## wp1739

## Source details

**Canonical URL:** [wp1739](https://www.imf.org/-/media/files/publications/wp/2017/wp1739.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp1739.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp1739.pdf.json)

---

### Data and sample
- Sample covers 17 advanced economies during 1978-2015; panel contains 229 years of fiscal consolidation episodes.
- Identified 73 episodes of fiscal consolidations in the sample.
- Size of reported fiscal adjustment episodes: from 0.04 percent of GDP to 4.74 percent of GDP.
- Average size of fiscal adjustment: 1.06 percent of GDP.
- Average duration of fiscal episodes: 3.3 years.
  - Shortest duration: 1 year (21 episodes).
  - Longest duration: 14 years (Canada).
- Temporal concentration: most episodes in the mid-1980s, late 1990s, and 2010-2015.
- Distribution: more than 35 percent of episodes have size between 0 and 0.5 percent of GDP; 5 percent of cases larger than 3 percent of GDP.
- Data level: general government.
- Data construction combined Devries et al. (2011), Alesina et al. (2015) expansion, and European Commission’s Stability and Convergence Programmes (first vintage) complemented with country budget sources to expand coverage to 2015.

### Measurement: Consolidation Promise Gap (CPG)
- Definition: CPG = PFA – RFA
  - PFA: size of planned fiscal adjustment (narrative approach).
  - RFA: size of realized fiscal adjustment (changes in the budget balance).
  - All expressed in percent of GDP.
- Sign interpretation:
  - Positive CPG: government delivered less fiscal adjustment than planned (underperformance).
  - Negative CPG: government implemented a larger adjustment than planned (overperformance).
- Alternative measures used: primary balance and structural balance variants.
- Empirical note: Average promise gap remains 0.3 percent of GDP under measures using total balance, primary balance, and structural balance.

### Key quantitative findings on promise gaps
- Fiscal promise gaps were sizeable: about 0.3 percent of GDP per year, or 1.1 percent of GDP over a typical 3-year adjustment episode.
- Average promise gap by sign:
  - Underperforming episodes (positive CPG): average promise gap = 2.8 percent of GDP.
  - Over-performing episodes (negative CPG): average promise gap = 1.2 percent of GDP.
- Country-level average promise gaps:
  - Largest positive promise gaps: Finland, Spain, Ireland, Italy.
  - Countries with negative average CPG (over-delivery): Denmark, United Kingdom, Germany, Canada, Sweden.

### Economic determinants (stylized facts and regression evidence)
- Stylized associations:
  - Higher initial debt and worse fiscal sustainability → smaller consolidation promise gaps.
  - Adverse economic conditions (large output gaps, high real GDP forecast errors) → smaller promise gaps.
- Panel regression (select exact estimates):
  - Lagged Debt: -0.026***; -0.028*** (standard errors (0.005), (0.007))
  - Change in output gap: -0.540***; -0.605*** (standard errors (0.070), (0.092))
  - Lagged GDP forecast error: -0.200; -0.202* (standard errors (0.133), (0.107))
  - Observations across specifications: 228, 231, 162, 162.
  - R-squared across specifications: 0.175, 0.287, 0.146, 0.467.
- Interpretation: Larger initial debt levels, improvements in the output gap, and larger forecast errors are associated with smaller consolidation promise gaps.

### Political determinants (PCA construction and regression results)
- Three political dimensions constructed via PCA: Electoral Proximity, Political Strength, Institutional Accountability.
  - % explained by factors: 0.39 (Proximity), 0.69 (Strength), 0.85 (Accountability).
- Selected factor loadings and uniqueness:
  - years in office: proximity loading 0.39; uniqueness 0.29
  - months to next election: proximity loading 0.45; uniqueness 0.28
  - margin of majority: strength loading 0.93; uniqueness 0.12
  - cabinet strength: strength loading 0.90; uniqueness 0.17
  - voice and accountability: accountability loading 0.87; uniqueness 0.22
  - control of corruption: accountability loading 0.96; uniqueness 0.07
- Political stylized facts:
  - Higher political strength and higher institutional accountability → lower consolidation promise gaps.
  - Higher electoral proximity → lower consolidation promise gaps (incumbent pressure to deliver).
  - Differences are more pronounced for Strength and Accountability than for Proximity.
- Regression estimates including PCA indices (select exact coefficients):
  - Lagged Debt: -0.020***; -0.021***; -0.063**; -0.057* (standard errors (0.007), (0.007), (0.029), (0.032))
  - Change in output gap: -0.557***; -0.545***; -0.497*; -0.536* (standard errors (0.085), (0.084), (0.292), (0.312))
  - Proximity: -0.072 (0.138) and 0.035 (0.356) (not statistically significant)
  - Strength: -0.531* (0.284) and -0.570 (1.079)
  - Accountability: -2.726* (1.561) and -3.202* (1.811)
  - Observations: 162, 162, 41, 41
  - R-squared: 0.404, 0.417, 0.839, 0.841
- Selected individual political variables (exact coefficients):
  - Recent elections: -0.388*** (0.147); -0.302 (0.307)
  - Margin of majority: -4.373** (2.007); -7.480** (3.546)
  - Government effectiveness: -3.899*** (1.310); -4.104** (1.942)
  - Lagged GDP forecast error: -0.173; -0.181*; -0.062; 0.026 (standard errors (0.113), (0.107), (0.162), (0.190))
  - Observations: 119, 162, 94, 65
  - R-squared: 0.375, 0.484, 0.534, 0.501
- Robustness: re-running analysis on the original Devries et al. (2011) database and on homogeneous SCP sample yields qualitatively similar results.

### Consequences: financial markets and voters
- Markets:
  - Delivering on promised consolidations rewarded by financial markets; underperformers punished.
  - 5-year government bond spreads (relative to Germany) increase at the start and immediately after consolidations; spike larger and more persistent for underperforming adjustments.
  - Static OLS estimates (underperformers, selected coefficients): CPG: 0.283*** (0.102); 0.247* (0.140). Observations: 55. R-squared: 0.128, 0.130, 0.166 in columns reported.
  - Overperformers: CPG: 0.260** (0.102) in one column; other columns show smaller or insignificant coefficients. Observations: 30.
- Voters:
  - Government popularity generally unaffected by sign of promise gaps; electorate response weaker than markets.
  - Static OLS estimates (underperformers, selected coefficients): CPG: -1.775** (0.829) in one specification; -0.866 (1.896) in another. Observations: 66 and 36. R-squared: 0.603 and 0.681.
  - Overperformers: CPG: 0.438 (1.465) and -1.029 (2.054) in reported columns. Observations: 57 and 35. R-squared: 0.748 and 0.792.
- Interpretation: markets react decisively to promise gaps (particularly underperformance and during recessions); voters show muted and statistically weaker reactions, with only marginal significance in some settings.

### Dynamics and identification: local projections and regime dependence
- Primary dynamic identification: local projection (LP) method following Jordà (2005).
  - Regression for each horizon k (k = 1,...,5) and controls identical to cross-sectional regressions.
  - Estimation uses Beck and Katz’s (1995) panel-corrected standard error (PCSE) estimator.
  - Lag length l = 2 selected by AIC; results robust to different lag structures.
- Rationale: LP preferred over PVARs and ARDL for transparent control specification, fewer identification assumptions, and reduced sensitivity to ordering and small-sample biases.
- Small-sample bias reported: 0.03 (1/T where T is 38).
- Regime-dependent IRFs estimated with LP conditioning on business cycle (real GDP growth normalized to zero mean and unit variance), equivalent to smooth transition autoregressive approach.
- LP impulse-response results (stylized):
  - After a CPG shock, 5-year government spreads increase strongly and significantly.
  - Government popularity declines gradually but response is statistically weaker.
  - Markets penalize underperformance more and for longer during recessions; voter reaction remains muted, with marginally significant decline around one year after shock in bad times.
  - Figures reported confidence bands equal 90 percent error bands.

### Policy implications and conclusions
- Core conclusions:
  - Fiscal consolidation promise gaps are sizeable and driven by both economic and political factors.
  - Smaller initial debt levels, improvements in the output gap, and smaller forecast errors are associated with narrower promise gaps.
  - Political conditions matter: recently elected governments, large parliamentary majorities, and high government effectiveness increase likelihood of delivering planned fiscal adjustments.
  - Financial markets punish underperformers and reward over-performers; voters are less responsive.
- Policy recommendations:
  - Policymakers should be cautious in formulating ambitious fiscal consolidations (especially when revenue driven).
  - Governments can improve delivery of promised spending cuts by increasing capacity to deliver public services efficiently.
  - Stronger starting conditions (lower fiscal imbalances) reduce subsequent promise gaps.
  - Policymakers should not deviate from initial adjustment plans hoping to offset market backlash with electoral gains because voters do not reward broken promises.

### Robustness checks and appendices (selected exact statistics)
- Robustness on Devries’ database (1978-2009), select results:
  - Lagged debt coefficient: -0.032*** (0.007); Observations: 162; R-squared: 0.225.
  - Change in output gap: -0.611*** (0.082); Observations: 165; R-squared: 0.368.
  - Lagged GDP forecast error: -0.404** (0.166); Observations: 96; R-squared: 0.428.
- Robustness on SCP database (1998-2015), select results:
  - Lagged debt: -0.026*** (0.009); Observations: 252; R-squared: 0.148.
  - Change in output gap: -0.242*** (0.052); Observations: 253; R-squared: 0.196.
  - Recent elections: -1.071* (0.586) and -1.289** (0.632) in specified models.
- Appendix 1 provides selected country-year consolidation sizes and composition (percent of GDP) for multiple countries and years (values preserved in source tables).

_Source: wp1739_

### References  _______________________________________________________________________________________  34

### wp1739 - References

### Data and Identification of Fiscal Consolidation Episodes
- Sample covers 17 advanced economies during 1978-2015.
- The authors create a new database of fiscal consolidations comparing narrative budget plans with actual fiscal performance in 17 OECD countries during 1978-2015.
- Identified 73 episodes of fiscal consolidations in the sample.
- The number of fiscal contractions per country ranges from two (Canada, Finland) to seven (France, United States).
- Identification follows a narrative approach relying on historical descriptions recorded in national budget laws, European Commission’s Stability and Growth Programs and OECD’s country reviews, updating Devries and others (2011) and Alesina and others (2015) for the 17-country sample.
- Discussion of alternative identification: contrast with the positive approach that associates consolidations with large changes in the cyclically adjusted primary balance (CAPB); cautions that CAPB may be influenced by many non-policy factors (example: stock price booms raising CAPB) and could bias empirical estimates.

### Key Findings on Promise Gaps
- Measured "promise gap" as the difference between budget plans and budget implementation.
- Fiscal promise gaps were sizeable: about 0.3 percent of GDP per year, or 1.1 percent of GDP over a typical 3-year adjustment episode.
- Economic factors and forecast errors were important determinants of differences between budget plans and fiscal outcomes.
- Political factors also played a role:
  - Greater electoral proximity is associated with smaller promise gaps.
  - Stronger political cohesion is associated with smaller promise gaps.
  - Higher accountability is associated with smaller promise gaps.

### Political Economy Interpretation
- Parties in government are motivated by both policy agendas and the willingness to remain in office; delivering promised policies increases chances of re-election.
- Political conditions such as proximity to elections and political fragmentation can influence fiscal outcomes.
- Literature notes ideology can affect composition of adjustments (revenue vs. expenditure), though ideology is not relevant for explaining the size of consolidation promise gaps in this sample.

### Market and Voter Reactions (Consequences)
- Governments that delivered on their promised fiscal consolidation plans were rewarded by financial markets.
- Delivering on consolidation promises was not associated with electoral punishment; governments delivering as promised were not penalized by voters.

### Contributions of the Paper
- Updates the narrative database from Devries and others (2011) and Alesina and others (2015) for the 17-country sample.
- Simultaneously considers the role of three political factors—electoral proximity, political strength, and institutional accountability—in explaining consolidation promise gaps.
- Examines consequences of promise gaps on market sentiment and governments’ popularity among the electorate.

### Structure and Further Analysis (paper organization)
- Section 2: data and definition of dependent variable.
- Section 3: economic and political causes of consolidation promise gaps.
- Section 4: reaction of markets and the electorate to promise gaps.
- Final section: conclusions and policy implications.
- Appendices listed include: Description of the Sample Selection; Robustness of Results in a Homogeneous Sample; The Consequences of Consolidation Promise Gaps: Regression Analysis.

*Source: wp1739 - References*

### 0.04 percent of GDP to 4.74 percent of GDP, with an average size of fiscal adjustment equal to 1.06

### wp1739 - 0.04 percent of GDP to 4.74 percent of GDP, with an average size of fiscal adjustment equal to 1.06 percent of GDP

### Sample of Fiscal Consolidation Episodes
- Coverage: 17 advanced economies between 1978-2015; 229 years of fiscal consolidation episodes in the panel database.
- Size of reported fiscal adjustment episodes: from 0.04 percent of GDP to 4.74 percent of GDP.
- Average size of fiscal adjustment: 1.06 percent of GDP.
- Average duration of fiscal episodes: 3.3 years.
  - Shortest duration: 1 year (21 episodes).
  - Longest duration: 14 years (Canada).
- Distribution over time:
  - Most episodes: mid-1980s, late 1990s, and 2010-2015.
  - More than 35 percent of episodes: size between 0 and 0.5 percent of GDP.
  - Only 5 percent of cases: adjustment larger than 3 percent of GDP.
- Table of country-specific consolidation sample years (selected examples preserved exactly as in source):
  - Australia: 1985-88, 1994-99, 2010-12, 2014-15
  - Austria: 1980-81, 1984, 1996-97, 2001-02, 2011-12, 2015
  - Belgium: 1982-85, 1987, 1990, 1992-94, 1996-97, 2010-15
  - Canada: 1984-97, 2010-15
  - Denmark: 1983-86, 1995, 2012
  - Finland: 1992-97, 2011
  - France: 1979, 1987, 1989, 1991-92, 1995-97, 1999-2000, 2011-15
  - Germany: 1982-1984, 1991-95, 1997-2000, 2003-04, 2006-07, 2011-12
  - Ireland: 1982-88, 2009-15
  - Italy: 1991-98, 2004-07, 2010-15
  - Japan: 1997-98, 2003-07
  - Netherlands: 1981-88, 1991-93, 2004-05, 2011-13, 2015
  - Portugal: 1983, 2000, 2002-03, 2005-07, 2010-15
  - Spain: 1983-84, 1989-90, 1992-97, 2009-15
  - Sweden: 1984, 1993-98, 2011, 2015
  - United Kingdom: 1979-82, 1994-99, 2010, 2012, 2014-15
  - United States: 1978, 1980-81, 1985-86, 1988, 1990-98, 2011, 2013-15
- Note: “Source: Author’s calculations. See Appendix 1 for details.”

### Measuring Consolidation Promise Gaps (CPG)
- Definition: CPG = PFA – RFA
  - PFA: size of planned fiscal adjustment (narrative approach).
  - RFA: size of realized fiscal adjustment (changes in the budget balance).
  - All expressed in percent of GDP.
- Sign interpretation:
  - Positive CPG: government delivered less fiscal adjustment than planned.
  - Negative CPG: government implemented a larger adjustment than planned.
- Alternative measures:
  - Using primary balance to correct for interest payments.
  - Using structural balance to account for economic cycle effects and one-off measures.
- Empirical note: Average promise gap remains 0.3 percent of GDP under measures using total balance, primary balance, and structural balance.
- Country-level findings (average promise gaps):
  - Largest positive promise gaps: Finland, Spain, Ireland, Italy.
  - Countries delivering larger-than-planned consolidations on average (negative CPG): Denmark, United Kingdom, Germany, Canada, Sweden.
- Data level: general government.

### Economic Determinants: Stylized Facts and Regression Results
- Variables considered:
  - Initial fiscal conditions: lagged public debt-to-GDP ratio (sustainability factor not included due to collinearity with debt).
  - Economic conditions: output gap (actual minus potential GDP), GDP forecast errors (3, 9, 15, 21 months; presented using 9 months ahead variant).
- Stylized facts (Figure 5 and text):
  - High initial debt and low fiscal sustainability are associated with smaller consolidation promise gaps.
  - Adverse economic conditions (large output gaps, high real GDP forecast errors) are associated with smaller promise gaps.
- Panel regression specification (Equation (1)):
  - Dependent variable: consolidation promise gap.
  - Regressors include lagged debt, change in output gap, lagged GDP forecast error, political variables.
  - Estimation: OLS with robust standard errors clustered at the country level.
- Key regression estimates (select exact coefficients and significance from tables):
  - Table 4 (Economic Determinants):
    - Lagged Debt: -0.026***; -0.028*** (standard errors in parentheses: (0.005), (0.007))
    - Change in output gap: -0.540***; -0.605*** (standard errors (0.070), (0.092))
    - Lagged GDP forecast error: -0.200; -0.202* (standard errors (0.133), (0.107))
    - Observations: 228, 231, 162, 162 (across specifications).
    - R-squared: 0.175, 0.287, 0.146, 0.467 (across specifications).
  - Interpretation: Larger initial debt levels, improvements in the output gap, and larger forecast errors lead to smaller consolidation promise gaps.

### Political Determinants: PCA Construction and Effects
- Political dimensions constructed via Principal Components Analysis (PCA):
  - Electoral Proximity:
    - Components: years in office, years left in current term, party of chief executive more time in office, months to next election.
    - Interpretation: higher proximity → greater electoral pressure to deliver → lower promise gaps.
    - Only the first principal component retained; first factor explains almost 40 percent of variance.
  - Political Strength:
    - Components: margin of parliamentary majority, low cabinet fragmentation (cabinet strength), executive control of all houses, weak opposition.
    - Interpretation: more political strength → tighter fiscal discipline → lower promise gaps.
    - Only the first principal component retained.
  - Political Accountability:
    - Components: voice and accountability, regulatory quality, government effectiveness, control of corruption, rule of law.
    - Interpretation: higher accountability → more fiscal discipline → lower promise gaps.
    - Only the first principal component retained.
- PCA descriptive statistics (Table 2 summary format preserved qualitatively in source).
- Factor loadings and uniqueness (Table 3 exact loadings preserved):
  - years in office: proximity factor loading 0.39; uniqueness 0.29
  - years left in current term: 0.41; uniqueness 0.28
  - party of chief executive more time in office: 0.37; uniqueness 0.29
  - months to next election: 0.45; uniqueness 0.28
  - margin of majority: strength loading 0.93; uniqueness 0.12
  - cabinet strength: 0.90; uniqueness 0.17
  - executive control of all houses: 0.76; uniqueness 0.42
  - weak opposition: 0.72; uniqueness 0.47
  - voice and accountability: accountability loading 0.87; uniqueness 0.22
  - regulatory quality: 0.89; uniqueness 0.10
  - government effectiveness: 0.82; uniqueness 0.13
  - control of corruption: 0.96; uniqueness 0.07
  - rule of law: 0.94; uniqueness 0.11
  - % explained by factors: 0.39 (proximity), 0.69 (strength), 0.85 (accountability)
- Political stylized facts (Figure 6 and text):
  - Higher political strength and higher institutional accountability are associated with lower consolidation promise gaps.
  - Higher electoral proximity is associated with lower consolidation promise gaps (incumbent pressure to deliver).
  - Differences between high and low values are more pronounced for Strength and Accountability than for Proximity.
- Regression results including PCA indices (Table 5 exact coefficients):
  - Lagged Debt: -0.020***; -0.021***; -0.063**; -0.057* (standard errors (0.007), (0.007), (0.029), (0.032))
  - Change in output gap: -0.557***; -0.545***; -0.497*; -0.536* (standard errors (0.085), (0.084), (0.292), (0.312))
  - Proximity: -0.072 (0.138) and 0.035 (0.356) in alternative specifications (not statistically significant)
  - Strength: -0.531* (0.284) and -0.570 (1.079)
  - Accountability: -2.726* (1.561) and -3.202* (1.811)
  - Observations: 162, 162, 41, 41
  - R-squared: 0.404, 0.417, 0.839, 0.841
- Selected individual political variables (Table 6 exact coefficients):
  - Lagged Debt: -0.035***; -0.026***; -0.033***; -0.014 (standard errors (0.009), (0.007), (0.011), (0.014))
  - Change in output gap: -0.304**; -0.575***; -0.493***; -0.026 (standard errors (0.123), (0.092), (0.139), (0.200))
  - Recent elections: -0.388*** (0.147); -0.302 (0.307)
  - Margin of majority: -4.373** (2.007); -7.480** (3.546)
  - Government effectiveness: -3.899*** (1.310); -4.104** (1.942)
  - Lagged GDP forecast error: -0.173; -0.181*; -0.062; 0.026 (standard errors (0.113), (0.107), (0.162), (0.190))
  - Observations: 119, 162, 94, 65
  - R-squared: 0.375, 0.484, 0.534, 0.501
- Robustness checks:
  - Re-running analysis on original Devries et al. (2011) database yields qualitatively similar results.
  - Appendix 2: estimation constrained to 2009 shows qualitatively similar results with generally more statistical significance.

### Consequences of Consolidation Promise Gaps
- Average promise gap sizes by sign:
  - Underperforming episodes (positive CPG): average promise gap = 2.8 percent of GDP.
  - Over-performing episodes (negative CPG): average promise gap = 1.2 percent of GDP.
- Voters’ reactions (Figure 7 stylized facts):
  - Government popularity is generally unaffected by the sign of promise gaps.
  - Slight decline in popularity in aftermath of underperforming episodes (evidence not strongly significant).
  - Interpretation: voters do not appear to reward governments for breaking consolidation promises; they may dislike broken promises.
- Market reactions (Figure 7 stylized facts):
  - 5-year government bond spreads (against Germany) increase at the start and immediately after consolidations in both over- and under-performance subsamples.
  - The spike in bond spreads is larger and more long-lasting for underperforming fiscal adjustments.
  - Interpretation: markets respond more decisively than voters to promise gaps; underperformance leads to higher bond spreads.

### Dynamic Impact Estimation: Impulse-Response Functions (IRFs)
- Methodology:
  - Local projections following Jorda (2005) used to generate IRFs.
  - Regression for each horizon k (k = 1,...,5) as specified in Equation (2).
  - Dependent Y: either 5-year government bond spreads (relative to Germany) or government popularity.
  - Controls: same vector of control variables as in regressions; lag length l = 2 (selected by AIC); results robust to different lag structures.
  - Estimation: Beck and Katz’s (1995) panel-corrected standard error (PCSE) estimator.
  - IRFs obtained by collecting estimated beta coefficients with confidence intervals from standard errors.
- Rationale for method choice:
  - PVARs considered inferior due to endogeneity treatment, ordering ambiguities, Choleski decomposition limitations, sensitivity to omitted variables and measurement errors.
  - Local projections provide direct estimation of dynamic responses with transparent control specification.

*Source: Author’s calculations.*

### Appendix 3 presents the results of a static approach to test the reaction of markets and voters to positive and

### wp1739 - Appendix 3 presents the results of a static approach to test the reaction of markets and voters to positive and

### Methodology: identification and estimation choices
- Primary dynamic identification: local projection method (LP). Advantages emphasized:
  - LP avoids identification and small-sample size limitations of larger structural systems.
  - Lagged dependent variables enter as control variables and are not used to derive IRFs.
  - Estimated IRFs’ confidence intervals are computed directly using standard errors of estimated coefficients without Monte Carlo simulations.
- ARDL alternative discussed:
  - Estimating an Autoregressive-Distributed-Lag (ARDL) model and computing IRFs is possible (Romer and Romer, 1989; Cerra and Saxena, 2008).
  - Drawbacks: ARDL-derived IRFs tend to be lag-sensitive and long-lasting effects may reflect one-type-of-shock models when the dependent variable is very persistent (Cai and Den Haan, 2009).
- Small-sample bias consideration:
  - Presence of a lagged dependent variable and country fixed effects could bias estimation of parameters in small samples (Nickell, 1981).
  - Finite sample bias in this study is around 0.03 (that is, 1/T, where T is 38).
- Regime-dependent responses:
  - To test dependence on the state of the business cycle, Equation (3) is estimated with z an indicator of the state of the economy (using the real GDP growth rate) normalized to have zero mean and unit variance.
  - This approach is equivalent to Granger and Teravistra’s (1993) smooth transition autoregressive model and uses more observations than estimating VARs for each regime, increasing stability and precision of IRFs.

### Impulse-response results (Figures 8 and 9): markets versus voters
- General findings from Figure 8 (LP estimates of Equation 2):
  - After a promise gap shock, 5-year government spreads increase strongly and are statistically significant at usual levels.
  - The reaction of voters is weaker: government popularity declines gradually but the impulse-response function is statistically weaker.
  - Note: Dotted lines in figures equal 90 percent confidence error bands.
- Regime-conditioned findings from Figure 9 (LP estimates of Equation 3, conditioned on business cycle phase):
  - Markets penalize incumbent governments more (raising government bond spreads) and for a longer period for not delivering on promises during bad times (recession).
  - Voter reaction remains muted; during bad times there is instead a marginally significant decrease in popularity one year after the promise gap shock.
  - Figures present responses for both Recession and Expansion regimes for:
    - Response: 5-year bond spreads
    - Response: Government Popularity
  - Note: Dotted lines equal 90 percent confidence error bands.

### Key empirical findings and statistics
- Finite sample bias: 0.03 (1/T where T is 38).
- Financial markets:
  - 5-year government spreads respond strongly and significantly to promise gap shocks.
  - Market penalty is larger and more persistent during recessions.
- Electorate:
  - Government popularity declines gradually after promise gap shocks.
  - Voter responses are statistically weaker than market responses; marginally significant decline during bad times at around one year after shock.

### Conclusions and policy implications (Section V)
- Core conclusions:
  - Fiscal consolidation promise gaps (distance between planned fiscal adjustments and actual consolidations) are sizeable.
  - Both economic and political factors affect promise gaps.
    - Smaller initial debt levels, annual improvements in the output gap and smaller forecast errors narrow promise gaps.
    - Political factors matter: newly elected governments, with a large margin of majority and high government effectiveness are more likely to deliver on planned fiscal adjustments.
  - Financial markets penalize (reward) underperformers (over-performers); the electorate is less responsive.
- Policy implications (bullet form preserved from text):
  - Policymakers should be more cautious in formulating ambitious fiscal consolidations (especially when they are revenue driven).
  - Governments can deliver on promised spending cuts by improving their capacity to deliver public services efficiently.
  - Stronger starting conditions (as reflected by lower fiscal imbalances) could lead to lower promise gaps later on.
  - Policy-makers should not be tempted to deviate from initial adjustment plans hoping to compensate a subsequent market backlash with a boost in popularity because voters do not react positively to broken promises.

### Appendix 1: Sample construction and selected country consolidation episodes (data sources and examples)
- Data construction:
  - Combined three sources:
    - Devries et al. (2011) covering 17 advanced economies between 1978 and 2009.
    - Alesina et al. (2015) narrative dataset expanded to 2013 for 11 countries (United States, United Kingdom, Austria, Belgium, Denmark, France, Germany, Italy, Ireland, Portugal, and Spain).
    - European Commission’s Stability and Convergence Programmes (first vintage) complemented with country budget sources to expand coverage to 2015.
  - By construction, sustained fiscal adjustments over various years would be compiled in tranches corresponding to the three sources.
- Selected country-year consolidation size and composition (percent of GDP) — values preserved exactly as in source:
  - Australia (Source: Commonwealth of Australia Budget)
    - 2010: Consolidation size 1.4; Tax size 0.15; Expenditure size 1.25; Tax-based 0; Expenditure-based 1
    - 2011: Consolidation size 1.9; Tax size 1.35; Expenditure size 0.55; Tax-based 1; Expenditure-based 0
    - 2012: Consolidation size 3; Tax size 1.7; Expenditure size 1.3; Tax-based 1; Expenditure-based 0
    - 2014: Consolidation size 1.2; Tax size 0.4; Expenditure size 0.8; Tax-based 0; Expenditure-based 1
    - 2015: Consolidation size 0.5; Tax size 0.55; Expenditure size -0.05; Tax-based 1; Expenditure-based 0
  - Austria (Source: Stability and Convergence Programme)
    - 2015: Consolidation size 0.1; Tax size 0.15; Expenditure size -0.05; Tax-based 1; Expenditure-based 0
  - Belgium (Source: Stability and Convergence Programme)
    - 2014: Consolidation size 0.4; Tax size -0.9; Expenditure size 1.3; Tax-based 0; Expenditure-based 1
    - 2015: Consolidation size 0.4; Tax size -0.65; Expenditure size 1.05; Tax-based 0; Expenditure-based 1
  - Canada (Source: Canada’s Economic Action Plan Budget)
    - 2010: Consolidation size 0.4; Tax size 0.4; Expenditure size 0; Tax-based 1; Expenditure-based 0
    - 2011: Consolidation size 0.3; Tax size 0.05; Expenditure size 0.25; Tax-based 0; Expenditure-based 1
    - 2012: Consolidation size 0.2; Tax size -0.1; Expenditure size 0.3; Tax-based 0; Expenditure-based 1
    - 2013: Consolidation size 0.4; Tax size 0.2; Expenditure size 0.2; Tax-based 0; Expenditure-based 0
    - 2014: Consolidation size 0.8; Tax size 0.25; Expenditure size 0.55; Tax-based 0; Expenditure-based 1
    - 2015: Consolidation size 0.2; Tax size 0.45; Expenditure size -0.25; Tax-based 1; Expenditure-based 0
  - Denmark
    - Denmark did not show evidence of fiscal consolidation in 2014 or 2015.
  - Finland (Source: Stability and Convergence Programme)
    - 2011: Consolidation size 1.9; Tax size 0.75; Expenditure size 1.15; Tax-based 0; Expenditure-based 1
  - France (Source: Stability and Convergence Programme)
    - 2014: Consolidation size 0.5; Tax size 0.05; Expenditure size 0.45; Tax-based 0; Expenditure-based 1
    - 2015: Consolidation size 0.1; Tax size -0.2; Expenditure size 0.3; Tax-based 0; Expenditure-based 1
  - Germany
    - Germany did not show evidence of fiscal consolidation in 2014 or 2015.
  - Italy (Source: Stability and Convergence Programme)
    - 2014: Consolidation size 0.4; Tax size -0.2; Expenditure size 0.6; Tax-based 0; Expenditure-based 1
    - 2015: Consolidation size 0.1; Tax size -0.3; Expenditure size 0.4; Tax-based 0; Expenditure-based 1
  - Ireland (Source: Stability and Convergence Programme)
    - 2014: Consolidation size 2.4; Tax size 0.3; Expenditure size 2.1; Tax-based 0; Expenditure-based 1
    - 2015: Consolidation size 1.2; Tax size -0.9; Expenditure size 2.1; Tax-based 0; Expenditure-based 1
  - Japan
    - CAPB in Japan in 2014 and 2015 improved, but episode not included in sample because published data is unavailable.
  - Netherlands (Source: Stability and Convergence Programme)
    - 2011: Consolidation size 1.9; Tax size 0.6; Expenditure size 1.3; Tax-based 0; Expenditure-based 1
    - 2012: Consolidation size 0.1; Tax size 0; Expenditure size 0.1; Tax-based 0; Expenditure-based 1
    - 2013: Consolidation size 0.8; Tax size 0.65; Expenditure size 0.15; Tax-based 1; Expenditure-based 0
    - 2015: Consolidation size 0.3; Tax size -0.6; Expenditure size 0.9; Tax-based 0; Expenditure-based 1
  - Portugal (Source: Stability and Convergence Programme)
    - 2014: Consolidation size 0.9; Tax size -0.4; Expenditure size 1.3; Tax-based 0; Expenditure-based 1
    - 2015: Consolidation size 1.4; Tax size -0.1; Expenditure size 1.5; Tax-based 0; Expenditure-based 1

*Source: wp1739 (Appendix 3 and Appendix 1 content) from the provided IMF PDF content.*

### 3.3 percent of GDP, leading to an improvement of the baseline for 2015.

### wp1739 - 3.3 percent of GDP, leading to an improvement of the baseline for 2015.

### Fiscal consolidation: country-level outcomes and composition (selected countries)
- Spain
  - 2015:
    - Consolidation size: 1.4 percent of GDP
    - Tax size: -0.1 percent of GDP
    - Expenditure size: 1.5 percent of GDP
    - Tax-based: 0
    - Expenditure-based: 1
    - Motivation: deficit reduction under the Stability Programme; headline target of DBP for 2015: headline deficit of 2.7 percent of GDP.
  - 2014:
    - Consolidation size: 1.7 percent of GDP
    - Tax size: 0.75 percent of GDP
    - Expenditure size: 0.95 percent of GDP
    - Tax-based: 0
    - Expenditure-based: 1
    - Motivation: deficit reduction; Stability Programme aimed to bring fiscal deficit below three percent of GDP in 2016 and reach a medium-term objective of a balanced budgetary position in structural terms in 2017.
  - Repeated statement (emphasis): Fiscal consolidation in 2015 amounted to 1.4 percent of GDP with expenditure cuts of 1.5 percent of GDP and tax cuts of 0.1 percent of GDP; motivated by deficit reduction under the Stability Programme, aiming to bring the fiscal deficit below three percent of GDP in 2016 and reach a structural balanced budget in 2017.

- Sweden
  - 2011:
    - Consolidation size: 0.6 percent of GDP
    - Tax size: -0.8 percent of GDP
    - Expenditure size: 1.4 percent of GDP
    - Tax-based: 0
    - Expenditure-based: 1
    - Note: roll-back of stimulus measures to local governments; effective consolidation solely expenditure-based.
  - 2015:
    - Consolidation size: 0.5 percent of GDP
    - Tax size: 0.1 percent of GDP
    - Expenditure size: 0.4 percent of GDP
    - Tax-based: 0
    - Expenditure-based: 1
    - Context: fiscal policy supported activity via automatic stabilizers; some permanent personal and corporate income tax cuts and increased expenditures for sickness benefits and asylum seekers decreased the fiscal balance.

- United Kingdom
  - 2014:
    - Consolidation size: 1.1 percent of GDP
    - Tax size: 0.05 percent of GDP
    - Expenditure size: 1.05 percent of GDP
    - Tax-based: 0
    - Expenditure-based: 1
    - Drivers: spending reduction announced in Autumn Statement 2013 and reduced costs of public service pensions; TME to be reduced by around a further £2 billion each year from 2016-2017.
  - 2015:
    - Consolidation size: 1.3 percent of GDP
    - Tax size: 0.2 percent of GDP
    - Expenditure size: 1.1 percent of GDP (noting textual emphasis: 1.01 percent of GDP appears in narrative)
    - Tax-based: 0
    - Expenditure-based: 1
    - Context: continuation of 2010 long-term plan to halve the deficit as a share of GDP; aim to achieve a surplus in 2019-2020; planned around £37 billion of further consolidation measures.

- United States
  - 2014:
    - Consolidation size: 1.6 percent of GDP
    - Tax size: 1.1 percent of GDP
    - Expenditure size: 0.5 percent of GDP
    - Tax-based: 1
    - Expenditure-based: 0
    - Drivers: spending caps in the Budget Control Act of 2011, increase in tax rates for top earners at the beginning of 2013, end of temporary payroll tax holiday.
  - 2015:
    - Consolidation size: 0.6 percent of GDP
    - Tax size: 0.95 percent of GDP
    - Expenditure size: -0.35 percent of GDP
    - Tax-based: 1
    - Expenditure-based: 0
    - Context: Bipartisan Budget Act of 2015 avoided a Federal shutdown, partly relieved automatic Federal spending cuts, relaxed the Federal debt limit; government purchases at Federal, State, and local levels added to consolidation.

### Appendix 2 — Robustness checks: homogeneous samples and main findings
- Methodological concerns addressed:
  - Narrative methodology subjectivity; assumed judgemental measurement errors are evenly distributed across three sources.
  - Potential structural breaks from mixing data sources between 1978 and 2015.
- Robustness strategy:
  - Replicated analysis on homogeneous sample of Devries et al. (2011) covering 17 advanced countries between 1978 and 2009 (tables A1.1 and A1.2).
  - Replicated analysis on a sample built only with Stability and Convergence Programs (SCP) submitted by the 28 EU Member States between 1998 and 2015 (tables A1.3 and A1.4).
  - Conclusion: tests on these two alternative homogeneous samples confirm robustness of main findings.

- Selected regression results (preserve exact coefficients and statistics reported)
  - Devries’ database (1978-2009), Table A2.1:
    - Specification (1): Lagged debt coefficient: -0.032*** (0.007); Constant: 1.892*** (0.638); Observations: 162; R-squared: 0.225.
    - Specification (2): Lagged debt coefficient: -0.051*** (0.012); Change in output gap coefficient: -0.611*** (0.082); Constant: 0.181 (0.450); Observations: 165; R-squared: 0.368.
    - Specification (3): Change in output gap coefficient: -0.619*** (0.119); Lagged GDP forecast error coefficient: -0.404** (0.166); Constant: -0.134 (0.736); Observations: 96; R-squared: 0.428.
    - Specification (4): Lagged GDP forecast error coefficient: -0.331*** (0.123); Constant: 3.479*** (0.908); Observations: 96; R-squared: 0.712.
  - Devries’ database (1978-2009), Table A2.2 (economic and political determinants):
    - Specification highlights:
      - Lagged debt coefficients: -0.073*** (0.017), -0.051*** (0.013), -0.061** (0.025), -0.017 (0.053) across specifications.
      - Change in output gap coefficients: -0.418** (0.157), -0.617*** (0.121), -0.440 (0.271), -0.346 (0.536).
      - Recent elections coefficient: -0.377* (0.393) in one specification; Margin of majority and Government effectiveness show large but less precisely estimated coefficients in some specifications.
      - Lagged GDP forecast error coefficients include -0.268** (0.126) and -0.329** (0.125) in specifications.
      - Observations: 73, 96, 41, 29; R-squared: 0.636, 0.713, 0.868, 0.798.
  - SCP’s database (1998-2015), Table A2.3:
    - Specification (1): Lagged debt -0.026*** (0.009); Constant 1.850*** (0.708); Observations: 252; R-squared: 0.148; Number of EU countries: 27.
    - Specification (2): Lagged debt -0.026*** (0.008); Change in output gap -0.242*** (0.052); Constant 0.571 (0.504); Observations: 253; R-squared: 0.196; Number of EU countries: 27.
    - Specification (3): Change in output gap -0.272*** (0.053); Lagged GDP forecast error -0.059 (0.046); Constant 0.509 (0.503); Observations: 224; R-squared: 0.122; Number of EU countries: 23.
    - Specification (4): Lagged GDP forecast error -0.107** (0.043); Constant 1.906*** (0.635); Observations: 224; R-squared: 0.263; Number of EU countries: 23.
  - SCP’s database (1998-2015), Table A2.4 (economic and political determinants):
    - Selected coefficients:
      - Lagged debt: -0.026*** (0.008), -0.026*** (0.009), -0.035*** (0.009), -0.033*** (0.009).
      - Change in output gap: -0.299*** (0.054), -0.271*** (0.053), -0.240*** (0.055), -0.266*** (0.057).
      - Recent elections: -1.071* (0.586) and -1.289** (0.632) in specified models.
      - Lagged GDP forecast error: -0.110** (0.043), -0.107** (0.043), -0.091** (0.045), -0.092** (0.045).
      - Government effectiveness: -1.798** (0.898) and -1.795** (0.889) in specified models.
      - Observations: 224, 224, 188, 188; R-squared: 0.275, 0.264, 0.314, 0.336; Number of EU countries: 23 in reported specifications.

### Appendix 3 — Consequences of consolidation promise gaps: regression analysis and implications
- Estimation approach:
  - Static OLS estimation of equation (A3.1) with robust standard errors clustered at the country level.
  - Dependent variables: 5-year government bond spreads (relative to Germany) and government popularity.
  - Key explanatory variable: Consolidation Promised Gap (CPG).
  - Controls: real GDP growth, public debt (percent of GDP), changes in unemployment, 21-month real GDP growth forecast errors, among others.
- Markets’ reaction to consolidation promise gaps (Table A3.1)
  - Main finding: financial markets tend to punish fiscal underperformance; markets react negatively when governments fail to meet planned fiscal adjustments.
  - Selected coefficients and statistics (preserve exact reported estimates):
    - Underperformers:
      - CPG: 0.283*** (0.102) in one specification; 0.247* (0.140) in another.
      - Real GDP growth: -0.058 (0.152) and -0.370 (0.262) in reported columns.
      - Public debt: 0.013 (0.008) in one column.
      - Observations: 55 across several columns.
      - R-squared: 0.128, 0.130, 0.166 in underperformer columns.
    - Overperformers:
      - CPG: 0.260** (0.102) in one column; 0.030 (0.329) and 0.077 (0.325) and 0.027 (0.303) in others.
      - Public debt: 0.039** (0.016) in one overperformer column.
      - Observations: 30 across overperformer columns.
      - R-squared: 0.000, 0.069, 0.180 in overperformer columns.
  - Note: Dependent variable is the 5-year government bond spreads relative to Germany. Time and country fixed effects omitted for parsimony. *, **, *** denote statistical significance at the 10, 5 and 1 percent levels.

- Voters’ reaction to consolidation promise gaps (Table A3.2)
  - Main finding: voters do not seem to react strongly to consolidation gaps, regardless of sign.
  - Selected coefficients and statistics:
    - Underperformers:
      - CPG: -1.775** (0.829) in one specification; -0.866 (1.896) in another.
      - Lagged GDP forecast error: 4.295*** (1.235) and 3.023 (3.533) in reported columns.
      - Change in unemployment rate: -2.489 (1.904) in one column.
      - Observations: 66 and 36 for underperformer columns.
      - R-squared: 0.603 and 0.681.
    - Overperformers:
      - CPG: 0.438 (1.465) and -1.029 (2.054) in reported columns.
      - Lagged GDP forecast error: 0.815 (1.483) and -4.351* (2.424).
      - Change in unemployment rate: 1.267 (3.170) in one column.
      - Observations: 57 and 35 for overperformer columns.
      - R-squared: 0.748 and 0.792.
  - Note: Dependent variable is government popularity as defined in the main text. Time and country fixed effects omitted for parsimony. *, **, *** denote statistical significance at the 10, 5 and 1 percent levels.

*Source: wp1739 - 3.3 percent of GDP, leading to an improvement of the baseline for 2015.*

### References

### wp1739 - References

### References

- Afonso, A. and J. T. Jalles, 2014, “Assessing Fiscal Episodes,” Economic Modelling, 37, 255-270.  
- Alesina, A., O. Barbiero, C. Favero, F. Giavazzi, and M. Paradisi, 2015, “Austerity in 2009-2013,” NBER Working Paper, No. 20827.  
- Angelopoulos, K., J. Malley, and A. Philippopoulos, 2012, “Optimal Taxation and the Skill Premium,” CESifo Working Paper, No. 3706.  
- Auerbach, A. J. and Y. Gorodnichenko, 2013, “Corrigendum: Measuring the Output Responses to Fiscal Policy,” American Economic Journal: Economic Policy, 5(3), 320-322.  
- Baldacci, E. and M. Kumar, 2010, “Fiscal Deficits, Public Debt, and Sovereign Bond Yields,” IMF Working Paper, WP/10/184.  
- Beetsma, R., M. Giuliodori, and P. Wierts, 2009, “Budgeting Versus Implementing Fiscal Policy in the EU,” CEPR Discussion Paper, No. 7285.  
- Beetsma, R., B. Bluhm, M. Giuliodori, and P. Wierts, 2013, “From Budgetary Forecasts to Ex-Post Fiscal Data: Exploring the Evolution of Fiscal Forecast Errors in the European Union,” Contemporary Economic Policy, 31 (4): 795-813.  
- Beetsma, R., J. Cimadomo, O. Furtuna, and M. Giuliodori, 2015, “The confidence effects of fiscal consolidations,” ECB Working Paper, No. 1770.  
- Beck, N. and J. Katz, “What to do (and not to do) with Time-Series Cross-Section Data,” American Political Science Review, 89(3), 634-647.  
- Boix C., 1997, “Political Parties and the Supply Side of the Economy: The Provision of Physical and Human Capital in Advance Economies, 1960-90”. American Journal of Political Science, 41(3), 814-845.  
- Cai, X. and W. Den Haan, 2009, “Predicting recoveries and the importance of using enough information,” CEPR Discussion Paper, No. 7508.  
- Cerra, V. and S. C. Saxena, 2008, “Growth Dynamics: The Myth of Economic Recovery,” American Economic Review, 98(1), 439-457.  
- Cruz, C., P. Keefer, and C. Scartascini, 2015, Database of Political Institutions 2015, Washington: Inter-American Development Bank: https://publications.iadb.org/handle/11319/7408?locale-attribute=en  
- Devries, P., J. Guajardo, D. Leigh, and A. Pescatori, “A New Action-based Dataset of Fiscal Consolidation,” IMF Working Paper, WP/11/128.  
- Ericsson, N. R., D. F. Hendry, and K. M. Prestwich, 1997, “The Demand for Broad Money in the United Kingdom, 1878-1993,” International Finance Discussion Papers, No. 596.  
- Escolano, J., L. Jaramillo, C. Mulas-Granados, and G. Terrier, “How Much is A Lot? Historical Evidence on the Size of Fiscal Adjustments,” IMF Working Paper, WP/14/179.  
- Frankel, J., 2011, “A Solution to Fiscal Procyclicality: The Structural Budget Institutions Pioneered by Chile,” NBER Working Paper, No. 16945.  
- Franzese, R., 2002, “Electoral and Partisan Cycles in Economic Policies and Outcomes”, Annual Review of Political Science, 5, 369-421.  
- Granger, C. W. J. and T. Terasvirta, 1993, Modelling Nonlinear Economic Relationships, Oxford: Oxford University Press.  
- Hendry, D., 1995, Dynamic econometrics, Oxford: Oxford University Press.  
- Jordá, O., 2005, “Estimation and Inference of Impulse Responses by Local Projections”, American Economic Review, 95(1): 161-182.  
- Mierau, J., R. Jong-A-Pin, and J. de Haan, 2007. "Do political variables affect fiscal policy adjustment decisions? New empirical evidence," Public Choice, 133(3), 297-319.  
- Morris, R. and L. Schuknecht, “Structural Balances and Revenue Windfalls,” ECB Working Paper, No. 737.  
- Mulas-Granados, C., 2003, “The Trade-Off between Growth & Equality and the Economic Impact of Alternative Fiscal Adjustment Strategies in the EU,” European Economy Group Working Paper, No. 20/2003.  
- Mulas-Granados, C., 2006, Economics, Politics and Budgets: The Political Economy of Fiscal Adjustments in Europe, London: Palgrave-Macmillan  
- Müller, W. C. and K. Strom, 1999, Policy, Office or Votes? How Political Parties in Western Europe Make Hard Decisions, Cambridge, Cambridge University Press  
- Nickell, S., 1981, “Biases in Dynamic models with Fixed Effects”, Econometrica, 49 (6): 1417-1426.  
- Konishi, H., 2006. "Spending cuts or tax increases? The composition of fiscal adjustments as a signal," European Economic Review, 50(6), 1441-1469, August.  
- Perotti R., 1998, “The Political Economy of Fiscal Consolidations,” Scandinavian Journal of Economics, 100 (1), 367-394.  
- Perotti, R. and Y. Kontopoulos, 2002, “Fragmented Fiscal Policy,” Journal of Public Economics, 86, 191-222.  
- Potrafke, N., 2011. "Does government ideology influence budget composition? Empirical evidence from OECD countries," Economics of Governance, 12(2), 101-134.  
- Romer, C. D. and D. H. Romer, 1989, “Does Monetary Policy Matter? A New Test in the Spirit of Friedman and Schwartz,” NBER Macroeconomic Annual, 4, 121-184.  
- Romer, C. D. and D. H. Romer, 2010, “The Macroeconomic Effects of Tax Changes: Estimates Based on a New Measure of Fiscal Shocks,” American Economic Review, 100(3), 763-801.  
- Rousseeuw, P. J., and K. Van Driessen, 1999, “A Fast Algorithm for the Minimum Covariance Determinant Estimator”, Technometrics, 41(3): 212-223.  
- Tavares, J., 2004, “Does Right or Left Matter? Cabinets, Credibility and Fiscal Policy,” Journal of Public Economics, 8(12).  
- Von Hagen, J., 2010, “The Sustainability of Public Finances and Fiscal Policy Coordination in the EMU,” CASE Network Studies and Analyses, No. 412.  
- Von Hagen, J., A. H. Hallett, and R. Straucht, 2001, “Budgetary Consolidation in EMU,” Economics Papers, 148, Brussels, European Commission.

*Source: wp1739 - References*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp1739.pdf_