## wpiea2024194-print-pdf - 2022

## Source details

**Canonical URL:** [wpiea2024194-print-pdf - 2022](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024194-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024194-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024194-print-pdf.pdf.json)

---

### Definition and measurement of expansion discourse
- Expansion discourse: support for spending on social matters, including support for the welfare state (e.g. health-, child-, elder care and pensions, social housing), education, and demand policies (Figure notes, Panel A).
- Advanced-economy measurement progressively includes platform calls for infrastructure, defense, the environment, and provision of incentives to businesses as potential government spending topics (Panel B).
- Baseline measure: draws only from Manifesto Project (MP) items most plausibly implying a fiscal cost and excludes items with both spending pressures and broader value judgments (e.g., public involvement in infrastructure, defense, the environment, incentives to firms).
- Comprehensive definition note: expansion discourse “accounts for well over a third of all platform content by the end of the sample period on average.”

### Trends over time and across country groups
- In advanced economies, the share of expansion discourse “increases since the 1990s” even after accounting for additional policy realms, with the main contribution from the baseline social-spending-centered measure.
- Parties across traditional families devote a progressively larger share of their platforms to support for government spending in more recent decades.
- Restraint content trends down among parties that traditionally championed fiscal prudence (conservatives, liberals, nationalists) up to the 1980s.
- Manifesto Project coverage and summary statistics (1960-2022):
  - Expansion (platform): N 4,504; Mean 12.7; Median 12.1; SD 7.90; Min 0; Max 68.4
  - Expansion (country-year): N 721; Mean 12.9; Median 12.5; SD 5.3; Min 1.5; Max 39.5
  - Restraint (platform): N 4,504; Mean 2.5; Median 1.1; SD 3.80; Min 0; Max 42.6
  - Restraint (country-year): N 721; Mean 2.5; Median 2.0; SD 2.20; Min 0; Max 21.0

### Party-family and U.S. party dynamics
- Six ideology families analyzed: left (LEF), socialists (SOC), liberals (LIB), Christian-democrats (CHR), conservatives (CON), and nationalists (NAT).
- U.S. presidential platforms, 1920-2020:
  - Expansion measured using MP variables per409, per504, per506; restraint using per414, per505, per507.
  - Historical highs in expansion discourse among Democrats since the 1996 elections; large bipartisan drop between the early 1980s and the 2000s.
  - Uptick in pro-restraint content in the 2012 Republican platform likely driven by criticism of the Affordable Care Act and historically high federal deficits after the global financial crisis.

### Polarization of fiscal discourse
- Polarization metric follows Dalton (2008): a standard deviation-like formula using party platform shares (Appendix B).
- Fiscal polarization constructed from relative pro-restraint positions: RR = FD_Restraint − FD_Expansion; dispersion weighted by party vote shares.
- Findings:
  - Fiscal polarization, relative to mean platform share devoted to fiscal topics, has historically been larger than relative polarization on non-fiscal issues.
  - In advanced economies, since the 1990s the distinction has blurred as parties converge toward relatively pro-government expansion platforms.

### Drivers — macro-fiscal conditions at elections
- Regression specification (country-election OLS, Equation (1)):
  - FD_d_{i,e} = α_i + β C_{i,e} + γ X_{i,e} + ε_{i,e}
  - FD_d is fiscal discourse in support of d = {Expansion, Restraint}; C is the one-year lag of the relevant economic indicator; α_i country fixed effect; X controls for government perimeter and data source dummies.
- Coefficient summary and interpretation:
  - On average, platforms drafted when the deficit is higher by 1 percentage point of GDP feature 0.22 percentage points less expansion content and 0.10 percentage point more restraint discourse (sample: 653 country-election years; coefficients significant at 1 percent).
  - Expansion and restraint discourse respond in opposite ways to the same macro-fiscal variable.
  - Fiscal discourse responds more systematically to fiscal variables (deficit, public debt) than to standard macro variables (real GDP growth, inflation).
  - For advanced economies under adverse macro conditions (lower real GDP growth and higher inflation) discourse is on average more conservative, though most estimates lack significance.
  - Results hold when using predetermined elections from the NELDA dataset.

### Macro-fiscal conditions at elections — heterogeneity (Figure 9)
- Regressions by time period and country group (90 percent country-clustered confidence intervals):
  - Expansion and restraint discourse diverge in response to macro-fiscal variables; responses vary by indicator, time period, and country group.
  - Fiscal variables (deficit, debt) show tighter links to platform content than growth or inflation.
  - For emerging markets, coefficients on fiscal conditions are smaller relative to advanced economies, suggesting differences in the ability to steer rhetoric toward fiscal conservatism.
- Robustness:
  - No meaningful difference when examining contemporaneous deficits (Figure A13).
  - Extended expansion definitions respond more to inflation in advanced economies and to growth in emerging markets since the 1990s (Appendix C).

### Fiscal events — public debt surges
- Debt surge definition (extension of Jalles and Medas (2022)):
  - SurgeCutoff_{i,t} = 1 if ∆D_{i,t,t−1} > max{μ_{∆D_i} + σ_{∆D_i}, Cutoff}, with cutoffs ranging from 1 to 5 percent of GDP; years since 2020 and 1st-percentile debt growth years excluded.
- Sample counts and regression sample:
  - 434 country-years with debt surged using 1 percent of GDP common cutoff; 336 cases using 5 percent cutoff.
  - Regression sample: 587 country-election years; 91 feature a contemporaneous debt surge; 76 a debt surge three years earlier; overall 219 country-election years associated with a surge in the same year or any of the three previous years.
- Empirical findings (Figure 10):
  - Support for expansion: surge episodes up to two calendar years before an election show negative point estimates; significantly negative only for shocks two years ahead (~1 percentage point less expansion discourse on average).
  - Support for restraint: reliably more support for restraint (up to 0.9 percentage points on average) in the first few years after a debt surge.
  - Point estimates increase slightly with higher common cutoffs, indicating larger adverse fiscal shocks elicit stronger political reactions toward fiscal moderation.
  - Interpretation: large increases in public debt by country historical standards tend to moderate support for expansion with effects taking time to materialize; restraint increases more immediately after a surge.

### Fiscal events — introduction of fiscal rules
- Focus: budget balance rules (operational limits on overall, primary, or cyclically adjusted balance) over 1985-2021 using IMF Fiscal Rules Dataset.
- Event-study setup (Equation (4)): staggered adoption, relative election-cycle distances to adoption, country and election-year fixed effects, standard errors clustered by country.
- Sample: 48 of 65 countries adopt a fiscal rule over 1985-2021; analysis considers earliest adoption per country.
- Empirical findings (Figure 11):
  - Support for expansion: virtually no impact from rule adoption; coefficients mostly negative but small and insignificant.
  - Support for restraint: after adoption, party platforms adopt a growing share of fiscal restraint content in subsequent elections.
  - Interpretation: formal fiscal frameworks do not significantly alter propensity to favor government spending but inject rhetoric with elements of fiscal discipline over several election cycles.

### Sensitivity checks for fiscal institutions and sample definition (Appendix D)
- Including adoption of any fiscal rule type yields qualitatively similar results (Figure A18, Panels A–B).
- Including countries that never adopt a fiscal rule produces weak but increasingly negative effects on expansion discourse and smaller effects on restraint (Figure A18, Panels C–D).
- Comparison with IMF Fiscal Council Dataset (2022): similar direction of effects from establishment of fiscal councils, independently or interacting with fiscal rules (Figure A19).

### Demand-side considerations: voter preferences (WVS/EVS evidence)
- Data: World Value Survey (WVS) and European Value Survey (EVS), 1989-2022.
- Survey items: preferences for more state ownership, government responsibility to provide for all citizens, income redistribution, taxing the rich and subsidizing the poor, state aid for unemployment, making people’s incomes equal; original scale 0 to 10 (higher = more support for government action).
- Findings:
  - No clear increasing trend in public support for government intervention comparable to manifesto-based expansion discourse trends.
  - Vote-share analysis:
    - Using country-specific medians: vote shares increasingly accrue to parties above (below) historical median for expansion (restraint).
    - Using country-by-decade medians: virtually no trend; vote-share shifts attenuate and are broadly stable once accounting for country-decade trends.
  - Interpretation: observed vote shifts likely reflect supply-side platform changes rather than rising average public demand.

### Structural change and inequality
- Inequality measures examined:
  - Composite Gini index (Milanovic, 2019), 1960-2015.
  - Wealth share ratio: top 10% vs bottom 50% (World Inequality Database), 1960-2022.
- Findings:
  - Inequality rising in advanced economies since the 1990s; mixed trends for EMDEs.
  - No consistent within-country relationship between fiscal discourse and inequality, except a positive association between the wealth ratio and expansion discourse.
  - Conclusion: no direct evidence that voter preferences or inequality trends fully explain long-term shifts in fiscal discourse; politicians appear to actively craft fiscal narratives.

### The aftermath of fiscal discourse changes — local projections framework
- Local projections model (Jordà 2005; Equation (5) expansion):
  - y_{i,t+h} − y_{i,t−1} = α_i + δ_t + β^e_h ΔExp_{i,t} + β^r_h ΔRes_{i,t} + θ_h E_{i,t} + γ_h X_{i,t} + ε_{i,t}
  - ΔExp_{i,t}, ΔRes_{i,t}: change in mean share of expansion and restraint platform content across two consecutive elections (non-zero only in election years).
  - Projection horizon up to 9 years; controls include lagged outcome changes and GDP growth terms as specified.
- Distribution of shocks (Figure A20):
  - Sample standard deviation of expansion shock: 4.57 percent (1960-2022 regression sample).
  - Sample standard deviation of restraint shock: 2.12 percent (1960-2022 regression sample).
  - 1990-2022 regression sample: expansion shock SD 4.74 percent; restraint shock SD 1.99 percent.

### Impulse response findings — fiscal and macro outcomes (Figures 15 and 16)
- Responses trace cumulative changes for a one standard deviation shock in expansion (reported) and restraint (reported); confidence bands: 68% and 90% country-clustered intervals.
- Key quantitative impulse responses (Figure 15):
  - Fiscal deficit:
    - One standard deviation increase in expansion discourse → 0.26 percent of GDP increase in the fiscal deficit by the end of the projection horizon.
    - One standard deviation increase in restraint discourse → 0.37 percent of GDP reduction in the fiscal deficit (significant toward the end of the horizon).
  - Government debt:
    - Expansion shock: benign but muted response.
    - Restraint shock: initial increase in debt-to-GDP ratio in near term (driven by temporary GDP decline), despite average deficit reduction.
  - GDP level:
    - Expansion shock: muted positive short-run effects; deficits grow over time as temporary boosts dissipate.
    - Restraint shock: GDP declines for several years following increase in restraint discourse.
  - Election-cycle dynamics: S-shaped impulse response of the budget deficit to an election-year dummy, consistent with political budget cycle literature.
- Components of the deficit (Figure 16):
  - Expansion shock:
    - Primary deficit: gradual increase driven by primary spending increases.
    - Government interest payments: relatively muted response.
    - Revenues: no dominant role in primary deficit increase.
    - Interpretation: expansion narratives translate more into spending initiatives than revenue cuts; discretionary spending drives deficit buildup.
  - Restraint shock:
    - Primary deficit: declines and gains significance earlier than overall deficit.
    - Revenues: positive point estimates throughout the horizon, suggesting revenue measures are a primary channel.
    - Primary spending: expenditure cuts contribute later in the horizon.
    - Government interest payments: short-run increase in interest payments-to-GDP ratio consistent with output decline and stable/increasing debt.
    - Interpretation: restraint narratives translate into revenue measures first, with expenditure control effects later; short-run output costs may temporarily raise debt ratios.

### Robustness and additional checks
- Starting sample in 1990 (Figure A21): qualitatively similar findings but starker — deficit responses almost twice as large by year 8 and gain significance earlier.
- Country-group heterogeneity (Figures A22–A23):
  - Advanced economies: patterns mirror full-sample; restraint shocks exhibit less pronounced short-term effects on deficits and GDP.
  - Emerging market and developing economies: restraint shocks lead to lower deficits short- and medium-term with enduring output costs but shorter-lived upticks in debt-to-GDP; expansion shocks less consequential.
- Expanded expansion definition (Figure A24): including infrastructure and defense yields larger and earlier significant responses in deficits and output than baseline.

### Conclusion (synthesis)
- Electoral campaigns amid deteriorating fiscal conditions feature more conservative fiscal talk (particularly around debt surges and adoption of constraining fiscal institutions).
- Over the long run, parties have increasingly competed over pro-government spending platforms with reduced emphasis on budget discipline.
- Long-term political forces can dominate short-run strategic adjustments driven by deficits, debt surges, and institutions like fiscal rules.
- Noted limitations and future directions: disentangling demand- and supply-side drivers, higher-frequency measure of fiscal discourse, identifying channels through which discourse affects public finances (cabinet vs opposition roles), and use of instrumental variables to strengthen identification.

*Source: wpiea2024194-print-pdf - Working Paper No. WP/2024/194*

### 2022.  In Panel A, expansion discourse encompasses support for spending on social matters, including support for the wel

### wpiea2024194-print-pdf - 2022

### Definition and measurement of expansion discourse
- Expansion discourse encompasses support for spending on social matters, including support for the welfare state (e.g. health-, child-, elder care and pensions, social housing), education, and demand policies (Figure notes, Panel A).
- For advanced economies (Panel B), measurement progressively includes platform calls for infrastructure, defense, the environment, and provision of incentives to businesses as potential government spending topics.
- Baseline measures define expansion discourse drawing only from Manifesto Project (MP) items that most plausibly imply a fiscal cost and exclude items that entail both potential spending pressures and broader value judgments (e.g., variables capturing party support for public involvement in infrastructure, defense, the environment, and incentives to firms).
- If a comprehensive definition is considered, expansion discourse “accounts for well over a third of all platform content by the end of the sample period on average.”

### Trends over time and across country groups
- The top dotted line in the figures displays the mean total platform share accounted for by all listed topics.
- Figure A6 provides the corresponding chart for emerging markets (not reproduced here).
- In advanced economies, the share of expansion discourse “increases since the 1990s” even after accounting for additional policy realms, with the main contribution provided by the baseline measure centered around social spending.
- Parties across traditional families devote a progressively larger share of their platforms in support of government spending in more recent decades (Figure 5, Panel A).
- Restraint content trends down among parties that traditionally championed fiscal prudence (conservatives, liberals, nationalists) up to the 1980s.

### Party-family and U.S. party dynamics
- The analysis considers six ideology families: left (LEF), socialists (SOC), liberals (LIB), Christian-democrats (CHR), conservatives (CON), and nationalists (NAT).
- Each bar in Figure 5 shows the party family-decade mean share of platform content in support of government expansion (Panel A) or fiscal restraint (Panel B); the 2010+ bar encompasses elections between 2010 and 2022.
- U.S. presidential platforms, 1920-2020 (Figure 6):
  - Panel A measures the share of each party’s platform advocating for government spending expansion using MP variables per409, per504, and per506.
  - Panel B measures restraint support using MP variables per414, per505, and per507.
  - Panel A displays historical highs in expansion discourse among Democrats since the 1996 elections, along with a large bipartisan drop between the early 1980s and the 2000s.
  - The observed uptick in pro-restraint content in the 2012 Republican platform is likely driven by criticism of the Affordable Care Act and historically high federal deficits following the global financial crisis.

### Polarization of fiscal discourse
- Polarization measurement follows Dalton (2008), using a standard deviation-like formula with party platform shares as constitutive elements (full methodology in Appendix B).
- Fiscal polarization: define the relative pro-restraint position of each platform as the difference between platform share of pro-restraint and pro-expansion content; compute dispersion in relative restraint positions weighting platforms by each party’s vote share.
- Overall ideological polarization: built on the RILE (right-left) scale from MP data.
- Non-fiscal polarization: computed after removing baseline fiscal variables from the RILE scale.
- Figure 7 findings:
  - Fiscal polarization, in proportion to the mean platform share devoted to fiscal topics, has historically been larger than relative polarization on non-fiscal issues.
  - In advanced economies, the distinction has become more blurred since the 1990s as parties have converged towards relatively pro-government expansion platforms.

### Drivers — macro-fiscal conditions at elections
- Research question: the degree to which fiscal discourse responds to macroeconomic and fiscal conditions around elections (platforms may respond to growth, fiscal risks, debt sustainability).
- Within-country visualization links fiscal discourse in an election year to the budget deficit using historical fiscal data from the Public Finance in Modern History (PFMH) database and IMF World Economic Outlook for recent periods.
- Estimated OLS regression at the country-election year level:
  - FD_d_{i,e} = α_i + β C_{i,e} + γ X_{i,e} + ε_{i,e}, (1)
  - where FD_d is fiscal discourse in support of d = {Expansion, Restraint}, i is country, e is election year.
  - C refers to the one-year lag of the relevant economic indicator for that country and election year.
  - α_i is a country fixed effect.
  - X is a vector of dummy variables capturing the government perimeter and data source of the economic information in use.
- Figure 8 (described): reports coefficient estimates and significance of lagged fiscal deficit in equations for expansion (Panel A) and restraint discourse (Panel B) and plots corresponding binscatter relationships for all countries over 1960-2022.
- The figure suggests that, within a given country, party platforms drafted when the deficit is higher by 1 percentage point of GDP feature (statement truncated in source).

*Source: wpiea2024194-print-pdf - 2022.*

### 0.22 percentage points less expansion content and 0.10 percentage point more restraint discourse

### 0.22 percentage points less expansion content and 0.10 percentage point more restraint discourse

### Fiscal discourse and lagged fiscal deficits at elections
- Estimated coefficients indicate on average 0.22 percentage points less expansion content and 0.10 percentage point more restraint discourse; coefficients are small in magnitude but significant at 1 percent level in a sample of 653 country-election years.
- Conditional binscatter and linear regressions (Equation (1)) using country, government perimeter, and data source fixed effects show:
  - Expansion and restraint discourse respond in opposite ways to the same macro-fiscal variable.
  - Fiscal discourse responds more systematically to fiscal variables (deficit, public debt) than to standard macroeconomic variables (real GDP growth, inflation).
  - For advanced economies, under adverse macroeconomic conditions (lower real GDP growth and higher inflation) discourse is on average more conservative (more restraint, less expansion), though most estimates are not significant.
  - Higher fiscal deficit or public debt ahead of an election anticipate party platforms with significantly more restraint and less government expansion content.
- The relationships hold when zooming into predetermined elections defined in the NELDA dataset.

### Macro-fiscal conditions at elections (Figure 9 summary)
- Plots show coefficients from regressions based on Equation (1) with 90 percent country-clustered confidence intervals.
- Panels compare:
  - Two time periods for advanced economies (Panels A and B).
  - Advanced versus emerging market and developing economies for 1990-2022 (Panels C and D).
- Key empirical patterns:
  - Expansion and restraint discourse diverge in response to macro-fiscal variables; responses depend on indicator, time period, and country group.
  - Fiscal discourse is more tightly linked to fiscal conditions (deficit, debt) than to growth or inflation.
  - For emerging markets, platform content also tends to be associated with fiscal conditions rather than growth and inflation, but coefficient estimates are smaller relative to advanced economies—suggesting differences in the ability to steer rhetoric toward fiscal conservatism.
- Robustness notes:
  - Figure A13 shows no meaningful difference when examining contemporaneous relationship between deficits and discourse.
  - Appendix C: extended discourse measures respond more robustly to inflation in advanced economies and to growth in emerging markets since the 1990s.
  - Time period heterogeneity may reflect larger variation of key economic indicators before the 1990s, larger sample for recent period, or better fiscal discourse measurement for later platforms.

### Fiscal events: public debt surges
- Debt surge definition (extension of Jalles and Medas (2022)):
  - SurgeCutoff_{i,t} = 1 if ∆D_{i,t,t−1} > max{μ_{∆D_i} + σ_{∆D_i}, Cutoff}, where cutoffs range from 1 to 5 percent of GDP.
  - First condition: change in debt-to-GDP across two consecutive years exceeds country mean plus one standard deviation (country-specific flexibility).
  - Second condition: change is at least as large as a common lower bound (1–5 percent of GDP).
  - Years since 2020 and with 1st-percentile debt growth are excluded in surge definition.
- Sample counts:
  - 434 country-years with debt surged using 1 percent of GDP common cutoff.
  - 336 cases using 5 percent of GDP cutoff.
  - Regression sample for surge analysis: 587 country-election year observations; 91 feature a contemporaneous debt surge; 76 a debt surge three years earlier; overall 219 country-election years associated to a debt surge in the same year or in any of the three previous years.
- Regression specification includes four debt surge indicators capturing occurrence up to three years before an election (Equation (3)).
- Empirical findings (Figure 10):
  - Panel A (Support for expansion): For surge episodes up to two calendar years before an election, all point estimates are negative; significantly so only for shocks two years ahead of the election date (about 1 percentage point less expansion discourse on average).
  - Panel B (Support for restraint): Discourse features reliably more support for restraint (up to 0.9 percentage points on average) in the first few years after a debt surge.
  - Point estimates slightly increase as the common cutoff is raised, suggesting larger adverse fiscal shocks elicit stronger political reactions toward fiscal moderation.
  - Interpretation: Large increases in public debt by a country’s historical standards tend to moderate support for expansion, but effects take time to materialize and can dissipate quickly; restraint increases more immediately in the years after a surge.

### Fiscal events: introduction of fiscal rules
- Focus: budget balance rules (operational limits on overall, primary, or cyclically adjusted balance) over 1985-2021 using IMF Fiscal Rules Dataset.
- Event-study regression (Equation (4)) uses staggered adoption of fiscal rules and measures relative election-cycle distances to adoption; country and election-year fixed effects included; standard errors clustered at country level.
- Sample considerations:
  - Only countries that adopted a fiscal rule over the 1985-2021 period are included for main analysis.
  - 48 of the 65 countries in the data adopt a fiscal rule over 1985-2021.
  - For each country, earliest fiscal rule adoption event is considered and assumed not interrupted thereafter.
- Empirical findings (Figure 11):
  - Both types of discourse evolve statistically similarly in treated countries and those yet to adopt a fiscal rule, though treated countries tend to exhibit weakly lower shares of expansion and especially restraint discourse on average.
  - Post-adoption dynamics diverge:
    - Panel A (Support for expansion): Virtually no impact on expansion discourse; mostly negative but small and insignificant coefficients.
    - Panel B (Support for restraint): On average, party platforms adopt a growing share of fiscal restraint content in elections following the institutional reform.
  - Interpretation: Formal fiscal frameworks do not significantly alter propensity to favor government spending, but they inject rhetoric with elements of fiscal discipline over the course of several election cycles.

*Source: wpiea2024194-print-pdf - 0.22 percentage points less expansion content and 0.10 percentage point more restraint discourse*

### Appendix D assesses the sensitivity of our results by relaxing the treatment and sample definition.

### Appendix D assesses the sensitivity of our results by relaxing the treatment and sample definition.

### Sensitivity checks for fiscal institutions and sample definition
- Figure A18 (Panels A and B): including the adoption of any type of fiscal rule as a possible fiscal event yields qualitatively similar results to the baseline.
- Figure A18 (Panels C and D): including countries that never adopt a fiscal rule produces weak but increasingly negative effects on expansion discourse, and smaller effects on restraint.
- Comparison with IMF Fiscal Council Dataset (2022): Figure A19 shows a similar direction of effects from the establishment of fiscal councils, either independently (Panels A and B) or as an interaction with fiscal rules (Panels C and D).

### Demand-side considerations: voter preferences (WVS/EVS evidence)
- Data sources and coverage:
  - World Value Survey (WVS) and European Value Survey (EVS), 1989-2022 (for countries in main analysis).
  - Survey items include preferences for more state ownership, government responsibility to provide for all citizens, income redistribution, taxing the rich and subsidizing the poor, state aid for unemployment, and making people’s incomes equal.
  - Original survey scale: 0 to 10 (higher values imply more support for government action or redistribution).
- Main findings (Figure 12):
  - Panel A: average trends in six survey items show no clear increasing trend in favor of government intervention comparable to that in the manifesto data.
  - Panel B: averaged responses by question type likewise show no comparable increasing trend.
  - Interpretation: no clear evidence that rising expansion discourse in party manifestos is paralleled by rising average public demand for government intervention in these survey items.
- Electoral response analysis (Figure 13):
  - Method: compute average vote share for parties with expansion and restraint discourse above/below:
    - country-specific median (left panels: Panels A and C)
    - country-by-decade median (right panels: Panels B and D)
  - Findings:
    - Left panels: vote shares increasingly accrue to parties above (below) their country’s historical median for expansion (restraint) discourse.
    - Right panels: virtually no trend when using country-by-decade medians; once account for country-decade trends, vote-share shifts attenuate and are broadly stable.
  - Interpretation: observed vote shifts likely reflect supply-side shifts in political platforms over time; attenuation when using country-by-decade medians suggests stability after accounting for evolving fiscal discourse norms.
- Role of structural change and inequality:
  - Structural transformations considered: globalization, skill-biased technological change, migration, ageing with generous pension systems.
  - Inequality measures examined (Figure 14):
    - Composite Gini index by Milanovic (2019) (1960-2015).
    - Wealth share ratio: top 10% vs bottom 50% from the World Inequality Database (1960-2022).
  - Findings:
    - Inequality rising in advanced economies since the 1990s; mixed trends for emerging market and developing economies.
    - No consistent within-country relationship between fiscal discourse and inequality, with the exception of a positive association between the wealth ratio and expansion discourse.
  - Conclusion: no direct evidence that voter preferences, as proxied by survey items or inequality trends, fully explain the observed long-term shifts in fiscal discourse; suggests politicians actively craft fiscal narratives.

### The aftermath of fiscal discourse changes — local projections framework
- Econometric setup (local projections, Jordà 2005):
  - Model equation (expanded for election and non-election years):
    y_{i,t+h} − y_{i,t−1} = α_i + δ_t + β^e_h ΔExp_{i,t} + β^r_h ΔRes_{i,t} + θ_h E_{i,t} + γ_h X_{i,t} + ε_{i,t}
  - Definitions:
    - y_{i,t}: macro-fiscal variable (fiscal deficit, components, government debt, GDP level).
    - ΔExp_{i,t}, ΔRes_{i,t}: change in mean share of expansion and restraint platform content across two consecutive elections (non-zero only in election years).
    - E_{i,t}: dummy for national election in year t.
    - X_{i,t}: controls including lagged change of outcome, current and lagged change in real GDP growth (except in GDP level regressions), indicators for government perimeter/data source switches. Both current and lagged versions of fiscal variables included.
  - Projection horizon: up to 9 years.
- Distribution of shocks (Figure A20):
  - Sample standard deviation of expansion shock: 4.57 percent (of a given party platform’s content).
  - Sample standard deviation of restraint shock: 2.21 percent.

### Impulse response findings — fiscal and macro outcomes (Figures 15 and 16)
- General presentation:
  - Impulse responses trace cumulative change in outcomes as time distance from an election-year fiscal discourse shock grows.
  - Responses reported for a one standard deviation shock in expansion (red) and restraint (blue); 68% and 90% country-clustered confidence intervals shown.
- Key quantitative findings (Figure 15):
  - Fiscal deficit response:
    - A one standard deviation increase in expansion discourse → 0.26 percent of GDP increase in the fiscal deficit by the end of the projection horizon.
    - A one standard deviation increase in restraint discourse → 0.37 percent of GDP reduction in the fiscal deficit (significant toward the end of the horizon).
  - Government debt:
    - Expansion shock: benign but muted response.
    - Restraint shock: initial increase in debt-to-GDP ratio in the near term (driven by temporary GDP decline), despite average deficit reduction.
  - GDP level:
    - Expansion shock: muted positive effects short-run; deficits grow over time as temporary boosts dissipate.
    - Restraint shock: GDP declines for several years following increase in restraint discourse.
  - Election-cycle dynamics (Panel D):
    - S-shaped impulse response of the budget deficit to an election-year dummy (θ_h), consistent with political budget cycle literature; deficit response peaks around election-associated horizons and turns negative in intervening periods.
- Components of the deficit (Figure 16):
  - Expansion shock:
    - Primary deficit: gradual increase following expansion shock (Panel A).
    - Primary spending: gradual increase drives the primary deficit (Panel B).
    - Government interest payments: relatively muted response (Panel C).
    - Revenues: no dominant role in primary deficit increase (Panel D).
    - Interpretation: expansion narratives translate more into spending initiatives than revenue cuts; discretionary spending drives the deficit buildup.
  - Restraint shock:
    - Primary deficit: declines, gaining significance earlier than for overall deficit (Panel A).
    - Revenues: positive point estimates throughout the projection horizon (Panel D), suggesting revenue measures are a primary channel.
    - Primary spending: expenditure cuts contribute later in the horizon (Panel B).
    - Government interest payments: short-run increase in interest payments-to-GDP ratio, consistent with output decline and stable/increasing debt (Panel C).
    - Interpretation: restraint narratives more readily translate into revenue measures first, with expenditure control effects appearing later; short-run output costs may temporarily raise debt ratios.

### Robustness and additional checks (Appendix references)
- Appendix E sensitivity checks (described in text):
  - Starting sample in 1990 (Figure A21): findings qualitatively similar but starker — deficit responses almost twice as large by year 8 and gain significance earlier, suggesting baseline reflects recent-decade dynamics.
  - Country-group heterogeneity (Figures A22 and A23):
    - Advanced economies: impulse patterns mirror longer sample; restraint shocks exhibit less pronounced effects on deficits and GDP in short term.
    - Emerging market and developing economies: restraint shocks lead to lower deficits over short- and medium-term with enduring output costs but shorter-lived upticks in debt-to-GDP; expansion shocks less consequential, with only a slow weak build-up in deficits.
  - Expanded definition of expansion discourse (Figure A24): including implied support for infrastructure and defense spending yields larger and earlier significant responses in deficits and output than baseline, indicating broader pro-intervention attitudes can translate into larger fiscal and economic responses.

### Conclusion (as summarized in supplied text)
- Main synthesis:
  - Electoral campaigns in the shadow of deteriorating fiscal conditions feature more conservative fiscal talk (particularly around debt surges and the adoption of constraining fiscal institutions).
  - In the long run, parties have increasingly competed over pro-government spending platforms with reduced focus on budget discipline.
  - Long-term political forces can dominate short-run strategic adjustments driven by deficits, debt surges, and institutional arrangements like fiscal rules.
- Limitations and future directions noted in the text:
  - Need to disentangle demand- and supply-side drivers of fiscal discourse and government size.
  - Measurement limitations: fiscal discourse observed only at elections rather than higher frequencies.
  - Need to establish channels through which discourse affects public finances, including the role of cabinets versus opposition and civil society.
  - Potential improvements: instrumental variables using external shocks to enhance identification.
  - Broader challenge: identify strategies to keep fiscal sustainability central in political discourse amid rising uncertainty about public finances.

*Source: wpiea2024194-print-pdf - Appendix D assesses the sensitivity of our results by relaxing the treatment and sample definition.*

### REFERENCES

### REFERENCES (wpiea2024194-print-pdf)

### Key bibliographic sources cited
- Aaskoven, Lasse.2020. “Do Fiscal Rules Reduce Political Polarization?” Comparative European Politics, 28: 630–658. 17
- Agyemang, Emma, and Chris Giles.2023. “Big government is back. How will we pay for it?” Financial Times, September 5, 2023. 2
- Alberto Alesina, Alberto, Carlo Favero, and Francesco Giavazzi.2019. Austerity: When It Works and When It Doesn’t. Princeton University Press. 6
- Alesina, Alberto.1987. “Macroeconomic Policy in a Two-Party System as a Repeated Game.” The Quarterly Journal of Economics, 102(3): 651–678. 5
- Alesina, Alberto, and Guido Tabellini.1990. “A Positive Theory of Fiscal Deficits and Government Debt.” The Review of Economic Studies, 57(3): 403–414. 5
- Arslanalp, Serkan, and Barry Eichengreen.2023. “Living with High Public Debt.” 2
- Ashok, Vivekinan, Ilyana Kuziemko, and Ebonya Washington.2015. “Support for Redistribution in an Age of Rising Inequality: New Stylized Facts and Some Tentative Explanations.” Brookings Papers on Economic Activity, 367–405. 6, 19
- Azzimonti, Marina, Marco Battaglini, and Stephen Coate.2016. “The costs and benefits of balanced budget rules: Lessons from a political economy model of fiscal policy.” Journal of Public Economics, 136: 45–61. 6
- Bertrand, Marianne, Esther Duflo, and Sendhil Mullainathan.2004. “How Much Should We Trust Differences-in-Differences Estimates?” The Quarterly Journal of Economics, 119(4): 249–275. 18
- Blanchard, Olivier.2023. “If markets are right about long real rates, public debt ratios will increase for some time. We must make sure that they do not explode.” Peterson Institute for International Economics, Realtime Economics, November 6, 2023. 2
- Borusyak, Kyrill, Xavier Jaravel, and Jann Spiess.2021. “Revisiting Event Study Designs: Robust and Efficient Estimation.” working paper. 18
- Bremer, Björn, and Reto Bürgisser.2023a. “Do Citizens Care about Government Debt? Evidence from Survey Experiments on Budgetary Priorities.” European Journal of Political Research, 62(1): 239–263. 6
- Bremer, Björn, and Reto Bürgisser.2023b. “Public Opinion on Welfare State Recalibration in Times of Austerity: Evidence from Survey Experiments.” Political Science Research and Methods, 11(1): 34–52. 6
- Brender, Adi, and Allan Drazen.2005. “Political Budget Cycles in New versus Established Democracies.” Journal of monetary Economics, 52(7): 1271–1295. 5, 25
- Brender, Adi, and Allan Drazen.2008. “How Do Budget Deficits and Economic Growth Affect Reelection Prospects? Evidence from a Large Panel of Countries.” American Economic Review, 98(5): 2203–20. 6
- Caselli, Francesca, and Philippe Wingender.2021. “Heterogeneous Effects of Fiscal Rules: The Maastricht Fiscal Criterion and the Counterfactual Distribution of Government Deficits.” European Economic Review, 136: 103748. 6
- Dalton, Russell J.2008. “The quantity and the quality of party systems: Party system polarization, its measurement, and its consequences.” Comparative Political Studies, 41(7): 899–920. 12, 43
- Dovis, Alessandro, Mikhail Golosov, and Ali Shourideh.2016. “Political Economy of Sovereign Debt: A Theory of Cycles of Populism and Austerity.” 5
- Funke, Manuel, Moritz Schularick, and Christoph Trebesch.2023. “Populist Leaders and the Economy.” American Economic Review, 113(12): 3249–3288. 6, 24
- Halac, Marina, and Pierre Yared.2022. “A Theory of Fiscal Responsibility and Irresponsibility.” National Bureau of Economic Research. 5
- Hatchondo, Juan Carlos, Leonardo Martinez, and Francisco Roch.2022. “Fiscal rules and the sovereign default premium.” American Economic Journal: Macroeconomics, 14(4): 244–273. 6
- Hyde, Susan D., and Nikolay Marinov.2012. “Which Elections Can Be Lost?” Political Analysis, 20(2): 191–201. 15
- IMF.2023. “Global Debt Is Returning to its Rising Trend.” https://www.imf.org/en/Blogs/Articles/2023/09/13/global-debt-is-returning-to-its-rising-trend. 2
- Jalles, João Tovar, and Paulo A. Medas.2022. “Economic Growth After Debt Surges.” IMF Working Paper No. 2022/159. 16
- Jordà, Òscar.2005. “Estimation and Inference of Impulse Responses by Local Projections.” American Economic Review, 95(1): 161–182. 4, 23
- Klomp, Jeroen, and Jakob de Haan.2013. “Political Budget Cycles and Election Outcomes.” Public Choice, 157(1/2): 245–267. 6
- Lagerborg, Andresa, Evi Pappa, and Morten O. Ravn.2023. “Sentimental Business Cycles.” The Review of Economic Studies, 90(3): 1358–1393. 5
- Laver, Michael, and Ian Budge.1992. Party Policy and Government Coalitions. New York: St. Martins Press. 43
- Lehmann, Pola, et al.2023. “The Manifesto Data Collection. Manifesto Project (MRG/CMP/MARPOR). Version 2023a.” 2, 7, 30
- Mauro, Paulo, Rafael Romeu, Ariel Binder, and Asad Zaman.2015. “A Modern History of Fiscal Prudence and Profligacy.” Journal of Monetary Economics, 76: 55–70. 12
- Milanovic, Branko.2019. “All the Ginis Dataset.” Version February 2019. Retrieved at https://stonecenter.gc.cuny.edu/research/all-the-ginis-alg-dataset-version-february-2019/. 20, 22
- Müller, Andreas, Kjetil Storesletten, and Fabrizio Zilibotti.2016. “The Political Color of Fiscal Responsibility.” Journal of the European Economic Association, 14(1): 252–302. 5
- Nordhaus, William D.1975. “The Political Business Cycle.” The Review of Economic Studies, 42(2): 169–190. 5
- Peltzman, Sam.1992. “Voters as Fiscal Conservatives.” The Quarterly Journal of Economics, 107(2): 327–361. 6
- Persson, Torsten, and Guido Tabellini.2003. “Do Electoral Cycles Differ Across Political Systems?” IGIER Working Papers No. 232. 5, 25
- Persson, Torsten, and Lars E. O. Svensson.1989. “Why a Stubborn Conservative Would Run a Deficit: Policy with Time-Inconsistent Preferences.” The Quarterly Journal of Economics, 104(2): 325–345. 5
- Rogoff, K.1990. “Equilibrium Political Budget Cycles.” American Economic Review, 80(1): 21. 5
- Rogoff, Kenneth.2022. “Emerging Market Sovereign Debt in the Aftermath of the Pandemic.” Journal of Economic Perspectives, 36(4): 147–166. 2
- Romer, Christina D., and David H. Romer.2004. “Choosing the Federal Reserve Chair: Lessons from History.” Journal of Economic Perspectives, 18(1): 129–162. 2, 5, 27
- Shiller, Robert J.2019. Narrative Economics: How Stories Go Viral and Drive Major Economic Events. Princeton University Press. 5
- Shi, Min, and Jakob Svensson.2006. “Political Budget Cycles: Do They Differ across Countries and Why?” Journal of Public Economics, 90(8-9): 1367–1389. 5, 25
- Singhal, Monica.2021. “Quantifying Preferences for Redistribution.” Manuscript. 6
- Volkens, Andrea, et al.2013. “Understanding and Validating the Left-Right Scale (RILE).” In Mapping Policy Preferences From Texts: Statistical Solutions for Manifesto Analysts. Oxford University Press. 43
- Yared, Pierre.2019. “Rising Government Debt: Causes and Solutions for a Decades-Old Trend.” Journal of Economic Perspectives, 33(2): 115–140. 2

### Appendix A — Manifesto Project data: methods and key definitions
- Data source: Manifesto Project (MP), version 2023a (Lehmann et al., 2023).
- MP objective: codify and quantify topic and policy content of electoral manifestos; platforms are authoritative pre-election party documents.
- Inclusion criterion: parties that gain at least one seat in a given parliamentary lower-house democratic election; presidential platforms generally reflect presidential campaigns.
- Coding process:
  - Coders: native-language political science scholars or students who pass English-language manifesto annotation training and test.
  - Text segmented into quasi-sentences (policy statements), each assigned a three-digit MP code.
  - Codes aggregated into platform-share variables by topic; codes assigned to seven broad domains: external relations, freedom and democracy, political system, economy, welfare and quality of life, fabric of society, social groups (Figure A3).
- Baseline fiscal discourse proxies constructed from MP variables:
  1. Baseline proxy for government expansion support (“expansion”): percentage share of platform statements calling for increases in:
     - per504: Welfare state expansion
     - per506: Education expansion
     - per409: Keynesian demand management
  2. Proxy for fiscal restraint support (“restraint”): percentage share of platform statements calling for:
     - per414: Economic orthodoxy (includes reduction of budget deficits)
     - per505: Welfare state limitation
     - per507: Education limitation
- Extensions: broader "expansion" definitions include per104 (Military: Positive), per411 (Technology and Infrastructure: Positive), per402 (Incentives: Positive), per501 (Environmental protection), per416.2 (Sustainability: Positive) for alternative specifications (Figures A6, A16, A17, A24).

### Appendix A — Coverage and descriptive statistics
- TABLE A1: MP coverage by economy, 1960-2022 — country-level # Elections, # Manifestos, Main Range, First Election (sample entries include Albania, Argentina, Armenia, Australia, Austria, etc.; United States: AE, # Elections 1632, Main Range 1960-2020, First Election 1920).
- FIGURE A2: Coverage evolution by country group, 1920-2022 (panels: available party platforms; covered country-elections).
- FIGURE A3: Domain evolution by country group (Panels A: AEs, 1960-2022; B: EMDEs, 1988-2022).
- TABLE A2 — FISCAL DISCOURSE: SUMMARY STATISTICS (PLATFORM%)
  - 1960-2022:
    - Expansion (platform): N 4,504; Mean 12.7; Median 12.1; SD 7.90; Min 0; Max 68.4
    - Expansion (country-year): N 721; Mean 12.9; Median 12.5; SD 5.3; Min 1.5; Max 39.5
    - Restraint (platform): N 4,504; Mean 2.5; Median 1.1; SD 3.80; Min 0; Max 42.6
    - Restraint (country-year): N 721; Mean 2.5; Median 2.0; SD 2.20; Min 0; Max 21.0
  - 1990-2022:
    - Expansion (platform): N 3,245; Mean 13.7; Median 13.0; SD 7.90; Min 0; Max 68.4
    - Expansion (country-year): N 492; Mean 14.0; Median 13.4; SD 5.3; Min 1.5; Max 39.5
    - Restraint (platform): N 3,245; Mean 2.2; Median 1.1; SD 3.20; Min 0; Max 25.9
    - Restraint (country-year): N 492; Mean 2.2; Median 1.7; SD 1.80; Min 0; Max 10.5
- TABLE A3 — FISCAL DISCOURSE: MEAN PLATFORM SHARE CHANGE ACROSS TIME PERIODS (IN P.P. AND %)
  - AEs, 1960-63 to 2016-19: Expansion Start,% 11.09; ∆ p.p. +5.56; ∆,% +50; Restraint Start,% 3.42; ∆ p.p. -1.59; ∆,% -47
  - AEs, 1960-63 to 2020-22: Expansion Start,% 11.09; ∆ p.p. +7.62; ∆,% +69; Restraint Start,% 3.42; ∆ p.p. -2.10; ∆,% -62
  - (Additional period comparisons and EMDE rows reported similarly in table.)

### Appendix B — Measuring fiscal and non-fiscal platform polarization: construction and measures
- Constructed relative pro-restraint position (RR) per party platform:
  - RR_{i,j,e} = FD_{Restraint_{i,j,e}} − FD_{Expansion_{i,j,e}}
  - FD_* are platform shares of the identified MP variables (restraint: per414, per505, per507; expansion: per504, per506, per409).
- Fiscal polarization at country-election:
  - Fiscal polarization_{i,e} = Σ_j w_{i,j,e} · (RR_{i,j,e} − RR_{i,e})^2
    - w_{i,j,e} = vote_{i,j,e} / Σ_j vote_{i,j,e} (party vote share among covered parties)
    - RR_{i,e} = Σ_j w_{i,j,e} · RR_{i,j,e} (weighted average relative restraint position)
- Relative fiscal polarization:
  - Relative fiscal polarization_{i,e} = Fiscal polarization_{i,e} / Σ_j w_{i,j,e} · (FD_{Restraint_{i,j,e}} + FD_{Expansion_{i,j,e}})
  - Interpretation: compares extent of party disagreement on relative pro-restraint positions to the average platform share devoted to the fiscal topics. The relative measure can exceed 1 (or 100%).
- Comparison with RILE-based polarization:
  - Non-fiscal RILE-based polarization constructed by adjusting the standard RILE (26 items) to remove overlapping fiscal components (notably per414, per505, per504, per506 adjustments).
- Empirical notes:
  - Figure A11: global maps of fiscal polarization (Panel A: 0-1 scale) and relative fiscal polarization (Panel B: percent of mean platform share), 2010-2022.
  - Figure A12: Conditional relationship between fiscal and non-fiscal polarization; residualized binscatter with country fixed effects shows higher non-fiscal polarization associates with higher fiscal disagreement; relationship about 80% steeper among EMDEs.

### Appendix C — Conditions at elections and fiscal discourse: robustness and alternative measures (summary of empirical approach and findings)
- Empirical strategy: regressions of fiscal discourse on lagged macro-fiscal variables with country, data source, and government perimeter fixed effects; winsorization applied to GDP growth (5-95th percentile) and CPI inflation (95th percentile); debt growth measured as mean yearly change in public debt-to-GDP over four years.
- Figures and robustness checks:
  - FIGURE A13: Fiscal discourse and same-year fiscal deficit, 1960-2022 (Panels for expansion and restraint).
  - FIGURE A14: Fiscal discourse and lagged fiscal deficit, predetermined elections, 1960-2022.
  - FIGURE A15: Macro-fiscal conditions at elections and discourse, predetermined elections — coefficients plotted for one-year lag of macro-fiscal variables across two time periods and across country groups; 90% confidence intervals with country-clustered standard errors.
  - FIGURE A16 and FIGURE A17: Alternative expansion definitions as outcomes (progressively comprehensive definitions: baseline social; add defense and infrastructure; add firm incentives; add environment) — period and AE vs EMDE comparisons; estimation follows Equation (1).
- Key methodological choices preserved:
  - Use of alternative expansion concept types (Type 1, Type 2, Type 3) to assess robustness (Figures A16–A17).
  - Winsorization rules and debt growth definition as noted above.

### Appendix D — Fiscal events: event-study evidence and institutional interactions
- Data sources:
  - Fiscal rules: IMF Fiscal Rules Dataset (2022).
  - Fiscal councils: IMF Fiscal Council Dataset (2022).
- Event-study specifications:
  - Estimation follows Equation (4), time dimension expressed in election cycles (not calendar years) relative to treatment.
  - FIGURE A18: Event studies including all fiscal rule types and untreated countries, 1985-2021 (Panels A–D: expansion and restraint outcomes; Panels C–D: first adoption of a budget balance rule, allowing adopters and never-adopters in sample).
  - FIGURE A19: Fiscal discourse response to fiscal councils, 1985-2021 (Panels for expansion and restraint; panels combining rules and councils). Only treated countries included in top panels; bottom panels restrict to countries ever adopting a fiscal rule for comparison.
- Inference:
  - Confidence intervals based on country-clustered standard errors.
  - Average election cycle length referenced from FIGURE A8 (mean cycle length about 3.7 years).

*Source: wpiea2024194-print-pdf - REFERENCES (appendices A–D, manifest data, coverage tables, figures, and methodological notes)*

### 1985. Panels A and B plot the evolution of fiscal discourse after the recorded introduction of a fiscal council. In Pane

### wpiea2024194-print-pdf - 1985. Panels A and B plot the evolution of fiscal discourse after the recorded introduction of a fiscal council. In Pane

### Event studies: fiscal councils and fiscal rules
- Panels A and B plot the evolution of fiscal discourse after the recorded introduction of a fiscal council.
- Panels C and D consider fiscal rules for the event studies: budget balance rules, which include any rule with operational limits to the overall, primary, or cyclically adjusted balance.
- For the post-rule introduction period, Panels C and D plot the discourse path both with and without an interaction term for the contemporaneous presence of a rule and a fiscal council.
- Confidence intervals are based on country-clustered standard errors.

### Distribution of fiscal discourse shocks in election years (Figure A20)
- Definition: A fiscal discourse shock is the change in the average share of expansion or restraint discourse in a country’s party platforms across two consecutive election cycles.
- Shocks included come from a local projection regression sample in a model of fiscal deficit as a function of restraint and expansion shocks and other controls as defined in Equation (5), where the outcome difference is computed at the first available time horizon (h= 0).
- Panels A and B: distribution histograms of the two types of shocks in two periods:
  - 1960-2022 regression sample standard deviation for the restraint discourse shock: 2.12% of a party platform’s content.
  - 1990-2022 regression sample standard deviation for the restraint discourse shock: 1.99% of a party platform’s content.
  - 1960-2022 regression sample standard deviation for the expansion discourse shock: 4.57% of a party platform’s content.
  - 1990-2022 regression sample standard deviation for the expansion discourse shock: 4.74% of a party platform’s content.
- Panels C and D: binscatter relationship between ∆Expansion and ∆Restraint in each period’s regression sample.
- Each panel reports the underlying country-fixed effect regression’s coefficient and country-clustered standard error with conventional significance markers.
- Baseline definition of expansion discourse is used in all cases.

### Local projections: responses to a discourse shock (Figures A21–A24)
- Estimation follows Equation (5). Data sources: Manifesto Project, PFMH, and WEO.
- Figures present impulse responses associated to a one standard deviation shock in restraint and expansion discourse.
- Presentation conventions across panels:
  - Bold blue line: estimated impulse responses for one standard deviation expansion shock.
  - Bold red line: estimated impulse responses for one standard deviation restraint shock.
  - Shaded areas (for expansion shocks) and dashed lines (for restraint) reflect the 68% and 90% country-clustered confidence intervals estimated for each impulse response.
- Figures and panels covered:
  - Figure A21 (1990-2022): Panel A: Fiscal deficits; Panel B: Government debt; Panel C: GDP level; Panel D: Primary deficits.
  - Figure A22 (AES sample): Panel A: Fiscal deficits; Panel B: Government debt; Panel C: GDP level; Panel D: Primary deficits.
  - Figure A23 (EMDES sample): Panel A: Fiscal deficits; Panel B: Government debt; Panel C: GDP level; Panel D: Primary deficits.
  - Figure A24 (Alternative definition): Panel A: Fiscal deficits; Panel B: Government debt; Panel C: GDP level; Panel D: Primary deficits.

*Working Paper No. WP/2024/194*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024194-print-pdf.pdf_