## _wp0641

## Source details

**Canonical URL:** [_wp0641](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp0641.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp0641.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp0641.pdf.json)

---

### Introduction and research question
- Monetary policy in the euro area is conducted within a multicountry, multicultural, and multilingual context following the formation of the European Economic and Monetary Union (EMU).
- Key research questions:
  - Have economic agents in different member countries been able to adjust and understand equally well the European perspective of monetary policy?
  - To what extent are expectations about monetary policy still influenced by different national and cultural backgrounds?
  - To what extent do differences in the ability to forecast monetary policy decisions reflect permanent information asymmetries related to geographic proximity to the ECB and country-specific factors?

### Data and measurement
- Database and coverage:
  - Source: surveys conducted by Reuters.
  - Professional financial analysts from 120 institutions in 24 countries.
  - Time period: 1999 to 2005.
  - About 4,500 observations; forecasts polled prior to ECB Governing Council meetings between January 1999 and January 2005.
  - Panel averages: on average about 30 institutions per meeting in early years and about 60 institutions toward the end of the sample.
  - About half the observations are from financial institutions with headquarters in euro area countries.
- Information collected:
  - Expected ECB policy rate for the meeting ahead (interest rate paid for the ECB’s main refinancing operations facility).
  - Policy rate with the highest probability (most likely rate).
  - Expected speed of ECB decision-making measured by expected number of meetings until the next change in the policy rate (meetings-to-change).
  - Probability distributions around analysts' point estimates, timing, direction, and size of expected changes, often expected year-end interest rate levels.
- Data rationale:
  - Survey responses are generally public and reflect institutions' client recommendations.
  - Most institutions participate regularly, enabling client evaluation of forecasting performance.

### Main empirical findings
- Heterogeneity in forecast accuracy:
  - Top 10 percent of institutions have a forecast error on average 8 basis points smaller than the error made by the worst 10 percent of performers.
  - Typical ECB policy rate change in the sample period: 25 basis points; observed magnitudes of differences reach about one-third of this change.
  - Observed average forecasting error for the full sample period: about 6 basis points.
  - Average absolute forecasting error by institution decile:
    - Most accurate institutions: around 2 basis points average error over 1999-2005.
    - Least accurate institutions: more than 10 basis points average error.
  - Difference between best and worst deciles by subperiod:
    - Around 12 basis points in the period prior to mid-2001.
    - Around 10 basis points in the post-2001 period.
  - Sample means of average forecasting errors: about 5 and 6 basis points in the pre- and post-2001 periods, respectively.
- Geography and information advantages:
  - Analysts headquartered in Frankfurt or working for institutions with a subsidiary in Frankfurt tend to have significantly lower forecasting errors.
  - Analysts headquartered in or with subsidiaries in London/United Kingdom show good performance, though coefficients are sometimes not significant at conventional levels.
  - Forecast accuracy tends to be lower with greater geographical distance to Frankfurt.
  - Financial institutions with headquarters or a subsidiary in Frankfurt show average forecasting errors lower by 1 to 1.5 basis points relative to peers (compared to the 8 basis points decile gap).
- Macro conditions and history:
  - A one percentage point larger inflation rate differential leads to forecasting errors larger by half a basis point on average.
  - A one percentage point larger unemployment rate differential increases forecasting errors by one-tenth of a basis point on average.
  - Deviations of national inflation and unemployment rates from the euro area average have significant positive effects on absolute forecasting error; the coefficient for relative inflation exceeds that for relative unemployment.
  - Analysts reporting from countries with historically higher central bank independence tend to make better forecasts (central bank independence matters primarily in the post-2001 subperiod).
- Time variation and learning:
  - Aggregate accuracy pattern: started at fairly high levels in 1999, decreased sharply in 2000, stabilized in 2001, and improved thereafter but not to 1999 levels.
  - Structural break: January 2001 identified as a breakpoint; sample split into pre-2001 and post-2001.
  - Geography coefficients (Frankfurt headquarter, Frankfurt subsidiary, vicinity) are more negative and statistically significant pre-2001 than post-2001.
  - Analysts headquartered in Frankfurt saw performance decline relative to sample average in post-2001; institutions with subsidiaries in Frankfurt saw only slightly weaker relative accuracy post-2001.
  - Confusion about the relative importance of national versus euro area macro conditions appears to have hampered prediction precision more post-2001.
- Systematic versus idiosyncratic heterogeneity:
  - A substantial part of published forecasts is explained by a systematic component extracted via bank-specific reaction functions (Taylor-type rules).
  - Decomposition: observed error eit = rt − rit^e is split into systematic component sit (based on Taylor-type rules using real-time euro area HICP, industrial production, M3, and consumer confidence) and unsystematic residual uit; analysis focuses on |sit|.
  - Geographic patterns for |sit| mirror overall observed error: analysts reporting for German banks or other financial institutions perform significantly better; Austrian and French forecasters significantly underperform.
  - Overall conclusion: differences in modeling the ECB’s policy decisions (systematic components) underlie observed differences in forecasting accuracy; some systematic differences are transitional (learning) while others persist.

### Econometric results and Taylor-type rules (selected quantitative findings)
- Sample sizes for regressions:
  - # observations: 4,242 (full period) for expected rate regressions; pre-2001: 1,414; post-2001: 2,828.
  - # observations for most likely rate regressions: 4,314 (full period); pre-2001: 1,507; post-2001: 2,807.
  - # observations for meetings-to-change: 3,199 (all countries); euro area # obs 1,754.
- Selected regression coefficients (Table 2: explaining forecasting error, coefficients in %):
  - Headquarters in Frankfurt: -0.011 *** (0.003) [full period]; -0.017 *** (0.005) [pre-2001]; -0.007 * (0.004) [post-2001].
  - Subsidiary in Frankfurt: -0.015 *** (0.003) [full period]; -0.017 *** (0.005) [pre-2001]; -0.015 *** (0.004) [post-2001].
  - HQ or Subsidiary in London: -0.002 (0.003) [full period]; -0.007 (0.005) [pre-2001]; 0.001 (0.003) [post-2001].
  - Vicinity of Frankfurt: -0.001 (0.003) [full period]; -0.012 ** (0.005) [pre-2001]; 0.003 (0.004) [post-2001].
  - Relative inflation: 0.005 *** (0.002) [full period]; 0.008 *** (0.002) [pre-2001]; 0.006 *** (0.002) [post-2001].
  - Relative unemployment: 0.001 ** (0.001) [full period]; 0.002 *** (0.001) [pre-2001]; 0.001 ** (0.001) [post-2001].
  - Central bank independence: -0.002 (0.001) [full period]; -0.004 ** (0.002) in one subcolumn.
- Table 4 (explaining Taylor-type rule systematic error; coefficients in %):
  - Headquarters in Frankfurt: -0.008 *** (0.002) [full period]; -0.011 *** (0.004) [pre-2001]; -0.006 * (0.003) [post-2001].
  - Subsidiary in Frankfurt: -0.010 *** (0.002) [full period]; -0.011 ** (0.004) [pre-2001]; -0.010 *** (0.003) [post-2001].
  - Vicinity of Frankfurt: 0.000 (0.002) [full period]; -0.009 ** (0.004) [pre-2001]; 0.004 (0.003) [post-2001].
  - Relative inflation: 0.005 *** (0.001) [full period]; other reported column entries include 0.855 *** (0.001) and 0.011 *** (0.002).
  - Relative unemployment: 0.001 *** (0.000) [full period]; other reported column entries include 0.247 *** (0.001) and 0.003 *** (0.001).
  - Central bank independence: -0.002 * (0.001) [full period]; -0.004 *** (0.001) [pre-2001].
- Taylor-type rule estimates (Table 5; short-run coefficients; Panel A, all groups):
  - HICP inflation: pre-2001 coef. 3.651 *** (1.112); post-2001 coef. 6.184 *** (0.562).
  - M3: pre-2001 coef. 2.556 *** (0.527); post-2001 coef. 3.467 *** (0.287).
  - Industrial production: pre-2001 coef. 1.761 *** (0.172); post-2001 coef. 0.606 *** (0.094).
  - Consumer confidence: pre-2001 coef. 0.001 (0.001); post-2001 coef. 0.032 *** (0.001).
  - Previous rate: pre-2001 coef. 0.948 *** (0.008); post-2001 coef. 0.827 *** (0.004).
- Most likely rate and meetings-to-change findings:
  - Systematic error behavior for most likely rate similar to expected rate.
  - Headquarters or subsidiary in London: negative and insignificant for expected rate but negative and significant for most likely rate; for meetings-to-change HQ/subsidiary in London plays a more significant role.
  - Meetings-to-change mean and dispersion: All countries # obs 3,199; mean 5.40 (number of meetings); pre-post-change diff. 5.773; st. dev. 6.51.

### Key tabular and country-level statistics (selected exact figures)
- Table 1 (a): Absolute Error for expected rate by Country (in %)
  - All countries: # obs 4,242; mean 0.060; pre-post-change diff. 0.077; st. dev. 0.005.
  - Euro area: # obs 2,350; mean 0.057; pre-post-change diff. 0.078; st. dev. 0.002.
  - Non-euro area: # obs 1,892; mean 0.062; pre-post-change diff. 0.076; st. dev. 0.009.
  - Country examples (mean): Austria 0.087; France 0.070; United Kingdom 0.062; Germany 0.053; USA 0.063.
- Table 1 (b): Absolute Error for most likely rate by Country (in %)
  - All countries: # obs 4,314; mean 0.035; pre-post-change diff. 0.092; st. dev. 0.005.
  - Euro area: # obs 2,421; mean 0.034; pre-post-change diff. 0.091; st. dev. 0.002.
  - Country examples (mean): Austria 0.067; France 0.043; United Kingdom 0.036; Germany 0.030; USA 0.051.
- Table 3: Absolute Systematic Error for expected rate by Country (in %)
  - All countries: # obs 4,242; mean 0.073; pre-post-change diff. 0.086; st. dev. 0.026.
  - Euro area: # obs 2,350; mean 0.073; pre-post-change diff. 0.087; st. dev. 0.025.
  - Country examples (mean): France 0.085; United Kingdom 0.074; Germany 0.069; USA 0.060.
- Figure notes:
  - Figure 1 shows average absolute forecast error in basis points by decile (1999-2005, pre-2001, post-2001); allocation of institutions to deciles changes slightly over time.
  - Figure 2 date labels include: 01may1998, 13sep1999, 25jan2001, 09jun2002, 22oct2003, 05mar2005; the figure shows mean and min/max of policy rate forecasts polled by Reuters and the actual policy rate decided at each meeting.

### Policy implications and recommendations
- Persistent heterogeneity in expectation formation implies potential problems for uniform monetary policy transmission across EMU countries.
- Presence of systematic heterogeneity related to geography, country-specific macro conditions, and forecasting models indicates persistent informational frictions and agglomeration effects in euro area financial markets.
- Policy recommendation:
  - There is room for policies that foster convergence to a common expectation-formation process.
  - Example policy lever: careful and targeted central bank communication to reduce persistent informational frictions and promote a more unified expectation-formation process.
- Additional policy-relevant observations:
  - Some heterogeneity is transitory (learning), but a substantial part persists—markets have yet to converge on a homogeneous view of the ECB.
  - The ECB should be aware of differences in agents' ability to understand and anticipate policy across the euro area and can consider targeted guidance to support convergence.

### Avenues for further research (selected)
- Explore links between ECB communication and expectations on ECB policy (e.g., whether speeches or interviews targeted at particular audiences help reduce systematic heterogeneity).
- Investigate whether certain forms of ECB communication focus regional audiences more than others.
- Apply the framework to other central banks (for example, the U.S. Federal Reserve) to assess geography's role in expectation formation beyond the European context.

*Source: _wp0641 - References*

### References .............................................................................................................

### _wp0641 - References

### Introduction and research question
- Monetary policy in the euro area is conducted within a multicountry, multicultural, and multilingual context following the formation of the European Economic and Monetary Union (EMU).
- Key research questions:
  - Have economic agents in different member countries been able to adjust and understand equally well the European perspective of monetary policy?
  - To what extent are expectations about monetary policy still influenced by different national and cultural backgrounds?
  - To what extent do differences in the ability to forecast monetary policy decisions reflect permanent information asymmetries related to geographic proximity to the ECB and country-specific factors?

### Data and measurement
- Novel database of monetary policy expectations:
  - Source: surveys conducted by Reuters.
  - Coverage: professional financial analysts from 120 institutions in 24 countries.
  - Time period: 1999 to 2005.
  - Information included:
    - Expected ECB policy rates for upcoming Governing Council meetings.
    - The probability distribution around analysts' point estimates.
    - Expectations about future monetary policy steps.
- Rationale for data quality:
  - Survey responses are generally public and reflect institutions' client recommendations.
  - Most institutions participate regularly, enabling clients to evaluate forecasting performance.

### Main empirical findings
- Large heterogeneity in forecast accuracy:
  - Top 10 percent of all institutions have a forecast error that is on average 8 basis points smaller than the error made by the worst 10 percent of performers.
  - The magnitude of differences reaches about one-third of the typical ECB policy rate change of 25 basis points during the sample period.
- Geography matters:
  - Significant cross-country variance, especially in the early years of the sample.
  - Proximity to informational hubs improves accuracy:
    - Analysts headquartered in Frankfurt or working for institutions with a subsidiary in Frankfurt tend to have better forecast accuracy.
    - Good performance also observed for analysts based in London/United Kingdom or working for institutions with a subsidiary in the City.
  - Forecast accuracy tends to be lower with greater geographical distance to Frankfurt for other locations.
- National macroeconomic conditions influence accuracy:
  - Predictions of ECB policies become less precise if the forecaster is located in a country where inflation or unemployment rates deviate from the euro area average.
- Historical institutional factors:
  - Analysts reporting predominantly from countries with a history of relatively high central bank independence tend to make better forecasts of ECB behavior.
- Systematic versus idiosyncratic heterogeneity:
  - Systematic component extracted via bank-specific reaction functions (Taylor-type rules) explains a significant part of published forecasts.
  - Some systematic differences are transitional (indicative of learning), while others are persistent.
- Time variation and learning:
  - Splitting the sample into pre–2001 and post–2001:
    - Analysts headquartered in Frankfurt saw their performance decline relative to the sample average in the post–2001 period.
    - Relative forecast accuracy of financial institutions with subsidiaries in Frankfurt became only slightly weaker in the post–2001 period.
    - Confusion by forecasters over the importance of national relative to euro area macroeconomic conditions for ECB policy decisions appears to have hampered prediction precision more post–2001 than before.
- Robustness across performance measures:
  - Similar and stable pattern of forecast accuracy across countries and institutions for both:
    - Expected level of the ECB policy rate.
    - Expected timing of ECB policy rate adjustments.
  - Evidence that determinants of expectations about the level also influence expectations about timing.

### Relation to existing literature
- Novel focus on explaining heterogeneity in monetary policy expectations within the euro area.
- Connects to literature on heterogeneous transmission of ECB policy, trade in goods and financial assets, and home bias in portfolio allocation.
- Cites empirical literature showing information asymmetries and agglomeration effects (examples include Froot and Stein, 1991; Coval and Moskowitz, 1999, 2001; Bae, Stulz, and Tan, 2005).
- Highlights role of geographic proximity, sociocultural ties, and information spillovers in financial agglomeration (examples include Thrift, 1994; Faulconbridge, 2003; Strauss-Kahn and Vives, 2005).

### Policy implications and recommendations
- Persistent heterogeneity in expectation formation implies potential problems for uniform monetary policy transmission across EMU countries.
- Presence of systematic heterogeneity related to geography, country-specific factors, and forecasting models indicates persistent informational frictions and agglomeration effects in euro area financial markets.
- Policy implication:
  - There is room for policies that foster convergence to a common expectation-formation process.
  - Example policy lever: careful and targeted central bank communication to reduce persistent informational frictions and promote a more unified expectation-formation process.

### Paper structure (as provided)
- Section II: Dataset based on the Reuters survey and key stylized facts.
- Section III: Analysis of determinants of differences in forecasting abilities across economic agents (geography, country-specific economic conditions, history).
- Section IV: Bank-specific Taylor-type rule estimates to distinguish systematic and unsystematic components of forecasting errors.

*Source: _wp0641 - References*

### Section V summarizes the main findings and discusses policy implications.

### _wp0641 - Section V summarizes the main findings and discusses policy implications.

### Data and key stylized facts
- Sample size and scope
  - The data consist of about 4,500 observations.
  - Forecasts come from analysts in 11 euro area and 13 other countries polled by Reuters prior to ECB Governing Council meetings between January 1999 and January 2005.
  - The panel encompasses on average about 30 financial institutions per Governing Council meeting in the first years and an average of about 60 institutions toward the end of the sample.
  - Overall, the sample covers 120 institutions.
  - About half the observations are from financial institutions with headquarters in euro area countries.

- Nature of the Reuters poll and constructed series
  - Reuters asked analysts to attach probabilities to different scenarios for the ECB’s policy rate (the interest rate paid for the ECB’s main refinancing operations facility), including probabilities of changes by 50 or 25 basis points, timing, direction, size, and often expected year-end interest rate levels.
  - Three summary series were constructed from the poll:
    - expected ECB policy rate (expected rate) for the meeting ahead,
    - policy rate with the highest probability (most likely rate),
    - expected speed of ECB decision-making measured by the expected number of meetings until the next change in the policy rate (meetings-to-change).

- Forecasting accuracy: cross-sectional and temporal patterns
  - The ability to anticipate ECB policy decisions differs markedly across economic agents.
  - Average absolute forecasting error by institution decile:
    - Most accurate institutions: around 2 basis points average forecast error over 1999-2005.
    - Least accurate institutions: more than 10 basis points average error.
    - Difference of up to 8 basis points between best and worst deciles.
    - Observed average forecasting error for the full sample period: about 6 basis points.
    - Difference between best and worst deciles by subperiod:
      - Around 12 basis points in the period prior to mid-2001.
      - Around 10 basis points in the post-2001 period.
  - Sample means of average forecasting errors:
    - About 5 and 6 basis points, respectively, in the pre- and post-2001 period.
  - Forecast dispersion:
    - Dispersion in forecast accuracy remained fairly constant (and large) over time.
    - Standard deviation of mean absolute forecast error across institutions within each country for the expected rate ranges from 0.06 to 0.14, with an average of 0.08.
    - Standard deviation of the average national forecasting errors is 0.02, considerably lower than within-country variation.

- Time pattern and regime change
  - Aggregating across countries, forecast accuracy:
    - Started at fairly high levels in 1999,
    - Decreased sharply in 2000,
    - Stabilized in 2001,
    - Improved thereafter, but not to 1999 levels.
  - Suggests a structural break around early 2001 when the ECB moved from a policy of interest rate hikes to lowering rates.
  - Absolute forecasting error in periods of changing policy rates is more than four times larger than in calmer periods.

- Bias and efficiency
  - Despite differences in forecast accuracy among institutions, the forecasts generally are unbiased and efficient.
  - For the expected rate, the average forecast closely tracks the actual (post-meeting) ECB policy rate during most of the period, while individual forecasts vary more widely.

### The role of geography, macro conditions, and history (overview)
- Objective and econometric setup
  - Interest is in explaining differences in absolute forecasting error e_it = ||r_t - e_it^r||, where r_t is the policy rate after Governing Council meeting t and e_it^r is analyst i’s published expectation.
  - General model form: e_it = α + β y_it + γ_t + u_it, where y_it is a vector of analyst-specific or analyst-and-meeting-specific explanatory variables, γ_t are meeting-fixed-effects, and u_it is a residual.
  - Meeting-fixed-effects are included to control for unobservables such as selection idiosyncrasies, time trends in number of institutions polled, and the change from bimonthly to monthly ECB meetings in late 2001.
  - Institution-varying explanatory variables preclude inclusion of institution-fixed effects.

- Explanatory variable dimensions (as used in the analysis)
  - Geography
    - Headquarters in Frankfurt: dummy variable; one for institutions that file forecasts from Frankfurt in the majority of cases, zero otherwise.
    - Subsidiary in Frankfurt: dummy variable; marks institutions that report at least once from Frankfurt, but from other locations in the majority of cases.
    - Headquarter or subsidiary in London: dummy variable; identifies institutions reporting at least once from London that are not already included in the first two categories.
    - Vicinity: dummy variable; includes institutions not classified into the first three categories but that have their headquarters or a subsidiary within 1,000 kilometers of Frankfurt.
  - Macro conditions
    - Relative inflation: difference between euro area and national HCPI inflation rates based on information available at the time of the Reuters poll regarding meeting t; set to zero for institutions headquartered outside the euro area.
    - Relative unemployment: equivalent variable for unemployment rates.
  - History
    - Central bank independence: dummy variable; 1 if the pre-1999 Grilli, Masciandaro, and Tabellini (1991) index for central bank independence for the country in which a bank is headquartered is above the (unweighted) euro area average; -1 if the index is below the euro area average; set to zero for non-euro area financial institutions.

### Main econometric findings and interpretation (for the expected rate)
- Geography and information advantages
  - Analysts working for institutions headquartered in Frankfurt or operating a subsidiary there tend to have significantly lower forecasting errors than others.
  - Possible interpretations:
    - Proximity to the ECB: observing the ECB up-close may improve prediction.
    - Informational agglomeration: informal information spillovers in a financial center.
    - Historical experience: familiarity with Bundesbank behavior pre-1999 may have provided useful human capital post-1999.
    - Financial expertise and agglomeration effects in information processing.
  - The coefficient for headquarter or subsidiary in London is negative (suggesting better performance) but not significant at conventional levels.

- Macro conditions matter
  - Deviations of national inflation and unemployment rates from the euro area average have a significant positive effect on the absolute forecasting error.
  - The coefficient estimate for relative inflation exceeds that for relative unemployment, signaling a greater potential for national inflation experiences to influence the precision of forecasters located in a particular euro area country.
  - Two interpretations:
    - Analysts retain a national perspective and may overweight national real-time data instead of properly aggregating to euro area signals.
    - High values of relative macro variables indicate increased overall forecasting uncertainty (easier to predict policy when member economies move in sync).

- Magnitude of effects
  - Financial institutions with their headquarters or a subsidiary in Frankfurt show an average forecasting error that is lower by 1 to 1.5 basis points than their peers, compared to the 8 basis points difference overall between the best and worst forecasters (as reported in the decile statistics).

- Robustness and further analysis
  - Table 2 presents OLS estimates for the full sample, with model (1) excluding and model (2) including “central bank independence” to check robustness and potential correlation with macro condition variables.
  - The analysis is extended by looking at subsamples and by examining the most likely rate and meetings-to-change variables (results broadly similar).

*Source: _wp0641 - Section V summarizes the main findings and discusses policy implications.*

### 1. With respect to the macroeconomic variables, a one percentage point larger inflation rate

### _wp0641 - 1. With respect to the macroeconomic variables, a one percentage point larger inflation rate

### Macroeconomic variables: quantitative effects on forecasting errors
- A one percentage point larger inflation rate differential leads to forecasting errors that are larger by half a basis point on average.
- A one percentage point larger unemployment rate differential increases forecasting errors by one-tenth of a basis point.
- Relative inflation, relative unemployment, and central bank independence matter significantly primarily in the post-2001 subperiod.

### Temporal break and sample split
- Standard tests indicate that January 2001 is a structural breakpoint in the empirical model; the sample is split about halfway into pre-2001 and post-2001 subperiods.
- Coefficients for geographic variables (headquarters in Frankfurt, subsidiary in Frankfurt, vicinity to Frankfurt) are more negative and statistically more significant during the pre-2001 subperiod than post-2001.
- The London headquarters/subsidiary negative coefficient is weakly present in the pre-2001 subperiod and disappears post-2001.
- The average of relative inflation triples from pre-2001 to post-2001 in the sample.

### Geography, learning, and institutional history
- A plausible explanation for the declining importance of geography is learning:
  - Forecasters throughout the euro area may have learned the ECB’s behavior, adopted necessary tools, or adjusted models to new realities.
  - Learning might also be passive (changes in macroeconomic framework or ECB policy setup improving model fit) or passive “unlearning” (failure to adjust previously well-performing models).
- Evidence in Section IV.C suggests both active and passive forms of learning may have played a role.
- Institutional history (central bank independence) is not significant in the full sample but matters more in the post-2001 subperiod:
  - Forecast accuracy is higher for financial institutions reporting from countries with long-standing traditions of independent central banking.
  - This result is estimated controlling for geographical factors (i.e., it measures institutional history additional to location effects).

### Endogenous variables beyond expected rate
- Results for macro conditions extend to the "most likely rate" and "meetings-to-next-change" endogenous variables.
- Geography’s explanatory power is somewhat weaker for most likely rate and meetings-to-next-change; institutional history has a larger role for these variables.
- Possible reason: alternative indicators of ECB activity (most likely rate, meetings-to-next-change) may be noisier than the expected policy rate.
- Forecasters in subsidiaries in Frankfurt may coordinate on a mean prediction for the ECB’s policy rate while differing on the full distribution or timing of the next rate change.
- The meetings-to-change variable contained outliers (upper 5 percent dropped due to likely typos); no such cleaning was necessary for expected rate and most likely rate.

### Decomposition of forecasting errors: systematic versus unsystematic
- Observed error for forecaster i at meeting t: eit = rt − rit^e (post-meeting policy rate minus published forecast).
- Forecast composed of:
  - Systematic component: rit^e (approximated by a backward-looking Taylor-type rule using real-time euro area HICP, industrial production, M3, and consumer confidence from Bloomberg as of the Friday prior to meeting t).
  - Unsystematic component: uit^ (gut feeling or analyst-specific residual).
- Forecaster-level regression (allowing upward vs downward interest rate cycle and pre-/post-2001 differentiation) estimates coefficients αi1, αi2, βi0, and βik, with residuals uit depicting the unsystematic component.
- Observed error decomposition: eit = sit + uit, where sit is the systematic component and uit the unsystematic; analysis focuses on |sit| (absolute systematic error).

### Findings on the systematic error (|sit|)
- Geographic patterns for |sit| mirror those for the overall observed error:
  - Analysts reporting for German banks or other financial institutions perform significantly better than the euro area average.
  - Austrian and French forecasters significantly underperform.
- Time variation similar to |eit|: relatively high accuracy at the beginning and end of the sample, higher average errors in between; average systematic error is as low (or lower) toward the end as at the beginning.
- |sit| is significantly lower for forecasters in the vicinity of Frankfurt, or reporting from headquarters or a subsidiary in Frankfurt; these effects are stronger during the pre-2001 subperiod.
- The coefficient for vicinity of Frankfurt turns significantly positive during the post-2001 subperiod.
- Relative inflation, relative unemployment, and central bank independence are significant determinants of |sit| primarily in the post-2001 subperiod.
- Overall conclusion: differences in modeling the ECB’s policy decisions (systematic, Taylor-type-rule-based components) underlie observed differences in forecasting accuracy.

*Source: Excerpt from the provided IMF working paper content (Tables and Appendices referenced in text).*

### Section III.A are also the best performers with respect to the systematic forecast error, and

### _wp0641 - Section III.A are also the best performers with respect to the systematic forecast error, and

### C. Dissecting the Taylor-Type Rules for the Expected Rate
- Purpose:
  - Investigate how models of ECB behavior differed across forecasters by sorting individual forecasters according to characteristics suggested by the analysis of forecasting errors and estimating group-specific Taylor-type rules.
- Benchmark:
  - Table 5 reports estimates for a pooled Taylor-type rule for the entire sample of financial institutions.
  - Short-run coefficients are reported as they are more meaningful for forecasts of imminent policy decisions.
  - The model allows coefficients to differ between the pre-2001 and post-2001 periods; the last column of Table 5 indicates whether coefficients changed between periods in a significant way.
- Aggregate findings:
  - Estimated reaction functions resemble standard central bank reaction functions.
  - On the nominal side, expectations for the ECB policy rate increase with HICP inflation and the growth rate of M3.
  - On the real side, changes in industrial production and the level of consumer confidence are positively related to the expected rate.
  - High estimated coefficients for previous rate (the policy rate prevailing before the meeting) suggest a high degree of interest rate smoothing and a protracted reaction to other variables.
- Time variation (pre-2001 vs. post-2001):
  - Analysts expected the ECB to react faster in the post-2001 period: the coefficient for previous rate drops (see last column of Table 5, Panel A).
  - Relative importance shifts: HICP inflation and consumer confidence gain importance; industrial production's relative weight decreases.
- Subgroup analysis (Panel B of Table 5):
  - Results similar across geographic, macroeconomic, and institutional-history splits; increased importance of HICP and consumer confidence and decreased weights on industrial production and previous rate are broadly shared.
  - Financial institutions positioned in Frankfurt gave greater weight to inflation but lower weight to money growth than institutions outside Frankfurt; most between-group differences are not statistically significant (limited power due to roughly 120 observations split into subgroups).
- Methodological note (footnote 21):
  - Decomposition of overall forecasting error into systematic and unsystematic parts in the previous section is based on individual, institution-specific Taylor-type rules; group-specific Taylor-type rules are the analytical tool used here.

### D. Do These Results Extend to Other Measures of Systematic Expectations?
- General approach:
  - The decomposition tools can be applied to alternative independent variables: most likely rate and meetings-to-change. Detailed results are in Appendices A1 and A2.
- Most likely rate:
  - Systematic forecast error behavior for most likely rate is similar to expected rate.
  - Country-by-country perspective (Table A1.2) identifies the same significant under- and over-performers as Table 3 for expected rate.
  - Regression results (Table A1.3) indicate comparable geographic, macroeconomic, and historical patterns for both variables.
  - Headquarters or subsidiary in London: estimated negative and insignificant for expected rate (Table 4) but negative and significant for most likely rate.
  - Intertemporal pattern: splitting sample pre-2001 and post-2001 (Table A1.3) yields results akin to Table 4. Relative inflation, relative unemployment, and central bank independence influence forecasting error primarily in the later half of the sample; significant effects of most geographic variables disappear post-2001.
  - Estimated Taylor-type rule coefficients and their changes over time for most likely rate mirror those for expected rate (Tables A1.4).
- Meetings-to-change:
  - Meetings-to-change can be modeled by a Taylor-type rule.
  - Distribution of systematic meetings-to-change forecasting errors across countries differs from expected-rate and most-likely-rate distributions, but geographic, macroeconomic, and historical patterns are broadly consistent.
  - Examples:
    - Forecasters headquartered in Germany do not perform better than average.
    - Forecasters from the United Kingdom outperform peers (Table A2.2).
  - Regression-based analysis (Table A2.3):
    - Geography and institutional history matter in the same way as for expected rate or most likely rate.
    - Headquarters or subsidiary in London plays a more significant role for meetings-to-change than for the other forecast variables.
  - Learning and time effects:
    - Learning plays a large role in explaining systematic errors for meetings-to-change (Table A2.3).
    - Being located in Frankfurt or London tends to significantly reduce forecasting errors pre-2001; geography variables do not matter during the post-2001 subperiod.
    - History of central bank independence is associated with better forecasting performance only post-2001.
  - Taylor-type rule findings (Table A2.4):
    - Rules are qualitatively very similar to those for expected rate and most likely rate.
    - All coefficients in the Taylor-type rule based on the full panel are estimated significant and have expected signs (Panel A).
    - Under rate hike expectations:
      - Higher inflation leads forecasters to expect an interest rate hike sooner (negative coefficient → fewer meetings before next change).
      - Larger M3, industrial production, and consumer confidence also imply sooner expected hikes.
      - If current interest rates are high under rate hike expectations, the urge to increase them further diminishes (positive coefficients).
    - Over time, changes replicate those for the other forecasts, except inflation has become less prominent; this change is rather uniform across groups (Panel B of Table A2.4).
- Methodological notes:
  - Interpretation of meetings-to-change depends on expected direction of monetary policy; coefficients should flip signs depending on expected direction (footnote 23).
  - For meetings-to-change Taylor-type rules, coefficients for periods of policy rate declines have been multiplied by (–1) to ensure comparability (footnote 25).

### V. Conclusions
- Context:
  - EMU assigns the ECB monetary policy for 12 countries with varying histories of inflation, policy strategies, and economic environments; policy is conducted euro-area-wide but in a multicountry context.
- Main empirical findings:
  - Using a database on forecasts of ECB policy decisions of 120 financial institutions in 24 countries since 1999, the study finds marked differences in ability to anticipate ECB decisions.
  - A substantial part of these differences is persistent and systematically explained by:
    - Geography (headquartered in Frankfurt or having a subsidiary in Frankfurt perform substantially better; some informational advantage for institutions based in the City of London).
    - Distance: institutions closer to Frankfurt show better forecasting performance.
    - Country-specific economic conditions: better forecasting when located in countries with low levels of inflation and/or unemployment relative to the (weighted) euro area average.
    - History: institutions in countries with a history of relatively high central bank independence tend to make better forecasts.
  - Most heterogeneity in forecasting performance is related to different models of ECB behavior; there is little evidence of learning.
- Interpretation:
  - Information asymmetries and agglomeration effects likely play a role in the ability to anticipate monetary policy; proximity and face-to-face contacts facilitate knowledge production and performance.
- Policy implications:
  - Expectations are crucial in the transmission of monetary policy.
  - The ECB should be aware of differences in agents' ability to understand and anticipate policy across the euro area.
  - Some heterogeneity is transitory, but a substantial part persists—markets have yet to converge on a homogeneous view of the ECB.
  - There is scope for continuous guidance of convergence through careful and targeted communication policy by the central bank.
- Avenues for further research:
  - Explore links between ECB communication and expectations on ECB policy (e.g., whether speeches or interviews targeted at particular audiences help reduce systematic heterogeneity).
  - Investigate whether certain forms of ECB communication focus regional audiences more than others.
  - Apply the framework to other central banks (e.g., U.S. Federal Reserve) to assess geography's role in expectation formation beyond the European context.

*Source: Excerpt from the supplied IMF PDF content unit.*

### References

### _wp0641 - References

### Bibliographic sources cited
- Angeloni, Ignazio, and Michael Ehrmann, 2003, “Monetary Transmission in the Euro Area: Early Evidence,” Economic Policy, Vol. 37, pp. 469-492.
- Angeloni, Ignazio, and Michael Ehrmann, 2004, “Euro Area Inflation Differentials,” ECB Working Paper No. 388 (Frankfurt: European Central Bank).
- Bae, Kee-Hong, René Stulz, and Hongping Tan, 2005, “Do Local Analysts Know More? A Cross-Country Study of the Performance of Local Analysts and Foreign Analysts,” NBER Working Paper 11697 (Cambridge, Massachusetts: National Bureau of Economic Research).
- Cecchetti, S., 2001, “Legal Structure, Financial Structure and the Monetary Policy Transmission Mechanism,” in The Monetary Transmission Process: Recent Developments and Lessons for Europe, ed. By Deutsche Bundesbank (London: Palgrave), pp. 170-194.
- Clark, Gordon, 2002, “London in the European Financial Services Industry: Locational Advantage and Product Complementarities,” Journal of Economic Geography, Vol. 2(4), pp. 433-53.
- Cook, Gary, Naresh Pandit, Jonathan Beaverstock, Peter Taylor, and Kathy Pain, 2004, “The Clustering of Financial Services in London,” Typescript, Liverpool University.
- Coval, J.D. and T.J. Moskowitz, 1999, “Home Bias at Home: Local Equity Preference in Domestic Portfolios,” Journal of Finance, Vol. 54 (6) pp. 2045 – 2073.
- Coval, J.D. and T.J. Moskowitz, 2001, “The Geography of Investment: Informed Trading and Asset Pricing,” Journal of Political Economy, Vol. 109(4), pp. 811 – 841.
- Dornbusch, R., C.A. Favero, and F. Giavazzi, 1998, “The Immediate Challenges for the European Central Bank,” Economic Policy Vol. 26, pp. 17-64.
- Dvorak, Tomas, 2005, “Do Domestic Investors Have an Information Advantage? Evidence from Indonesia,” Journal of Finance, Vol. 60(2), pp. 817 – 839.
- Ehrmann, Michael, and Marcel Fratzscher, 2005, “Equal Size, Equal Role? Interest Rate Interdependence Between the Euro Area and the United States,” Economic Journal Vol. 115, pp. 930-950.
- Faulconbridge, James, 2004, “London and Frankfurt in Europe’s Evolving Financial Centre Network,” Area, Vol. 36(3), pp. 235-44.
- Froot, K.A., and Jeremy Stein, 1991, “Exchange Rates and Foreign Direct Investment: An Imperfect Capital Markets Approach,” Quarterly Journal of Economics, Vol. 106 (November 1991), pp. 1191-1217.
- Ginsburg, Victor, Shlomo Weber, and Ignacio Ortuño-Ortín, 2005, “Disenfranchisement in Linguistically Diverse Societies: The Case of the European Union,” Journal of the European Economic Association, Vol. 3(4), pp. 946-965.
- Gordon, Roger H., and A. Lars Bovenberg, 1996, “Why Is Capital So Immobile Internationally? Possible Explanations and Implications for Capital Income Taxation,” American Economic Review Vol. 86(5), pp. 1057 – 1075.
- Grilli, Vittorio, Donato Masciandaro, and Guido Tabellini, 1991, “Political and Monetary Institutions and Public Financial Policies in the Industrial Economies,” Economic Policy, Vol. 13, pp. 341-92.
- Grote, Michael, 2004, “The Rise (and Fall?) of the Financial Center Frankfurt—Locations of Foreign Banks in Germany 1949-2003,” Typescript, University of Frankfurt.
- Krugman, Paul, 1998, “What’s New About the New Economic Geography?,” Oxford Review of Economic Policy, Vol. 14(2), pp. 7-17.
- Mihov, I., 2001, “Monetary Policy Implementation and the Transmission in the European Monetary Union,” Economic Policy, Vol. 33, pp. 369-402.
- Mojon B., and G. Peersman, 2003, “A VAR Description of the Effects of Monetary Policy in the Individual Countries of the Euro Area,” in Monetary Policy Transmission in the Euro Area, ed. by I. Angeloni, A. Kashyap, and B. Mojon, (London: Cambridge University Press), pp. 56-74.
- Portes, Richard, Hélène Rey, and Yonghyup Oh, 2001, “Information and Capital Flows: The Determinants of Transactions in Financial Assets,” European Economic Review Vol. 45, pp. 783–96.
- Sauer, Stephan, and Jan-Egbert Sturm, 2003, “Using Taylor Rules to Understand ECB Monetary Policy,” CESifo Working Paper No. 1110 (Munich: Center for Economic Studies).
- Strauss-Kahn, Vanessa, and Xavier Vives, 2005, “Why and Where Do Headquarters Move?,” CEPR Discussion Paper No. 5070, May (London: Center for Economic Policy Research).
- Thrift, Nigel, 1994, “On the Social and Cultural Determinants of International Financial Centres: The Case of the City of London,” in Money, Power and Space, ed. by S. Corbridge, N. Thrift, and R. Martin, (Oxford: Blackwell), pp. 327-355.
- Tschoegl, Adrian, 2000, “International Banking Centers, Geography, and Foreign Banks,” Financial Markets, Institutions and Instruments, Vol. 9(1), pp. 1-32.

### Key tabular findings and descriptive statistics
- Table 1 (a): Absolute Error for expected rate by Country (in %)
  - All countries: # obs 4,242; mean 0.060; pre-post-change diff. 0.077; st. dev. 0.005.
  - Euro area: # obs 2,350; mean 0.057; pre-post-change diff. 0.078; st. dev. 0.002.
  - Non-euro area: # obs 1,892; mean 0.062; pre-post-change diff. 0.076; st. dev. 0.009.
  - Country examples (mean): Austria 0.087; France 0.070; United Kingdom 0.062; Germany 0.053; USA 0.063.
  - p-values for unbiasedness and efficiency tests reported per country; many entries show p-value 0.000 for panel aggregates.
- Table 1 (b): Absolute Error for most likely rate by Country (in %)
  - All countries: # obs 4,314; mean 0.035; pre-post-change diff. 0.092; st. dev. 0.005.
  - Euro area: # obs 2,421; mean 0.034; pre-post-change diff. 0.091; st. dev. 0.002.
  - Country examples (mean): Austria 0.067; France 0.043; United Kingdom 0.036; Germany 0.030; USA 0.051.
- Table 1 (c): Absolute Error for meetings-to-change by Country (in number of meetings)
  - All countries: # obs 3,199; mean 5.40; pre-post-change diff. 5.773; st. dev. 6.51.
  - Euro area: # obs 1,754; mean 5.31; pre-post-change diff. 5.621; st. dev. 6.48.
  - Country examples (mean): Austria 5.72; United Kingdom 5.59; Germany 5.18; Portugal 6.17; Sweden 6.44.
- Table 3: Absolute Systematic Error for expected rate by Country (in %)
  - All countries: # obs 4,242; mean 0.073; pre-post-change diff. 0.086; st. dev. 0.026.
  - Euro area: # obs 2,350; mean 0.073; pre-post-change diff. 0.087; st. dev. 0.025.
  - Country examples (mean): France 0.085; United Kingdom 0.074; Germany 0.069; USA 0.060.

Notes attached to tables:
- “-- indicates missing values; "diff." shows significance level of a t-test of whether the country-specific mean is different from the mean for all countries.”
- “***, **, and * indicate significance at the 99%, 95%, and 90% levels, respectively.”
- “The unbiasedness test and efficiency test are as defined in the text; p-values of above 0.90 indicate that unbiasedness or efficiency are rejected at the 90% level.”

### Regression and econometric results (selected coefficients and sample sizes)
- Table 2: Explaining the Forecasting Error: Geography, Macro Conditions, and History
  - # observations: 4242 (full period), 1414 (pre-2001), 2828 (post-2001).
  - Distance variables (selected coefficients, coef. and std. err.):
    - Headquarters in Frankfurt: -0.011 *** (0.003) [full period]; -0.017 *** (0.005) [pre-2001]; -0.007 * (0.004) [post-2001].
    - Subsidiary in Frankfurt: -0.015 *** (0.003) [full period]; -0.017 *** (0.005) [pre-2001]; -0.015 *** (0.004) [post-2001].
    - HQ or Subsidiary in London: -0.002 (0.003) [full period]; -0.007 (0.005) [pre-2001]; 0.001 (0.003) [post-2001].
    - Vicinity of Frankfurt: -0.001 (0.003) [full period]; -0.012 ** (0.005) [pre-2001]; 0.003 (0.004) [post-2001].
  - Macro conditions (selected coefficients):
    - Relative inflation: 0.005 *** (0.002) [full period]; 0.008 *** (0.002) [pre-2001]; 0.006 *** (0.002) [post-2001]; 0.010 *** (0.002) in one column.
    - Relative unemployment: 0.001 ** (0.001) [full period]; 0.002 *** (0.001) [pre-2001]; 0.001 ** (0.001) [post-2001]; 0.003 *** (0.001) in one column.
  - History:
    - Central bank independence: -0.002 (0.001) in full period; -0.004 ** (0.002) in one subcolumn.
  - Notes: ***, **, and * indicate significance at the 99%, 95%, and 90% levels, respectively.
- Table 4: Explaining the Taylor-Type Rule Systematic Error: Geography, Macro Conditions, and History across Subperiods
  - # observations: 4242 (full period), 1414 (pre-2001), 2828 (post-2001).
  - Distance variables (selected coefficients):
    - Headquarters in Frankfurt: -0.008 *** (0.002) [full period]; -0.011 *** (0.004) [pre-2001]; -0.006 * (0.003) [post-2001].
    - Subsidiary in Frankfurt: -0.010 *** (0.002) [full period]; -0.011 ** (0.004) [pre-2001]; -0.010 *** (0.003) [post-2001].
    - Vicinity of Frankfurt: 0.000 (0.002) [full period]; -0.009 ** (0.004) [pre-2001]; 0.004 (0.003) [post-2001].
  - Macro conditions (selected coefficients):
    - Relative inflation: 0.005 *** (0.001) [full period]; 0.855 *** (0.001) appears in a pre-2001 column; 0.006 *** (0.001) [pre-2001/post-2001 mix]; 0.011 *** (0.002) in one column.
    - Relative unemployment: 0.001 *** (0.000) [full period]; 0.247 *** (0.001) in a pre-2001 column; 0.002 *** (0.000) in another column; 0.003 *** (0.001) in one column.
  - History:
    - Central bank independence: -0.002 * (0.001) [full period]; -0.004 *** (0.001) [pre-2001].
  - Notes: ***, **, and * indicate significance at the 99%, 95%, and 90% levels, respectively.

### Taylor-type rule estimation (Table 5) — selected coefficients and significance
- Panel (A) All groups, pre-2001 vs. post-2001 (selected coefficients shown with significance):
  - HICP inflation: pre-2001 coef. 3.651 *** (std. err. 1.112); post-2001 coef. 6.184 *** (0.562).
  - M3: pre-2001 coef. 2.556 *** (0.527); post-2001 coef. 3.467 *** (0.287).
  - Industrial production: pre-2001 coef. 1.761 *** (0.172); post-2001 coef. 0.606 *** (0.094).
  - Consumer confidence: pre-2001 coef. 0.001 (0.001); post-2001 coef. 0.032 *** (0.001).
  - Previous rate: pre-2001 coef. 0.948 *** (0.008); post-2001 coef. 0.827 *** (0.004).
- Panel (B) By group: multiple subgroup coefficient estimates reported with tests for equality:
  - Notes: ***, **, * indicate significance at the 99%, 95%, and 90% levels, respectively.
  - A Y indicates rejection at the 90% level of a t-test of equality of the pre-2001 and post-2001 coefficients.
  - B Y indicates rejection at the 90% level of a t-test of equality of the coefficients of the two groups pre-2001.
  - C Y indicates rejection at the 90% level of a t-test of equality of the coefficients of the two groups post-2001.

### Figures and graphical notes
- Figure 1: Monetary Policy Forecasting Error by Decile for expected rate
  - Shows average absolute forecast error in basis points by financial institutions across deciles 1–10 for 1999-2005, pre-2001, and post-2001.
  - Note: “The figure shows the average absolute forecast error in basis points by financial institutions, ranging from those 10% of institutions with the lowest forecast errors in decile 1 to those 10% with the highest error in decile 10. The fact that the average errors for the full sample period can be (somewhat) larger or smaller than for both subperiods is owed to the fact that the allocation of institutions to deciles changes slightly over time.”
- Figure 2: Actual Rate and expected rate Based on the Overall Sample of Forecasters
  - Date labels include: 01may1998, 13sep1999, 25jan2001, 09jun2002, 22oct2003, 05mar2005.
  - Note: “The figure shows the mean as well as the minimum and maximum of the policy rate forecasts polled by Reuters before a particular ECB Council meeting. In addition, the actual policy rate as decided in the respective meeting is depicted.”
  - Source listed in figure: Reuters; authors’ calculations.

*Italic: Content extracted from the provided PDF chapter/section titled "_wp0641 - References".*

### Appendix 1: Results for

### Appendix 1: Results for most likely rate

### A1.1 — Explaining the Forecasting Error: Geography, Macro Conditions, and History (ordered probit; absolute forecast errors take discrete values of 0 bp, 25 bp, or 50 bp)
- Distance (coefficients and std. err. shown for full period, pre-2001, post-2001 columns):
  - Headquarters in Frankfurt: coef. -0.252; std. err. 0.141 (pre-2001/post-2001/full period also reported: -0.143/0.256; -0.298/0.172; -0.102/0.154; -0.073/0.271; -0.055/0.191)
  - Subsidiary in Frankfurt: coef. -0.264; std. err. 0.151 (additional column values: -0.345/0.280; -0.213/0.182; -0.156/0.158; -0.294/0.287; -0.039/0.192)
  - HQ or Subsidiary in London: coef. -0.130; std. err. 0.114 (additional column values: 0.088/0.214; -0.227/0.138; -0.100/0.115; 0.092/0.214; -0.171/0.140)
  - Vicinity of Frankfurt: coef. 0.045; std. err. 0.122 (additional column values: -0.192/0.234; 0.081/0.148; 0.089/0.124; -0.198/0.234; 0.174/0.153)
- Macro conditions:
  - Relative inflation: coef. 0.197; std. err. 0.066 (additional: 0.060/0.129; 0.287/0.081; 0.231/0.067; 0.081/0.131; 0.344/0.083) — significance: ***
  - Relative unemployment: coef. 0.048; std. err. 0.020 (additional: 0.021/0.033; 0.094/0.028; 0.057/0.021; 0.020/0.033; 0.112/0.029) — significance: **
- History:
  - Central bank independence: coef. -0.135; std. err. 0.054 (additional: -0.081/0.099; -0.205/0.066) — significance: **
- Sample sizes:
  - # observations: 4314 (full period), 1507 (pre-2001), 2807 (post-2001)
- Notes:
  - ***, **, and * indicate significance at the 99%, 95%, and 90% levels, respectively.

### A1.2 — Absolute Systematic Error of Mean Expectations in Taylor-Type Rule (in %) — country-level means, 1999–2005
- Table structure: for each country, # obs, mean, diff. (significance of difference vs. all countries mean), st. dev., annual values 1999–2005, and 1999-2005 aggregated columns for pre- and post-Feb. 2001 meetings.
- Selected country entries (preserve reported values exactly):
  - Austria: # obs 105; mean 0.095; diff. ***; st. dev. 0.103; annual values include 0.012, 0.089, 0.091, 0.143, 0.148, 0.053, 0.019; 1999-2005 pre/post: 0.240/0.082/0.065/0.105
  - Belgium: # obs 8; mean 0.133; diff. *; st. dev. 0.179; limited annual entries (many --)
  - Denmark: # obs 149; mean 0.064; st. dev. 0.092; annual values include 0.018, 0.087, 0.084, 0.089, 0.085, 0.010; 1999-2005 columns: 0.000/0.269/0.034/0.059/0.069
  - United Kingdom: # obs 1,442; mean 0.069; st. dev. 0.090; annual values include 0.026, 0.105, 0.075, 0.069, 0.100, 0.023; 1999-2005 columns: 0.025/0.250/0.043/0.076/0.066
  - Germany: # obs 1,299; mean 0.064; diff. **; st. dev. 0.092; annual values include 0.020, 0.087, 0.064, 0.089, 0.091, 0.021; 1999-2005 columns: 0.023/0.274/0.035/0.059/0.066
  - Euro area (aggregate): # obs 2,421; mean 0.068; st. dev. 0.092; 1999-2005 columns: 0.021/0.094/0.067/0.090/0.096; 1999-2005 pre/post: 0.024/0.025/0.267/0.042/0.068/0.068
  - Non-euro area (aggregate): # obs 1,893; mean 0.068; st. dev. 0.089; 1999-2005 columns: 0.027/0.101/0.074/0.072/0.095; 1999-2005 pre/post: 0.025/0.023/0.249/0.043/0.073/0.066
  - All countries: # obs 4,314; mean 0.068; st. dev. 0.091; annual values 1999–2005: 0.024, 0.097, 0.070, 0.082, 0.095; 1999-2005 pre/post columns: 0.024/0.024/0.259/0.042/0.070/0.067
- Notes:
  - -- indicates missing values; "diff." shows significance level of a t-test of whether the country-specific mean is different from the mean for all countries. ***, **, and * indicate significance at the 99%, 95%, and 90% levels, respectively.

### A1.3 — Explaining the Taylor-Type Rule Systematic Error: Geography, Macro Conditions, and History (linear regressions; coefficients in %)
- Distance:
  - Headquarters in Frankfurt: coef. -0.013; std. err. 0.003 (additional columns: -0.012/0.005; -0.013/0.004; -0.010/0.003; -0.013/0.005; -0.005/0.004) — significance: ***
  - Subsidiary in Frankfurt: coef. -0.011; std. err. 0.003 (additional columns: -0.013/0.005; -0.010/0.004; -0.009/0.003; -0.014/0.005; -0.004/0.004) — significance: ***
  - HQ or Subsidiary in London: coef. -0.005; std. err. 0.002 (additional columns: -0.006/0.005; -0.004/0.003; -0.005/0.002; -0.006/0.005; -0.002/0.003) — significance: **
  - Vicinity of Frankfurt: coef. -0.002; std. err. 0.003 (but some columns show -0.013/0.005 significant at ***, and 0.008/0.003 significant at **)
- Macro conditions:
  - Relative inflation: coef. 0.007; std. err. 0.001 (additional: 0.000/0.003; 0.009/0.002; 0.008/0.002; 0.000/0.003; 0.012/0.002) — significance: ***
  - Relative unemployment: coef. 0.001; std. err. 0.000 (additional: 0.000/0.001; 0.001/0.001; 0.001/0.000; 0.000/0.001; 0.002/0.001) — significance: * / ** / ***
- History:
  - Central bank independence: coef. -0.003; std. err. 0.001 (some columns report 0.000/0.002 and -0.007/0.001) — significance: ***
- Sample sizes:
  - # observations: 4314 (full period), 1507 (pre-2001), 2807 (post-2001)
- Notes:
  - ***, **, and * indicate significance at the 99%, 95%, and 90% levels, respectively.

### A1.4 — Taylor-Type Rule Estimates for most likely rate (selected coefficients; pre-2001 vs. post-2001)
- Panel (A) All groups — pre-2001 vs. post-2001 (coef. and std. err. reported):
  - HICP inflation: pre-2001 coef. 2.608; std. err. 1.414; post-2001 coef. 4.082; std. err. 0.765 — significance pre: *; post: ***
  - M3: pre-2001 coef. 2.466; std. err. 0.685; post-2001 coef. 2.839; std. err. 0.391 — significance: ***
  - Industrial production: pre-2001 coef. 1.266; std. err. 0.221; post-2001 coef. 0.740; std. err. 0.127 — significance: ***
  - Consumer confidence: pre-2001 coef. 0.002; std. err. 0.001; post-2001 coef. 0.028; std. err. 0.001 — significance: * (pre), *** (post)
  - Previous rate: pre-2001 coef. 0.956; std. err. 0.011; post-2001 coef. 0.852; std. err. 0.006 — significance: ***
- Panel (B) By group — coefficients displayed for multiple group comparisons (Vicinity, Macro conditions, History, comparison groups such as outside Frankfurt vs. HQ or subsidiary in Frankfurt; low CB independence vs. high; high inflation vs. low, etc.). Selected reported entries (preserve values exactly where shown):
  - Vicinity group HICP inflation: 2.216 (pre-2001), 3.992 *** (post-2001), 3.617, 4.689 ***
  - M3 in Vicinity: 2.843 *** (pre), 3.038 *** (post), 1.858, 2.242 ***
  - Previous rate in Vicinity: 0.960 *** (pre), 0.843 *** (post), 0.949 ***, 0.872 *** (post)
  - Macro conditions group and History group report analogous sets of coefficients for HICP inflation, M3, Industrial production, Consumer confidence, Previous rate with many entries significant at ***, **, or *.
- Notes:
  - ***, **, * indicate significance at the 99%, 95%, and 90% levels, respectively.
  - A Y indicates rejection at the 90% level of a t-test of equality of the pre-2001 and post-2001 coefficients.
  - B Y indicates rejection at the 90% level of a t-test of equality of the coefficients of the two groups pre-2001.
  - C Y indicates rejection at the 90% level of a t-test of equality of the coefficients of the two groups post-2001.

*Source: Appendix tables and notes from the supplied IMF content unit.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2006/_wp0641.pdf_