few instances did greater clarity go hand in hand with higher volatility during the global financial crisis.
  - Conclusion: reducing noise using clear reports is possible but challenging, especially in times of crisis.

### Motivation and literature context
- Central banks have increased use of communication to increase transparency and guide expectations (inflation, growth, future policy).
- Prior evidence shows central bank communications matter for financial markets (references cited in source).
- Quality of communication (clarity) may matter in addition to quantity; clearer communications may raise signal-to-noise ratios and reduce uncertainty.
- Counterarguments and trade-offs exist: greater transparency is not always welfare-enhancing; there may be an optimal intermediate degree of transparency; increased transparency can in some cases increase volatility as markets react to new information.

### Empirical strategy and intuition
- Intuition: clearer formulation by central banks helps agents understand communications, reducing uncertainty over policy and enabling more informed price formation, potentially lowering return volatility.
- Focus on return volatility as a short-term measure of how uncertainty relates to price formation.
- The analysis focuses on executive summaries of inflation reports rather than full texts because they are most likely to be read by market participants.
- Readability approach uses objective text characteristics (word lengths, sentence lengths) to measure clarity; avoids subjective content analysis.

### Readability measure: Flesch-Kincaid (FK)
- FK grade level for English texts (expresses reading difficulty as number of years of education needed):
  - FK = 0.39 * (words / sentences) + 11.8 * (syllables / words) − 15.59
- Interpretation: higher words/sentences or syllables/words raises FK, indicating lower clarity (more education required to comprehend).
- Stylized examples in source illustrate FK sensitivity: 
  - “We think inflation will be below two percent next year” → FK = 4.8.
  - Replace “think” with “expect” → FK = 6.0.
  - Add “over the next twelve months” → FK = 6.7.

### Data and markets analyzed
- Central banks and texts: Czech National Bank, European Central Bank (editorial of the Monthly Bulletin), Bank of England, Sveriges Riksbank; mainly executive summaries.
- Market instruments:
  - Treasury bills and government bond yields across maturities ranging from overnight up to 5 year rates.
  - Stock market returns for main indices: PX (Czech Republic), Eurostoxx50 (euro area), OMX30 (Sweden), FTSE100 (United Kingdom).
- Event-window design: ten-day event windows around publication of inflation reports; standard deviations computed using ten days for event windows; natural logarithm of standard deviations taken for dependent variables.

### Volatility computation and econometric specifications
- Post-event volatility measure:
  - ln(σ_post_t) = ln( sqrt( ∑_{i=1}^{10} (r_{t+i} − μ_r)^2 / 9 ) )
  - ln(σ_post_t) denotes volatility for the post-event window; r_t denotes yield changes or returns on the day of communication; μ_r denotes averages for r_t over the ten-day post-event window.
- Main regression estimated for each central bank:
  - ln(σ_post_t) = α + β_crisis * CRISIS_t + β_FK * FK_t + β_FKcrisis * (FK_t * CRISIS_t) + β_pre * ln(σ_pre_t) + β_pol * (Σ_{k=0}^{30} Δi^p_{t−k}) + Y′_t γ + ε_t
  - CRISIS dummy: equals 1 after 14 September 2008, and zero for earlier periods.
  - Interaction term (FK_t * CRISIS_t) captures changes in the clarity–volatility relationship during the crisis.
  - Controls included:
    - Pre-event volatility ln(σ_pre_t) to account for baseline volatility (difficulties drafting clear reports when volatility is high).
    - Recent changes in policy stance: Σ_{k=0}^{30} Δi^p_{t−k} (average change in the policy rate in the 30-day period prior to report release).
    - Year dummy vector Y_t (additional time dummies not preferred due to limited observations).
  - Standard errors computed using White (1980) approach.
- Ten-day horizon chosen to study longer-run effects of communication beyond hourly/daily windows; ambition to capture beneficial effects that persist beyond immediate market reaction.

### Implementation choices and limitations noted
- Use of objective FK readability measure avoids subjective content analysis but focuses on textual form (word/sentence/syllable structure).
- Focus on executive summaries because of higher likelihood of being read.
- Only two control variables included due to limited success in explaining variation in textual clarity with fundamentals and limited sample size.
- A crisis period includes unconventional monetary policies and forward guidance, but analysis of clarity of these policies is left for future work.
- Authors note that high-frequency methods (e.g., GARCH, realized volatility) could be used in future research; current study uses ten-day windows to capture longer-horizon effects.

*Italic source: WP/14/175 — Does the Clarity of Inflation Reports Affect Volatility in Financial Markets? Aleš Bulíř, Martin Čihák, and David-Jan Jansen (September 2014), Section 1.*

### Section 2

### _wp14175 - Section 2

### Conceptual framework: clarity and volatility
- Clarity measured by the Flesch-Kincaid (FK) grade level; lower FK grade levels indicate higher levels of clarity.
- β_FK measures the percentage change in volatility related to unit changes in the FK grade level.
- A positive β_FK implies that lower FK grade levels (greater clarity) coincide with lower volatility.
- The crisis interaction β_FKcrisis captures whether the relationship between clarity and volatility differs in crisis years; the sum β_FK + β_FKcrisis reflects the net effect during crisis years.
- Neither expected inflation, expected deviations from inflation targets, nor voting records can robustly explain variations in clarity (footnote in source).

### Data description
- Summary statistics reported (Table 2) for clarity of inflation reports and financial market volatility for: Czech Republic, euro area, Sweden, United Kingdom.
- Reported statistics include means, standard deviations, 10th, 50th, 90th percentiles, and number of observations.
- General patterns:
  - In all four cases, stock market volatility is higher than volatility of interest rates.
  - Volatility levels in stock returns: higher in Sweden and the euro area than in the Czech Republic and the United Kingdom.
  - Volatility generally higher at the short end of the yield curve, except in the Czech Republic.
  - FK grade level statistics indicate cross-country variation, partly reflecting original report languages.
- Time-series perspective emphasized: changes over time are more relevant than cross-country differences.
- Figure 1 documents a decrease in clarity around the start of the global financial crisis in 2008 (annual averages 1997–2013).

### Estimation approach and auxiliary methods
- Main regression uses equation 3 with coefficients β_FK and β_FKcrisis; dependent variables are volatility levels for interest rates of various maturities and stock returns.
- Tables 3 and 4 report parameter estimates for the Czech Republic and the euro area (Table 3) and Sweden and the United Kingdom (Table 4).
- F-statistics and p-values (based on Wald tests) reported for hypothesis β_FK + β_FKcrisis = 0 to test whether clarity and volatility are related during crisis years.
- Rolling-window regressions used for the euro area:
  - Window length: four years.
  - Windows shifted forward by six months in each subsequent regression.
  - First regression covers 2004 to 2007; subsequent regressions shift by six months.
  - Rolling regressions do not include an interaction term between clarity and the crisis dummy.
  - Figure 2 shows coefficient for FK grade level for two-year and three-year interest rates across rolling samples.
  - Vertical markers: first vertical line = last sample that only uses pre-crisis observations; second vertical line = first sample that only includes observations after September 2008.

### Main estimation results: pre-crisis period
- Some evidence that greater clarity coincides with lower volatility in financial markets prior to the crisis.
- Statistically significant negative relationship only for the ECB (Table 3, panel B).
- For the ECB’s Monthly Bulletin, clarity shows the clearest connection with medium-term interest rates.
- Point estimate size roughly 0.20 for β_FK in the ECB case:
  - Interpretation: A point estimate of 0.20 indicates that volatility declines by 20% if the FK grade level of an inflation report decreases by one unit.
  - Decreasing FK grade level by one unit implies the average person needs one year of schooling less to sufficiently comprehend the inflation report.
  - Such clarity improvements can be realized by straightforward textual edits.
- The size 0.20 is comparable to estimates from Humphrey-Hawkins testimonies analyzed in Jansen (2011).

### Main estimation results: crisis and early-crisis dynamics
- During the financial crisis as a whole, evidence of a negative effect of clarity on volatility largely disappears; in some cases the relationship may turn positive.
- Bank of England evidence:
  - For BoE communications and volatility of FTSE100 returns (Table 4, panel B), estimated parameter for clarity in crisis years equals 0.44 (p=0.02).
- ECB during crisis:
  - β_FKcrisis is smaller than zero but not statistically significant for the ECB (Table 3, panel B).
  - Bottom row of Table 3 indicates inability to reject the null that β_FK + β_FKcrisis = 0.
- Rolling-window results for euro area (Figure 2) suggest three empirical points:
  1. When samples do not exclusively include crisis observations, point estimates for β_FK fluctuate around 0.20, consistent with Table 3.
  2. Once only observations after September 2008 are included, there is a steady decline towards zero of the estimates for β_FK.
  3. In the middle period (samples start to include some crisis observations), there is an increase in the point estimates for β_FK for both two-year and three-year rates; these point estimates are significantly different from zero at the 5% level.
- Interpretation: Volatility of government bond yields became more responsive to clarity of Monthly Bulletins during the early stages of the crisis — more (less) clarity coincided with lower (higher) return volatility.

### Evidence of positive clarity–volatility association in some crisis contexts
- Some indications that during crisis years clearer communications have gone hand in hand with higher levels of volatility (i.e., positive relationship):
  - Sweden: sum β_FK + β_FKcrisis is negative for the one-year and the five-year maturity (Table 4, panel A) — indicates heterogeneous patterns across maturities and countries.
  - Euro area: rolling-window point estimates show a downward trend and become negative (but not significantly different from zero) as more crisis-period observations are included.
- The findings illustrate that increased transparency can, at times, create news rather than reduce noise (citing Blinder et al. 2008, Bloom 2014).
- The paper notes it would be interesting for future work to investigate under what conditions the relationship between clarity and volatility becomes positive.

### Conclusions and policy implications
- Long-run effects of clear central bank communication on volatility are small, but indications point to beneficial effects in some periods.
- Evidence:
  - Prior to and during the early stages of the financial crisis, clarity of reports and asset return volatility were negatively related.
  - During the financial crisis as a whole, the negative relationship largely disappeared and may have, to a limited extent, become positive.
- Two main implications:
  1. No guarantee that greater clarity of inflation reports will always coincide with reduced return volatility, but clear communication is sometimes able to reduce noise in financial markets.
  2. Reducing noise by publishing clear inflation reports is not without challenges during crises; other communication channels (press conferences, speeches, interviews) may have different effects on financial markets and merit further research.

*Source: _wp14175 - Section 2*

### Section 3

### _wp14175 - Section 3

### Summary statistics (Table 2)
- Cut-off date: 31/8/2013.
- Czech Republic
  - FK level (reports): mean 14.10, sd 0.91, p10 13.1, p50 14.0, p90 14.95, count 53
  - O/N rate volatility: mean -2.9, sd 1.6, p10 -5.0, p50 -2.9, p90 -0.8, count 399
  - 3 months rate volatility: mean -3.1, sd 0.6, p10 -3.8, p50 -3.2, p90 -2.3, count 036
  - 1 y rate volatility: mean -3.5, sd 0.6, p10 -4.2, p50 -3.5, p90 -2.8, count 3036
  - 2 y rate volatility: mean -3.6, sd 0.6, p10 -4.3, p50 -3.6, p90 -2.8, count 3036
  - 5 y rate volatility: mean -3.4, sd 0.6, p10 -4.2, p50 -3.4, p90 -2.6, count 3014
  - Stocks volatility: mean 0.1, sd 0.5, p10 -0.5, p50 0.1, p90 0.7, count 3822
- Euro area
  - FK level (reports): mean 16.00, sd 0.81, p10 14.9, p50 16.0, p90 16.9, count 175
  - O/N rate volatility: mean -3.2, sd 1.3, p10 -5.2, p50 -3.0, p90 -1.6, count 3779
  - 3 months rate volatility: mean -4.1, sd 0.9, p10 -5.2, p50 -4.1, p90 -3.0, count 2341
  - 1 y rate volatility: mean -3.6, sd 0.7, p10 -4.4, p50 -3.7, p90 -2.7, count 2341
  - 2 y rate volatility: mean -3.3, sd 0.5, p10 -4.0, p50 -3.4, p90 -2.6, count 2341
  - 5 y rate volatility: mean -3.3, sd 0.4, p10 -3.8, p50 -3.3, p90 -2.7, count 2341
  - Stocks volatility: mean 0.2, sd 0.5, p10 -0.5, p50 0.2, p90 0.8, count 3822
- Sweden
  - FK level (reports): mean 12.40, sd 0.91, p10 11.2, p50 12.4, p90 13.6, count 58
  - O/N rate volatility: mean -4.0, sd 1.4, p10 -5.9, p50 -4.2, p90 -2.4, count 2157
  - 3 months rate volatility: mean -4.0, sd 0.7, p10 -5.0, p50 -3.9, p90 -3.1, count 3809
  - 1 y rate volatility: mean -3.7, sd 0.7, p10 -4.6, p50 -3.7, p90 -2.9, count 2923
  - 2 y rate volatility: mean -3.3, sd 0.5, p10 -3.9, p50 -3.3, p90 -2.7, count 3821
  - 5 y rate volatility: mean -3.2, sd 0.4, p10 -3.7, p50 -3.2, p90 -2.7, count 3821
  - Stocks volatility: mean 0.2, sd 0.5, p10 -0.4, p50 0.2, p90 0.9, count 4345
- United Kingdom
  - FK level (reports): mean 12.60, sd 0.81, p10 11.5, p50 12.6, p90 13.5, count 67
  - O/N rate volatility: mean -3.6, sd 2.7, p10 -7.4, p50 -2.4, p90 -0.6, count 3164
  - 3 months rate volatility: mean -4.3, sd 0.9, p10 -5.5, p50 -4.3, p90 -3.2, count 4130
  - 1 y rate volatility: mean -3.6, sd 0.6, p10 -4.4, p50 -3.6, p90 -2.9, count 4337
  - 2 y rate volatility: mean -3.3, sd 0.5, p10 -3.9, p50 -3.3, p90 -2.7, count 4337
  - 5 y rate volatility: mean -3.2, sd 0.4, p10 -3.7, p50 -3.1, p90 -2.7, count 4337
  - Stocks volatility: mean -0.1, sd 0.5, p10 -0.7, p50 -0.1, p90 0.6, count 4345
- Notes: Columns denote the mean, standard deviation, 10th percentile, median, 90th percentile, and the number of observations.

### Clarity and volatility: regression results (Table 3, Panel A — Czech Republic)
- Regression specification: FK grade level and FK*Crisis interaction (crisis dummy equals 1 after 14 September 2008). Dependent variables: volatilities of interest rates (overnight to 5 year) and stock market returns.
- Coefficients (standard errors in parentheses)
  - Overnight (column 1): FK level (reports) -0.01 (0.26); FK*Crisis 0.17 (0.50)
    - Observations 49; Adjusted R2 0.43; F-statistic 0.15; p-value 0.70
  - 3 month (column 2): FK level (reports) 0.10 (0.11); FK*Crisis -0.01 (0.45)
    - Observations 41; Adjusted R2 0.22; F-statistic 0.04; p-value 0.84
  - 1 year (column 3): FK level (reports) 0.07 (0.12); FK*Crisis 0.25 (0.31)
    - Observations 41; Adjusted R2 0.32; F-statistic 1.17; p-value 0.29
  - 2 year (column 4): FK level (reports) 0.10 (0.13); FK*Crisis 0.46 (0.35)
    - Observations 41; Adjusted R2 0.30; F-statistic 2.62; p-value 0.12
  - 3 year (column 5): FK level (reports) 0.09 (0.10); FK*Crisis 0.44 (0.43)
    - Observations 41; Adjusted R2 0.21; F-statistic 1.55; p-value 0.23
  - 5 year (column 6): FK level (reports) 0.09 (0.09); FK*Crisis 0.52 (0.38)
    - Observations 41; Adjusted R2 0.26; F-statistic 2.81; p-value 0.11
  - Stock returns (column 7): FK level (reports) -0.06 (0.10); FK*Crisis -0.08 (0.14)
    - Observations 53; Adjusted R2 0.65; F-statistic 1.57; p-value 0.22

### Clarity and volatility: regression results (Table 3, Panel B — euro area)
- Coefficients (standard errors in parentheses)
  - Overnight (col 1): FK level (reports) -0.09 (0.15); FK*Crisis 0.05 (0.31)
    - Observations 166; Adjusted R2 0.63; F-statistic 0.02; p-value 0.88
  - 3 month (col 2): FK level (reports) 0.04 (0.17); FK*Crisis -0.09 (0.23)
    - Observations 106; Adjusted R2 0.66; F-statistic 0.12; p-value 0.73
  - 1 year (col 3): FK level (reports) 0.22 (0.11); FK*Crisis -0.17 (0.16)
    - Observations 106; Adjusted R2 0.60; F-statistic 0.21; p-value 0.64
  - 2 year (col 4): FK level (reports) 0.20 (0.10) ∗; FK*Crisis -0.15 (0.16)
    - Observations 106; Adjusted R2 0.48; F-statistic 0.17; p-value 0.68
  - 3 year (col 5): FK level (reports) 0.17 (0.09) ∗∗; FK*Crisis -0.18 (0.14)
    - Observations 106; Adjusted R2 0.40; F-statistic 0.01; p-value 0.91
  - 5 year (col 6): FK level (reports) 0.15 (0.09) ∗; FK*Crisis -0.14 (0.14)
    - Observations 106; Adjusted R2 0.37; F-statistic 0.01; p-value 0.93
  - Stock returns (col 7): FK level (reports) -0.00 (0.06); FK*Crisis -0.13 (0.14)
    - Observations 175; Adjusted R2 0.47; F-statistic 1.07; p-value 0.30
- Significance notation: ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01

### Clarity and volatility: regression results (Table 4, Panel A — Sweden)
- Coefficients (standard errors in parentheses)
  - Overnight (col 1): FK level (reports) 0.23 (1.58); FK*Crisis -0.10 (1.97)
    - Observations 22; Adjusted R2 -0.31; F-statistic 0.02; p-value 0.88
  - 3 month (col 2): FK level (reports) -0.11 (0.11); FK*Crisis 0.19 (0.39)
    - Observations 49; Adjusted R2 0.31; F-statistic 0.04; p-value 0.84
  - 1 year (col 3): FK level (reports) -0.07 (0.14); FK*Crisis -0.26 (0.15)
    - Observations 39; Adjusted R2 0.38; F-statistic 10.34; p-value 0.00
  - 2 year (col 4): FK level (reports) -0.19 (0.15); FK*Crisis 0.03 (0.18)
    - Observations 50; Adjusted R2 0.14; F-statistic 2.63; p-value 0.12
  - 5 year (col 5): FK level (reports) -0.00 (0.08); FK*Crisis -0.21 (0.11) ∗
    - Observations 50; Adjusted R2 0.43; F-statistic 7.95; p-value 0.01
  - Stock returns (col 6): FK level (reports) 0.02 (0.10); FK*Crisis -0.10 (0.16)
    - Observations 50; Adjusted R2 0.65; F-statistic 0.39; p-value 0.54

### Clarity and volatility: regression results (Table 4, Panel B — United Kingdom)
- Coefficients (standard errors in parentheses)
  - Overnight (col 1): FK level (reports) 0.05 (0.50); FK*Crisis 0.04 (0.64)
    - Observations 47; Adjusted R2 0.88; F-statistic 0.04; p-value 0.84
  - 3 month (col 2): FK level (reports) -0.05 (0.14); FK*Crisis -0.59 (0.42)
    - Observations 63; Adjusted R2 0.61; F-statistic 2.65; p-value 0.11
  - 1 year (col 3): FK level (reports) -0.06 (0.09); FK*Crisis -0.13 (0.25)
    - Observations 66; Adjusted R2 0.54; F-statistic 0.72; p-value 0.40
  - 2 year (col 4): FK level (reports) -0.09 (0.10); FK*Crisis 0.10 (0.22)
    - Observations 66; Adjusted R2 0.29; F-statistic 0.00; p-value 0.97
  - 3 year (col 5): FK level (reports) -0.08 (0.11); FK*Crisis 0.12 (0.18)
    - Observations 66; Adjusted R2 0.15; F-statistic 0.06; p-value 0.81
  - 5 year (col 6): FK level (reports) -0.03 (0.10); FK*Crisis 0.09 (0.17)
    - Observations 66; Adjusted R2 0.03; F-statistic 0.18; p-value 0.67
  - Stock returns (col 7): FK level (reports) -0.14 (0.12); FK*Crisis 0.58 (0.21) ∗∗∗
    - Observations 66; Adjusted R2 0.48; F-statistic 5.98; p-value 0.02
- Significance notation: ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01

*Source: _wp14175 - Section 3 (cut-off date 31/8/2013).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14175.pdf_