## 1. Mean Level (Later Period) and Change in CBI

## Source details

**Canonical URL:** [1. Mean Level (Later Period) and Change in CBI](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08119.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08119.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08119.pdf.json)

---

### Introduction and scope
- Focus: reforms bearing on monetary policy (not financial regulation and supervision).
- Three reform areas emphasized:
  - Revisions to legal statutes to increase institutional independence from the executive.
  - Enhancements of accountability as central banks become more autonomous.
  - Increases in central bank transparency, related to changes in monetary policy conduct (notably the introduction of inflation targeting).
- New measures introduced: indices of central bank independence (CBI) and transparency covering a comprehensive set of countries and permitting comparison over time.

### Construction of indices and data coverage
- Independence index compares:
  - The original index constructed by Cukierman and others (1992) covering the 1980s.
  - The authors’ index based on a database of central bank laws held by the IMF and current through 2003.
- Transparency index methodology broadly follows Geraats (2002) and Eijffinger and Geraats (2006), with comparison between:
  - The late 1990s (based on Fry and others (2000) survey results).
  - 2006 (based on authors’ reading of central bank websites and published documents).
- Indices facilitate time comparisons and broader country coverage than some prior work.

### Main empirical findings
- Controlling for country fixed effects by taking first differences, there is robust evidence that central bank independence (CBI) reduces inflation; result robust to controls for endogeneity and measurement error via instrumental variables estimation.
- Greater central bank transparency leads the private sector to make greater use of information provided by the central bank when making forecasts, consistent with predictions of a simple signal extraction model.

### Theoretical and empirical context
- Theoretical rationale: time inconsistency problem and delegation solutions (conservative central banker or independent policymaker with mandate).
- Empirical literature: mixed findings historically; this paper exploits time dimension and finds a relatively robust negative relationship between CBI and inflation when using first differences.

---

### Measures of central bank independence (CBI)
- Index choice and sample
  - Focus: Cukierman, Webb, and Neyapti (CWN) measure.
  - Updated index computed for all but three of the seventy-two countries in the original CWN sample (unable to code these three due to unavailability of legislation in the IMF’s central bank law database (CBLD)).
  - CBLD was current through end-2003 at the time the index was computed.
  - An additional twenty-seven countries were coded.
- CWN index structure — four components:
  1. Appointment procedures for the head of the central bank.
  2. Resolution of conflict between central bank and executive.
  3. Use of an explicit policy target.
  4. Rules limiting lending to government.
- Cross-country and subgroup patterns
  - Evidence of increased independence across all countries.
  - For developing and emerging market economies: all components have shown a statistically significant increase since the 1980s.
  - For advanced economies: only components 2 and 4 have shown a significant increase; components 1 and 3 were already relatively high.
  - Reforms focused on relations with the executive in advanced economies; reforms were more comprehensive in emerging and developing countries.

- Regression correlates of change in CBI (D.CBI) — selected estimates (Table 2)
  - Specification explains change in CWN index by initial values CBI0, INF0, GDP0, DEMOC0, OPEN0, REGIME0.
  - Selected coefficient estimates (standard errors in parentheses):
    - CBI0 -0.609*** (0.206)
    - INF0 0.227** (0.109)
    - DEMOC0 0.021*** (0.007)
    - GDP0 -4.639 (3.604)
    - OPEN0 -0.256 (0.412)
    - REGIME0 -0.031** (0.013)
    - INF0*DEMOC0 -0.031 (0.030)
  - Observations: 61.
  - R2 values reported: 0.29, 0.30, 0.46 across columns.
  - Interpretation:
    - Reform related to low initial CBI (negative CBI0) and high prior inflation (positive INF0).
    - Reform more likely in initially more democratic systems (positive DEMOC0).
    - No robust evidence that reform is more likely in initially richer or more open economies.
    - Reforms appear more ambitious where initial exchange rate regimes were less flexible (negative REGIME0).

---

### Measures of central bank transparency (TRANS)
- Taxonomy and index construction
  - Adopt Geraats (2002) taxonomy: five aspects — political, economic, procedural, policy, and operational transparency.
  - For each aspect two questions chosen (total of ten); most interpreted as binary to reduce subjectivity.
  - Coding relied primarily on publicly available English-language materials; 1998 data from Fry and others (2000) Bank of England survey; 2006 coding carried out in 2006 with cross-checking in 2007.
- Components defined
  - Political transparency: delineation of roles and measurable objectives.
  - Economic transparency: release of economic information, including forecasts.
  - Procedural transparency: publication of how policy decisions are reached (minutes, voting records).
  - Policy transparency: release of information on the policy decision immediately after meetings, with detailed reasoning.
  - Operational transparency: publication on transmission mechanism, forecast accuracy, and accounting for past errors.

- Cross-country changes and measurement caveats
  - Overall increase in transparency scores not statistically significant for all countries; advanced economies experienced a statistically significant increase.
  - Two components driving the increase for advanced economies: economic and policy transparency.
  - Emerging market and developing countries: no statistically significant increase for the sample as a whole.
  - Measurement issues:
    - 1998 data are self-reported (possible upward bias).
    - English-language coding in 2006 may understate transparency for some countries (language bias).
    - Measured steep declines for China and Russia (declines of 0.5 or above) may reflect measurement error, language issues, or true decline—interpret with caution.

- Summary table snippet (mean levels and changes as presented)
  - Political: Trans - 1.83 -.06 .83 -.02 .84 -.12
  - Economic: Trans - 2.75 .12* .82 .15*** .63 .06
  - Procedural: Trans - 3.21 -.03 .17 -.07 .25 .04
  - Policy: Trans - 4.81 .27*** .90 .40*** .64 .04
  - Operational: Trans - 5.42 -.08 .43 -.10 .39 -.06
  - Total: Trans .60 .04 .63 .07** .55 -.01
  - Observations: 40 (All Countries), 36 (Advanced Economies), 23 (Emerging Markets)
  - Change significance: * significant at 10%; ** significant at 5%; *** significant at 1%.

- Correlates of transparency (TRANS1) — selected regression highlights (Table 4)
  - Variables considered: CBI1, OPEN1, REGIME1, RQ1, VA1.
  - Selected coefficient estimates (standard errors in parentheses):
    - CBI1 0.605*** (0.176)
    - RQ1 0.251*** (0.064)
    - OPEN1 -0.954** (0.393) (not consistent across specs)
    - REGIME1 0.048*** (0.010)
    - VA1 0.166*** (0.050) (when included)
    - Component correlations: CBI1.1 0.407* (0.205); CBI3.1 0.201** (0.096)
  - Observations: 40; reported R2: 0.56, 0.48, 0.63, 0.46, 0.57 across columns.
  - Interpretation:
    - Greater transparency associated with more independent central banks (CBI1), better governance (higher RQ1, VA1), and more flexible de facto exchange rate regimes (REGIME1).
    - Most robust CBI–transparency link is with CBI component 3 (existence of a well-defined target).

---

### Effects of central bank independence on inflation
- Background and econometric challenges
  - De jure measures may not capture de facto independence; turnover rate of central bank governors used as de facto proxy.
  - Endogeneity and omitted variable bias are concerns.

- OLS and first-difference results (Table 5 summary)
  - Levels regressions show weak relationship between de jure CBI and inflation; turnover rate related to inflation with predicted sign in earlier studies.
  - First-differenced specification yields predicted negative effects of both de jure and de facto independence measures on inflation.
  - Component-level result: CBI3 (single, well-specified inflation or price-level target) most likely to deliver low inflation:
    - Change regression reported: CBI3 -0.196** (0.084).

- Quantitative example (text)
  - Average increase in de jure CBI (around .25) associated with a drop in inflation of more than 5 percentage points compared to a situation where CBI does not increase at all (for countries whose inflation was already in single digits; with a greater drop for countries with higher initial inflation).

- Instrumental variables (IV) results (Table 6)
  - De jure independence instrumented with governance measures: Rule of Law (RL1) and Voice and Accountability (VA1).
  - IV estimates (2SLS and LIML) — selected estimates (standard errors in parentheses):
    - 2SLS: D.CBI -0.937*** (0.328)
    - LIML: D.CBI -0.938*** (0.329)
    - LIML with component: D.CBI3 -0.798** (0.352)
  - Instrument diagnostics:
    - Cragg-Donald (C-D) statistics and Stock–Yogo critical values discussed; some evidence of weak instruments by C-D.
    - Anderson–Rubin (A-R) test p-values indicate significance (A-R [p] 0.007, 0.007, 0.002, 0.007 across columns reported).
    - Overid test p-values: instruments not rejected (e.g., Overid test [p] 0.910, 0.910, 0.230, 0.86).
  - Robustness: using an unweighted CBI measure (LVAU) yields nearly identical results.

---

### Effects of transparency on private-sector information use and forecast behavior
- Concept and measure
  - Private-sector use of private vs. public information measured by VAR/MSE ratio:
    - VAR = variance of private-sector forecasts for inflation in the following year.
    - MSE = mean square error of those forecasts (using next year’s actual inflation).
    - VAR/MSE lies on the unit interval and declines when public information becomes relatively more accurate.
  - Signal-extraction model: r = VAR/MSE = (γβ)/(αβ + γβ) and ∂r/∂α ≤ 0 (higher public-signal accuracy α reduces r).

- Empirical results (Table 7) — selected estimates (standard errors in parentheses)
  - Sample limited by forecast dataset coverage: maximum sample size of 28 observations.
  - Baseline OLS (column (1)):
    - TRANS1 -0.322** (0.146)
    - OPEN1 -0.853** (0.398)
    - RQ1 0.117*** (0.0402)
    - Constant 0.423*** (0.0963)
    - Observations: 28; R2 = 0.27.
  - Component results (columns (2)–(6)):
    - T2.1 (Economic transparency): -0.273** (0.105)
    - T5.1 (Operational transparency): -0.205*** (0.0627)
    - T1.1, T3.1, T4.1 less consistently significant.
  - Full specification (column (7)):
    - Economic and operational components remain negative and statistically significant.
    - Observations: 28; R2 = 0.47.

- Quantitative interpretation (text)
  - A one standard deviation increase in the transparency score (around .22) is associated with a seven percentage point decline in the VAR/MSE ratio.
  - Illustrative mapping: if public and private information were initially equally accurate (α = β) and initial r = 0.26, a .07 decrease in r corresponds to an increase in public information accuracy of around 70 percent (α1 = 1.7*β).

- Interpretation
  - Greater transparency, particularly economic transparency (release of forecasts) and operational transparency, is associated with greater private-sector use of public information and reduced dispersion across forecasters.
  - Publication of forecasts and related information improves the informativeness of the public signal.

---

### Conclusions and policy implications
- Summary of findings
  - Marked increase in CBI since indices were first constructed in the late 1980s; developing and emerging market economies showed the most comprehensive increases across all CBI dimensions.
  - Increases in independence concentrated on relations between executive and monetary authority (dispute resolution, central bank financing) for both advanced and emerging/developing economies.
  - Increased independence related to: higher initial inflation, lower initial CBI, more democratic initial political systems, and initially less flexible exchange rate arrangements.
  - De jure legal independence shows limited direct effect on inflation in levels regressions, but first-difference and IV approaches uncover a statistically significant negative effect of CBI on inflation.
  - Transparency scores increased significantly only for advanced economies over the 1998–2006 period; transparency positively correlated with governance quality, central bank independence, and exchange rate flexibility.
  - Greater transparency is associated with more use of public information by private-sector forecasters (reduced VAR/MSE), especially when transparency involves public release of central bank forecasts and operational information.

- Policy and research implications
  - Institutional reforms enhancing CBI—especially those that include a well-specified policy target—appear most effective in reducing inflation; attention to de facto independence and governance preconditions is important.
  - Enhancing transparency (economic and operational) can substantially improve the informativeness of the public signal and reduce forecast dispersion; releasing forecasts is central to these transparency benefits.
  - Measurement and identification issues remain important:
    - Transparency measurement sensitive to language and self-reporting differences across periods.
    - Better instruments for both CBI and transparency are a priority for future work to identify causal channels more cleanly.
  - The transparency index methodology can be replicated to extend country coverage and build time series going forward.

---

### Time periods and data construction (key specifics)
- Current period considered to cover 2002-06, although in some cases data available only through 2005.
- Initial period for CBI/turnover: 1987-91.
- Initial period for transparency: 1997-2001 (1998 survey used).
- Central bank law database (CBLD) current through end-2003 when authors constructed the index.
- Turnover current period covers 1995-2004 (constructed using Morgan Stanley’s 2005 Central Bank Directory and central bank websites).
- Political/institutional variables from World Bank Aggregate Governance Indicators: original data annual for 2002-05 and biannual for 1996-2000; second period measure is average of 2002-05.
- Democracy score from Polity IV Dataset; variables available annually through 2004.
- Real GDP per capita and openness from Penn World Tables, version 6.2; annual data available through 2004 (current period measures based on 2002-04).
- De facto exchange rate regime data monthly (January 1990 through June 2004), updated to end-2006; regime scored 1 to 8 where 1 is most restrictive and 8 is least restrictive.
- Inflation data (annual, percent per annum) from IMF’s International Financial Statistics, generally available up to 2006; inflation transformed as π/(100+π).
- Forecast data from Consensus Economics (forecast for following calendar year); coverage mostly during the 1990s through end-2006.
- Transparency coding examples (exact codings preserved):
  - 1.1: statutory objective — 1 / .5 / 0 as defined.
  - 1.2: explicit numerical target — 1 / 0.
  - 2.1: publish surveys for inflation expectations — 1 / 0.
  - 2.2: publish forward-looking analyses — 1 / .5 / 0.
  - 3.1: publish minutes — 1 / 0.
  - 3.3: publish voting patterns — 1 / 0.
  - 4.1 and 4.2: publish explanations on day policy changed / not changed — 1 / 0.
  - 5.1: publish discussion of risks to outlook — 1 / .5 / 0.
  - 5.2: publish discussion of shocks or forecast errors after the fact — 1 / 0.
- Overall transparency index: unweighted average of 5 categories; each category is unweighted average of its subcategories.

*Source: _wp08119 - 1.     Mean Level (Later Period) and Change in CBI (IMF working paper excerpt).*

### 1.     Mean Level (Later Period) and Change in CBI ...........................................................7

### 1.     Mean Level (Later Period) and Change in CBI

### Introduction and scope
- Focus: reforms bearing on monetary policy (not financial regulation and supervision).
- Three reform areas emphasized:
  - Revisions to legal statutes to increase institutional independence from the executive.
  - Enhancements of accountability as central banks become more autonomous.
  - Increases in central bank transparency, related to changes in monetary policy conduct (notably the introduction of inflation targeting).
- New measures introduced: indices of central bank independence (CBI) and transparency covering a comprehensive set of countries and permitting comparison over time.

### Construction of indices and data coverage
- Independence index compares:
  - The original index constructed by Cukierman and others (1992) covering the 1980s.
  - The authors’ index based on a database of central bank laws held by the IMF and current through 2003.
- Transparency index methodology broadly follows Geraats (2002) and Eijffinger and Geraats (2006), with comparison between:
  - The late 1990s (based on Fry and others (2000) survey results).
  - 2006 (based on authors’ reading of central bank websites and published documents).
- Indices facilitate time comparisons and broader country coverage than some prior work (Eijffinger and Geraats, 2006; Cukierman, Miller, and Neyapti, 2002; Jácome and Vázquez, 2005).
- Related work: Crowe and Meade (2007); Arnone, Laurens, and Segalotto (2006); Dincer and Eichengreen (2007).

### Main empirical findings
- Two headline results emphasized:
  - Controlling for country fixed effects by taking first differences, there is robust evidence that central bank independence (CBI) reduces inflation. This result is robust to controls for endogeneity and measurement error via instrumental variables estimation.
  - Greater central bank transparency leads the private sector to make greater use of information provided by the central bank when making forecasts, consistent with predictions of a simple signal extraction model.

### Discussion of theoretical and empirical context
- Theoretical rationale for CBI:
  - Time inconsistency problem (Kydland and Prescott, 1977; Barro and Gordon, 1983).
  - Delegation solutions: conservative central banker (Rogoff, 1985) or independent policymaker with suitable incentives and mandate (Walsh, 1995).
  - Models of independence: goal independence vs. instrument independence (Debelle and Fischer, 1995; Fischer, 1995).
- Critics (e.g., Blinder, 1998) argue time inconsistency may be less relevant for modern central banks, especially in industrial countries; alternative motivations include political economy factors (Crowe, 2006a; Cukierman and Gerlach, 2003; Goodman, 1991; Hayo, 1998; Lohmann, 1998).
- Empirical literature has mixed findings:
  - Early studies (Alesina and Summers, 1993) found CBI associated with lower inflation in a narrow subset of industrial countries.
  - Later, broader-sample studies found more equivocal results (Eijffinger and de Haan, 1996; Berger, de Haan and Kooi, 2000; Berger, Eijffinger, and de Haan, 2001; Klomp and de Haan, 2007).
  - This paper revisits the relationship exploiting time dimension and finds relatively robust negative relationship between CBI and inflation when using first differences.

### Findings on transparency trends and effects
- Aggregate trend:
  - Over the subset of industrial and emerging market economies analyzed, overall levels of transparency have not increased significantly since the late 1990s.
  - Possible reasons: short measurement period and methodology differences between 1998 and 2006; some sharp drops in reported transparency may reflect methodological artifacts.
- Notable exceptions:
  - Creation of the European Central Bank (ECB) and introduction of inflation targeting in a number of countries since 1998 are associated with large and statistically significant increases in recorded transparency (Crowe and Meade, 2007).
- Effects of transparency:
  - Transparency can improve outcomes when information asymmetries cause inefficiencies.
  - Transparency can be costly in second-best environments when the central bank’s informational advantage helps offset other inefficiencies.
  - Empirically, greater transparency is associated with increased use of central bank-provided information by the private sector in forecasts (consistent with a signal extraction framework; Geraats, 2002).

### Methodological notes and robustness
- Identification strategy for CBI–inflation relationship:
  - Use of country fixed effects via first differences to control for time-invariant unobservables.
  - Instrumental variables estimation used to address endogeneity and measurement error; the negative effect of CBI on inflation remains robust.
- Transparency measurement caveats:
  - Comparisons across time may be biased by different data collection methodologies in the late 1990s vs. 2006.

*Source: _wp08119 - 1.     Mean Level (Later Period) and Change in CBI*

### section 5 of the paper. We find that greater transparency—in particular relating to the release of

### _wp08119 - section 5 of the paper. We find that greater transparency—in particular relating to the release of

### Measures of central bank independence (CBI)
- Index choice and sample
  - Focus: Cukierman, Webb, and Neyapti (CWN) measure.
  - Updated index computed for all but three of the seventy-two countries in the original CWN sample (unable to code these three due to unavailability of legislation in the IMF’s central bank law database (CBLD)).
  - CBLD was current through end-2003 at the time the index was computed.
  - An additional twenty-seven countries were coded (selected for economic significance and inclusion in the CBLD).

- CWN index structure
  - Four components: (1) appointment procedures for the head of the central bank; (2) resolution of conflict between central bank and executive; (3) use of an explicit policy target; (4) rules limiting lending to government.
  - First and third components relate more to internal central bank procedures and policies; second and fourth relate to relations between central bank and executive.

- Cross-country and subgroup patterns (textual findings)
  - Table 1 (summary): evidence of increased independence across all countries.
  - For developing and emerging market economies: all components have shown a statistically significant increase since the 1980s.
  - For advanced economies: only the second and fourth components (disputes with the executive and lending to government) have shown a significant increase—first and third components were already relatively high.
  - Key reforms focused on relations with the executive, rather than the central bank’s own goals and procedures, in advanced economies; reforms were more comprehensive in emerging and developing countries.

- Regression correlates of change in CBI (D.CBI) — main results (Table 2)
  - Regression specification: change in CWN index explained by initial values for independence (CBI0), inflation (INF0), real GDP per capita (GDP0), democracy (DEMOC0), openness (OPEN0), and de facto exchange rate regime (REGIME0).
  - Selected coefficient estimates (as presented):
    - CBI0 -0.609*** (0.206)
    - INF0 0.227** (0.109)
    - DEMOC0 0.021*** (0.007)
    - GDP0 -4.639 (3.604)
    - OPEN0 -0.256 (0.412)
    - REGIME0 -0.031** (0.013)
    - INF0*DEMOC0 -0.031 (0.030) (interaction not significantly different from zero)
  - Observations: 61 (for the reported specifications).
  - R2 values reported: 0.29, 0.30, 0.46 across columns.
  - Interpretation:
    - Reform related to low initial levels of CBI (negative CBI0 coefficient) and high prior inflation (positive INF0 coefficient).
    - Reform more likely in countries with initially more democratic political systems (positive DEMOC0).
    - No evidence that reform is more likely in initially richer or more open economies (GDP0 and OPEN0 not robustly significant).
    - Reforms appear more ambitious in countries with less flexible initial exchange rate regimes (negative REGIME0 coefficient).

### Measures of central bank transparency (TRANS)
- Taxonomy and index construction
  - Adopt Geraats (2002) taxonomy: five aspects of transparency — political, economic, procedural, policy, and operational transparency.
  - For each of the five components, two questions were chosen (total of ten questions); most interpreted as binary to reduce subjectivity.
  - Coding relied primarily on publicly available English-language materials (noted language bias concern); 1998 data came from Fry and others (2000) Bank of England survey; 2006 coding carried out in 2006 with cross-checking in 2007.

- Components defined
  - Political transparency: delineation of roles and measurable objectives (e.g., explicit policy targets).
  - Economic transparency: release of economic information, including forecasts.
  - Procedural transparency: publication of how policy decisions are reached (minutes, voting records, transcripts).
  - Policy transparency: release of information on the policy decision immediately after meetings, with detailed account of reasoning.
  - Operational transparency: publication on transmission mechanism, forecast accuracy, and accounting for past errors.

- Cross-country changes and measurement caveats
  - Table 3 (summary): overall increase in transparency scores not statistically significant for all countries; advanced economies experienced a statistically significant increase.
  - Two components driving the increase for advanced economies: economic and policy transparency.
  - Emerging market and developing countries: no statistically significant increase in transparency scores for the sample as a whole.
  - Noted measurement issues:
    - 1998 data are self-reported (possible upward bias).
    - English-language coding in 2006 may understate transparency for some countries (language bias).
    - Measured steep declines in transparency for China and Russia (declines of 0.5 or above) may reflect measurement error, language issues, or true decline—suggested interpretation with caution.

- Summary table snippet (mean levels and changes; as presented)
  - Trans - 1.83 -.06 .83 -.02 .84 -.12 (Political) (standard deviations in parentheses)
  - Trans - 2.75 .12* .82 .15*** .63 .06 (Economic)
  - Trans - 3.21 -.03 .17 -.07 .25 .04 (Procedural)
  - Trans - 4.81 .27*** .90 .40*** .64 .04 (Policy)
  - Trans - 5.42 -.08 .43 -.10 .39 -.06 (Operational)
  - Trans .60 .04 .63 .07** .55 -.01 (Total)
  - Observations: 40 (All Countries), 36 (Advanced Economies), 23 (Emerging Markets)
  - Change significance: * significant at 10%; ** significant at 5%; *** significant at 1%

- Correlates of transparency (TRANS1) — regression highlights (Table 4)
  - Variables considered: independence (CBI1), openness (OPEN1), exchange rate regime (REGIME1), regulatory quality (RQ1), voice and accountability (VA1).
  - Selected coefficient estimates (as presented):
    - CBI1 0.605*** (0.176)
    - RQ1 0.251*** (0.064)
    - OPEN1 -0.954** (0.393) (not consistently correlated across specifications)
    - REGIME1 0.048*** (0.010)
    - VA1 0.166*** (0.050) (when included)
    - Component correlations: CBI1.1 0.407* (0.205); CBI3.1 0.201** (0.096) — most robust correlation with the third component (existence of a well-defined target).
  - Observations: 40 in reported columns; R2 values reported: 0.56, 0.48, 0.63, 0.46, 0.57.

- Interpretation
  - Greater transparency associated with:
    - More independent central banks (CBI1).
    - Better governance (higher RQ1, VA1).
    - More flexible de facto exchange rate regimes (REGIME1).
  - Most robust CBI-transparency link is with CBI component 3 (existence of a well-defined target).

### Effects of central bank independence on inflation
- Background and econometric challenges
  - De jure measures of independence may fail to capture de facto independence; turnover rate of central bank governors used as a de facto proxy.
  - Endogeneity and omitted variable bias are important concerns.

- OLS and first-difference results (Table 5 summary)
  - Period regressions (levels) confirm weak relationship between de jure CBI and inflation; turnover rate (TURNOVER) related to inflation with predicted sign in earlier studies.
  - First-differenced specification (exploiting two data points) yields predicted negative effects of both de jure and de facto independence measures on inflation.
  - Component-level result: CBI3 (single, well-specified inflation or price-level target) is the aspect most likely to deliver low inflation (reported coefficient in change regressions: CBI3 -0.196** (0.084)).

- Quantitative example from text
  - Average increase in de jure CBI (around .25) associated with a drop in inflation of more than 5 percentage points compared to a situation where CBI does not increase at all (for countries whose inflation was already in single digits; with a greater drop for countries with higher initial inflation).

- Instrumental variables (IV) results (Table 6)
  - De jure independence instrumented with governance measures: Rule of Law (RL1) and Voice and Accountability (VA1).
  - IV estimates (2SLS and LIML) find a statistically significant negative effect of D.CBI on D.INF (change in inflation).
  - Selected IV results (as presented):
    - 2SLS: D.CBI -0.937*** (0.328)
    - LIML: D.CBI -0.938*** (0.329)
    - LIML with component: D.CBI3 -0.798** (0.352)
  - Instrument diagnostics:
    - Cragg-Donald (C-D) statistics and critical values discussed; some evidence of weak instruments based on C-D compared to Stock and Yogo critical values.
    - Anderson-Rubin (A-R) test p-values show significance (A-R [p] 0.007, 0.007, 0.002, 0.007 across columns reported), indicating robustness of the negative effect even under weak-instrument concerns.
    - Overid test p-values indicate instruments are not rejected (e.g., Overid test [p] 0.910, 0.910, 0.230, 0.86 in reported columns).

- Robustness
  - Using an unweighted CBI measure (LVAU) yields nearly identical results.

### Effects of transparency on private-sector information use and forecast behavior
- Concept and measure
  - Private-sector use of private vs. public information measured by VAR/MSE ratio:
    - VAR = variance of private-sector forecasts for inflation in the following year.
    - MSE = mean square error of those forecasts (using next year’s actual inflation).
    - The VAR/MSE ratio lies on the unit interval and declines when public information becomes relatively more accurate.
  - A simple signal-extraction model demonstrates r = VAR/MSE = (γβ)/(αβ + γβ) and ∂r/∂α ≤ 0 (higher public-signal accuracy α reduces r).

- Empirical results (Table 7)
  - Sample limited by forecast dataset coverage: maximum sample size of 28 observations.
  - Baseline OLS specification (column (1)) finds predicted negative correlation between transparency (TRANS1) and VAR/MSE ratio (r):
    - TRANS1 -0.322** (0.146)
    - OPEN1 -0.853** (0.398) (in baseline)
    - RQ1 0.117*** (0.0402)
    - Constant 0.423*** (0.0963)
    - Observations: 28; R2 = 0.27.
  - Components considered individually (columns (2)–(6)):
    - T2.1 (Economic transparency) coefficient: -0.273** (0.105)
    - T5.1 (Operational transparency) coefficient: -0.205*** (0.0627)
    - Other components less robust individually (T1.1, T3.1, T4.1 not consistently significant).
  - Full specification including all components (column (7)):
    - Economic and operational components remain negative and statistically significant (T2.1 and T5.1).
    - Observations: 28; R2 = 0.47.

- Quantitative interpretation from text
  - A one standard deviation increase in the transparency score (around .22) is associated with a seven percentage point decline in the VAR/MSE ratio.
  - Illustrative mapping to the model: if public and private information were initially equally accurate (α = β) and initial r = 0.26, a .07 decrease in r corresponds to an increase in public information accuracy of around 70 percent (α1 = 1.7*β).

- Interpretation
  - Greater transparency, particularly economic transparency (release of forecasts) and operational transparency, is associated with greater private-sector use of public information and reduced dispersion across forecasters.
  - Results support the notion that publication of forecasts and related information improves the informativeness of the public signal.

### Conclusions and implications
- Summary of findings
  - Marked increase in CBI since indices were first constructed in the late 1980s; developing and emerging market economies showed the most comprehensive increases across all CBI dimensions.
  - Increases in independence concentrated on relations between executive and monetary authority (dispute resolution, central bank financing) for both advanced and emerging/developing economies.
  - Increased independence related to: higher initial inflation, lower initial CBI, more democratic initial political systems, and initially less flexible exchange rate arrangements.
  - De jure legal independence shows limited direct effect on inflation in levels regressions, but first-difference and IV approaches uncover a statistically significant negative effect of CBI on inflation (supporting theoretical predictions).
  - Transparency scores increased significantly only for advanced economies over the 1998–2006 period; transparency positively correlated with governance quality, central bank independence, and exchange rate flexibility.
  - Greater transparency is associated with more use of public information by private-sector forecasters (reduced VAR/MSE), especially when transparency involves the public release of central bank forecasts (economic transparency) and operational information.

- Policy and research implications
  - Institutional reforms enhancing CBI—especially those that include a well-specified policy target—appear most effective in reducing inflation; attention to de facto independence and governance preconditions is important.
  - Enhancing transparency (economic and operational) can substantially improve the informativeness of the public signal and reduce forecast dispersion; releasing forecasts is central to these transparency benefits.
  - Measurement and identification issues remain important:
    - Transparency measurement sensitive to language and self-reporting differences across periods.
    - Better instruments for both CBI and transparency are a priority for future work to identify causal channels more cleanly.
  - The transparency index methodology can be replicated to extend country coverage and build time series going forward.

*Italic line: Source: _wp08119 - section 5 of the paper. We find that greater transparency—in particular relating to the release of (IMF working paper excerpt, section 5).*

### 1.         Time         Periods

### _wp08119 - 1.         Time         Periods

### Time periods and general construction
- The current period is considered to cover 2002-06, although in some cases data is available only through 2005.
- For the initial period:
  - Analysis using the central bank independence and turnover data covers the period 1987-91.
  - Analysis using the transparency data covers the period 1997-2001, unless otherwise stated.
- Generally, data for the current and initial periods are constructed from averages of the available data (usually at an annual frequency) over the relevant period.

### Central Bank Independence (CBI) data
- Data for the 1980s is taken from Cukierman, Webb, and Neyapti (1992).
- Data for the current period is constructed by the authors using the IMF’s Central Bank Laws Database, which collates the text of central bank laws and was current through 2003 when the authors constructed the index.
- The scores reflect the authors’ interpretation of the laws’ provisions, following Cukierman, Webb, and Neyapti’s methodology.
- Table A1 summarizes the index and its components.

### Turnover data
- Data for the 1980s (covering 1980-89) taken from Cukierman, Webb, and Neyapti (1992).
- Data for the current period (covering 1995-2004) constructed by the authors using Morgan Stanley’s 2005 Central Bank Directory, with some additional material on governors’ terms taken from central bank websites.

### Transparency data
- Data for the late 1990s is taken from Fry and others (2000), based on responses by central banks to a survey conducted in 1998 by the Bank of England.
- Data for the current period was constructed by the authors based on their reading of documents and information available in English on central banks’ websites during 2006 and early 2007.
  - The source for question 1.1 is derived from the CBI measure and is therefore ultimately based on the central bank law database.
- The subindices and overall index are constructed using the methodology outlined in Table A2.

### Political and institutional data
- Six political and institutional variables are taken from the World Bank’s Aggregate Governance Indicators Dataset (Kaufmann, Kraay and Mastruzzi, 2006).
  - The original data are annual for 2002-05 and biannual for 1996-2000.
  - For analysis using the CBI/Turnover data, no observations are available for the initial period (1987-1991).
  - For analysis using the transparency data, the measure for the first period (1997-2001) is the average of 1998 and 2000.
  - In all cases, the measure for the second period (2002-2006) is the average of the four available years (2002-05) for each variable.
- One further political variable – the democracy score – is obtained from the Polity IV Dataset (Marshall and Jaggers, 2001).
  - These variables are available annually through 2004.

### Real GDP per capita and openness
- Variables are taken from the Penn World Tables, version 6.2 (Summers and Heston, 1988).
  - Annual data is available through 2004 (so that current period measures are based on 2002-04 data only).
  - The openness measure is the sum of exports and imports divided by GDP, all at current prices.

### De facto exchange rate regime
- Data derived from the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions (de facto methodology adopted in 1997); retroactively updated in Bubula and Ötker-Robe, 2002.
- The data is monthly (from January 1990 through June 2004). The authors updated the information to end-2006 using the semi-annual updates available at the IMF site.
- For the initial period (CBI data) the measure is therefore derived only from the 1990-91 data.
- The data scores the de facto exchange rate regime from 1 to 8:
  - 1 is the most restrictive (no national currency)
  - 8 is the least restrictive (independently floating)
  - each increase in score corresponds to a decrease in the degree of restrictiveness
- For each time period considered, the median measure for the period in question is taken.

### Inflation
- Annual inflation data (period average inflation, percent per annum) is taken from the IMF’s International Financial Statistics.
  - Data is generally available up to 2006.
- The percentage inflation rate (π) is transformed using the inflation tax transform: π/(100+π).

### Inflation forecasts
- Forecast data is taken from Consensus Economics, covering all available periods.
  - Data for most countries is available monthly; for some countries, only bi-monthly data are available.
  - Coverage commences in different years (mostly during the 1990s) and is available through end-2006.
  - In all cases, the forecast considered here is for the following calendar year.

### Country samples (Table A1) — Central Bank Independence
- Common to both samples — Advanced Economies:
  - Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Israel, Italy, Japan, Korea, Luxembourg, Netherlands, New Zealand, Norway, Portugal, Singapore, Spain, Sweden, Switzerland, United Kingdom, United States
- Common to both samples — Emerging Market and Developing Economies:
  - Argentina, Bahamas, Barbados, Bolivia, Botswana, Brazil, Chile, China, Colombia, Congo, Costa Rica, Egypt, Ethiopia, Ghana, Honduras, Hungary, India, Indonesia, Kenya, Lebanon, Malaysia, Malta, Mexico, Morocco, Nepal, Nicaragua, Nigeria, Pakistan, Peru, Philippines, Poland, Qatar, Romania, Serbia, South Africa, Tanzania, Thailand, Turkey, Uganda, Uruguay, Venezuela, Zambia, Zimbabwe
- Crowe-Meade only Emerging Market and Developing Economies:
  - Albania, Belarus, Bosnia and Herzegovina, Bulgaria, Croatia, Czech Republic, Ecuador, El Salvador, Estonia, Guatemala, Jamaica, Kazakhstan, Kuwait, Latvia, Lithuania, Macedonia, Namibia, Oman, Paraguay, Russia, Slovakia, Slovenia, Sri Lanka, Trinidad and Tobago, Tunisia, Turkmenistan, United Arab Emirates
  - Note: We dropped these countries due to unavailability of 2003 central bank law.
- Cukierman, Webb, and Neyapti (1992) only:
  - Panama, Samoa, Taiwan, Province of China
- Notes:
  - We use the country classification scheme used by the IMF in its World Economic Outlook publication.
  - Serbia replaces Yugoslavia in the updated sample.
  - We group the Crowe-Meade independence scores using the classification of advanced and emerging market/developing economies in the IMF’s World Economic Outlook. However, we use Cukierman, Webb, and Neyapti (1992) country groupings when replicating their results; the main difference is that they classify Greece, Israel, Korea, Portugal, and Singapore as developing countries.

### Country samples (Table A1) — Transparency
- Advanced Economies (24):
  - Australia, Austria, Belgium, Canada, Finland, France, Germany, Greece, Iceland, Ireland, Israel, Italy, Korea, Luxembourg, Netherlands, New Zealand, Norway, Portugal, Singapore, Spain, Sweden, Switzerland, United Kingdom, United States
- Emerging Market and Developing Economies (13):
  - Argentina, China, Czech Republic, Hungary, India, Indonesia, Mexico, Poland, Russia, Slovakia, Slovenia, South Africa, Turkey
- Note: Countries with formal inflation targeting regimes are noted in italics in the source table.

### Construction and coding of transparency measure (Table A2)
- Transparency categories follow Eijffinger and Geraats (2006) and Fry and others (2000) survey questions.
- Coding examples (preserving exact coding):
  - 1.1: Is there a statutory objective?
    - 1: Single objective of price stability or price stability objective does not conflict with other objectives
    - .5: Price stability objective potentially conflicts with other objectives
    - 0: Objectives do not include price stability or no objective
  - 1.2: Is there an explicit numerical target for prices or inflation?
    - 1: Yes
    - 0: No
  - 2.1: Does the central bank publish surveys that could be used to estimate inflation expectations?
    - 1: Yes
    - 0: No
  - 2.2: Does central bank publish any forward-looking analyses such as forecasts?
    - 1: Words AND numbers / figures
    - .5: Words OR numbers / figures
    - 0: Neither
  - 3.1: Does central bank publish minutes of policy meetings?
    - 1: Yes
    - 0: No
  - 3.3: Does central bank publish voting patterns of monetary policy committee?
    - 1: Yes
    - 0: No
  - 4.1: Does central bank publish explanations on day policy changed?
    - 1: Yes
    - 0: No
  - 4.2: Does central bank publish explanations on day policy does not change?
    - 1: Yes
    - 0: No
  - 5.1: Does central bank publish discussion of risks to outlook or forecast?
    - 1: Words AND numbers / figures
    - .5: Words OR numbers / figures
    - 0: Neither
  - 5.2: Does central bank publish discussion of shocks or forecast errors after the fact?
    - 1: Yes
    - 0: No
- Source notes:
  - Source for question 1.1 is from the Crowe-Meade CBI measure based on 2003 law.
  - Source for all other questions are website and publications of the central banks. For some countries, information in English is provided with a delay; the authors did not consider the language of the information when measuring transparency.
- Overall transparency index defined as the unweighted average of 5 categories; each category is unweighted average of all subcategories.

*Source: _wp08119 - 1.         Time         Periods (PDF chapter).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08119.pdf_