## 2.1  Survey data collection and sample

## Source details

**Canonical URL:** [2.1  Survey data collection and sample](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025197-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025197-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025197-print-pdf.pdf.json)

---

### Data collection and sample design
- Survey administered between April 2024 and May 2024 using the survey company YouGov.
- Respondents: at least 18 years of age, surveyed in their native language, selected from pre-profiled online panelists.
- YouGov panel access: over 22 million registered members across more than 40 countries; respondents who fully complete the survey receive compensation of varying amounts and forms, including cash and gift cards.
- Final sample: 27,202 respondents (ranging from 2,013 to 2,338 per country).
- Countries surveyed: Argentina, Australia, Brazil, Canada, France, Germany, Hungary, Italy, Japan, the Netherlands, Poland, the UK, and the US.
- Samples selected to be nationally representative by age, gender, and region; additional variables used in some countries (socioeconomic class in the UK; race/education in the US).
- Sampling weights constructed to ensure representativeness of the final sample.

### Quality procedures and cleaning
- Sampling with replacement to minimize non-response bias; each respondent could only take the survey once.
- Questionnaire administered individually and focused solely on study-related questions to reduce cognitive fatigue.
- Median completion time was 19 minutes (interquartile range: 20 minutes).
- Only respondents who completed the survey were included; exclusions: respondents finishing in less than 1 minute (speeders) and inattentive respondents who failed attention checks.
- Debt variables winsorized by country at the 98 percent level to reduce impact of outliers.
- Subsample analyses use both qualitative and quantitative assessments of debt where available.

### Questionnaire modules (overview)
- Background information:
  - Gender, age, income, education level, employment status, marital status, number of children, residence, financial constraints, savings, and portfolio choices.
  - Political leanings, social class, receipt of pensions or need-based government transfers, and self-identified net beneficiary status.
  - Main sources of economic news, engagement with economic issues, trust in government, and views on the government’s role in public service provision.
- Knowledge and beliefs module:
  - Factual knowledge on relationships among government spending, tax revenues, budget deficits, and debt.
  - Randomized framing: half the respondents in each country asked about tax revenue–deficit relationships in the context of increased taxes; the other half about decreasing taxes. Similar randomizations for spending–deficit and deficit–debt relationships.
  - Questions on perceptions of country total debt as a share of GDP and expectations for debt over the next five years.
  - Questions on whether current levels of taxes, spending, and debt are high or low.
- Treatment (information experiment):
  - Information experiment designed to evaluate how informing respondents about the level and trajectory of debt influences beliefs about fiscal policy direction and respondent behavior.
  - Detailed descriptions of information treatments provided in Section 6 (not reproduced here).
- Policy views and own behavior:
  - After randomized information intervention, respondents asked probabilistic expectations that the government will increase taxes or cut government spending, time horizon for adjustments, and beliefs about incidence across tax and spending categories.
- Beliefs regarding efficacy and mechanisms:
  - Views on impact of public debt on household, whether higher public debt today will be paid by higher taxes and/or lower government spending in the future, distributional impact of higher debt, and whether inflation will have to increase to lower debt.

*Source: wpiea2025197-print-pdf - 2.1  Survey data collection and sample (IMF).*

### 3  Descriptive statistics: knowledge and prior beliefs

#### Knowledge of fiscal variables — key findings
- Overall knowledge of factual relationships among taxes, spending, deficits, and debt: described as weak with significant cross-country variation.
- Country examples:
  - Netherlands: nearly 60 percent accurately understand the relationship between changes in tax revenues and budget deficits.
  - Argentina: 39 percent accurately understand that relationship.
- Asymmetries in understanding:
  - 58 percent recognize that higher government spending leads to higher budget deficits.
  - 42 percent believe that lower spending reduces budget deficits.
  - Nearly half understand that lower tax revenues increase budget deficits; only 43 percent correctly identify the relationship between tax revenue increases and budget deficits.
- Knowledge index construction:
  - Transform underlying knowledge variables into z-scores; average z-scores across items; standardize that average; regress knowledge index on socioeconomic characteristics with country fixed effects.
- Correlates of greater accuracy:
  - Possessing financial assets is significantly correlated with more accurate knowledge in most countries.
  - College degree and higher income are significantly associated with accuracy (with country variation).
  - Effect of age generally positive with heterogeneity across countries.
  - Reliance on traditional media sources and valuing being informed about economic policies associated with greater accuracy.
- Machine learning (random forests):
  - Primary predictors of knowledge of fiscal relationships: owning financial assets and being aged 55 and above.

#### Beliefs regarding debt levels — key findings
- Perceptions of current and five-year-ahead debt:
  - Respondents systematically under-estimate public debt in countries with debt-to-GDP ratios exceeding 100 percent of GDP in 2023 (US, UK, Canada, Japan, France, and Italy).
  - Extent of under- or over-estimation influenced by initial country debt levels; people more likely to underestimate debt in high-debt countries.
- Correlates of perceived debt-to-GDP ratios:
  - Respondents with financial assets and older respondents more likely to report higher estimates; results statistically significant.
  - Misperception defined as absolute deviation of prior beliefs from the true value as of 2023; errors lower among older respondents, respondents receiving old age benefits, and respondents who value being informed about economic policy.
- Expectations of future debt:
  - Strong positive association between perceptions of current debt and expectations of future debt.
  - In countries where debt is projected to rise, 74 percent of respondents underestimating future debt versus 50 percent in countries where debt is not expected to rise.
- Qualitative priors:
  - 72 percent of respondents perceive debt to be somewhat high or very high.
  - A higher share of respondents believe tax levels are high compared to government spending.
  - Nearly half the respondents in Argentina believe that debt will decrease in five years.
- Quantitative–qualitative correlation:
  - Medians: 'Very high' median debt of 75 percent of GDP; 'Somewhat high' median 45 percent of GDP; 'Not high' median 30 percent of GDP.
  - Qualitative beliefs used as proxy for perceived debt levels since half the sample did not provide a quantitative estimate.

*Source: wpiea2025197-print-pdf - 3  Descriptive Statistics (IMF).*

### 4  Policy beliefs — stylized facts and incidence

#### Likelihood and time horizon of fiscal adjustment
- Perceived likelihood:
  - Majority perceive a high probability of tax increases (50-80 percent).
  - Only one-third report a high likelihood of spending cuts.
  - US and Brazil: approximately 30 percent assign a probability of less than 25 percent to spending cuts.
  - Argentina: 24 percent assign a probability of greater than 75 percent to spending cuts.
- Perceived time-horizon:
  - Most respondents who anticipate tax increases believe these will occur within two years.
  - Perceived time horizon for spending cuts varies across countries.
  - Argentina and Italy: 77 and 69 percent, respectively, believe spending cuts are imminent.
  - Belief that spending cuts are imminent is much less prevalent in Brazil and Japan.

#### Perceptions across groups and incidence of adjustment
- Correlates of expectations:
  - Ownership of financial assets correlated with higher probability of expecting spending cuts; no strong relationship with expectation of tax increases.
  - Women tend to expect tax increases more than men and are less likely to assign high probability to spending cuts.
  - Older respondents generally less likely to expect spending cuts; older respondents show higher expectations for tax increases in the U.S., Germany, and Argentina.
  - Recipients of need-based transfers and respondents who self-report staying informed about economic policy show heightened expectations for both tax increases and spending cuts.
  - Self-identified net tax payers exhibit higher expectations of tax increases and lower likelihood assigned to spending cuts.
  - Right-wing orientation generally associated with higher expectations of tax increases and lower expectations of spending cuts (exceptions: Argentina, Hungary, and Japan).
- Incidence among respondents reporting over a 50 percent probability:
  - Over 70 percent in all countries predict higher taxes on middle-income households.
  - Expected spending cuts by category: Social programs 58 percent; Public infrastructure 61 percent; Defense spending 38 percent.
- Socioeconomic patterns in incidence expectations:
  - Higher-income respondents more likely to expect tax increases on wealthy households, wealth/estate taxes, and taxes on large corporations.
  - Prime-age respondents more likely to anticipate sales tax increases and taxes affecting middle-income households and small businesses.
  - Recipients of need-based transfers more likely to anticipate cuts to education spending; lower-income respondents more likely to expect cuts to education than higher-income respondents.
  - Individuals aged 55+ and the retired are less likely to anticipate cuts to pensions, but recipients of old age benefits are more likely to expect pension cuts.
- Overall interpretation:
  - Individuals generally expect the burden of tax increases and spending cuts to fall on themselves.

*Source: wpiea2025197-print-pdf - 4  Policy Beliefs (IMF).*

### 4.2  Role of beliefs and knowledge — predictors and lifetime exposure

#### Predictors of expectations about fiscal policy
- Beliefs that taxes, spending, and debt levels are high, and expectations of future debt increases, correlate with higher expectations of tax increases.
- Greater trust in government associated with higher expectations of expenditure cuts and lower expectation of tax increases.
- Random forest analysis: key predictors of tax increase likelihood include perceptions of government debt levels and trajectory, prior beliefs about tax levels, trust in government, and knowledge of fiscal relationships.
- Country fixed effects are primary predictors of expectations regarding spending cuts, highlighting country-specific influences.

#### Measure of lifetime exposure to fiscal consolidation
- Lifetime exposure Ai defined as:
  - Ai = (Σ_{k=1}^{age_i−1} w_i(k,λ) C_{age_i−k}) / (Σ_{k=1}^{age_i−1} w_i(k,λ)), where w_i(k,λ) = ((age_i − k)/age_i)^λ and C_{(age_i−k)} indicates a consolidation episode in the k-th year prior to the survey.
- Recency bias implemented with λ = 1 so weights decrease linearly from the year before the survey to zero at age 20; consolidation episodes experienced prior to age 20 are weighted as 0.
- Robustness: parallel measures for experiences of inflation and sovereign debt; alternative measure is simple sum of consolidation episodes.

#### Past consolidation experiences and belief outcomes
- Respondents with greater experience of consolidation are less likely to perceive existing debt, spending, or tax levels as high (effects generally robust to contemporaneous macro controls but lose significance with country fixed effects).
- A one standard deviation increase in exposure to fiscal consolidation predicts a 0.08 standard deviation increase in the belief that debt will rise in the future (a 2.8 percent increase relative to the control group mean).
- Greater lifetime exposure to consolidation associated with higher expectations of future tax increases (insignificant with country fixed effects) and significantly less likely to predict expectations of spending cuts.
- Greater exposure to past consolidations correlates with:
  - Less optimism about government’s ability to reduce or stabilize debt.
  - A one standard deviation increase in past exposure increases the belief that debt will be harmful for future taxpayers by 0.04 standard deviation (1 percent increase relative to control mean).
  - A one standard deviation increase increases belief that higher debt today will have to be paid off in the future by 0.03 standard deviations (1 percent increase relative to control mean).
  - Greater inclination to believe inflation may rise in the future due to elevated public debt.
- Experienced fiscal consolidation is strongly and negatively associated with trust in government across and within countries.
- Effects driven by respondents who experienced substantial spending-based consolidations; tax-based consolidation effects are small and statistically insignificant.

*Source: wpiea2025197-print-pdf - 4.2  Role of beliefs and knowledge (IMF).*

### 6  Causal effects of information about government debt (treatments and channels)

#### Information treatments (three-layered)
- Treatment 1: informed about actual debt-to-GDP ratio in their country in 2023, shown a figure contrasting 2023 with pre-pandemic average, with a short script explaining debt-to-GDP and its interpretation.
- Treatment 2: Treatment 1 information plus equation: "Debt in the future = Debt issued today + interest paid on today’s debt + government spending today − taxes collected by the government today."
- Treatment 3: Treatments 1 and 2 plus explicit statement that "Economic forecasters predict that government debt in the US will remain high relative to historical standards in the coming years." (applied analogously in other countries’ scripts)

#### Debt-sample separation and main treatment effects
- Debt-stable sub-sample: Netherlands, Germany, Poland, and Hungary (2023 debt decreased or remained stable relative to historical average).
- Debt-increased sub-sample: remaining nine countries (2023 debt increased relative to historical average).
- Treatment effects on expectations:
  - Debt-stable sample: all three treatments lower expectations of tax increases by 0.6 - 0.11 standard deviations relative to control (1.7 - 3.2 percent decrease relative to control group mean).
  - Debt-increased sample: all three treatments increase expectations of spending cuts by 0.04 standard deviations relative to control (1.7 percent increase relative to control group mean).
  - In the debt-stable sample, no consistent statistically significant effect on spending cuts.
- Comparing treatments:
  - No significant differences between Treatments 1 and 2, even after accounting for prior knowledge.
  - Treatment 3 increases likelihood of expecting tax increases relative to those only informed of current debt levels; driven by respondents who learn their country’s future debt levels are expected to remain high.
- Interpretation:
  - In countries failing to control debt, agents more prone to anticipate cuts to benefits.
  - In countries successful at keeping debt under control, taxation appears the more relevant margin of adjustment.
  - Differences may reflect agents’ perceptions about whether fiscal issues stem from excessive spending or historical patterns where taxation has been the adjustment margin.

*Source: wpiea2025197-print-pdf - 6  Causal effect of information about government debt (IMF).*

### 6.3  Channels of treatment effects: heterogeneous debt priors

#### Treatment effects by qualitative priors
- Debt-stable sample:
  - Treatment effects reduce expectations of tax increases by 0.09 - 0.14 standard deviations for respondents who perceive debt as "high" or "somewhat high" (2.6 - 4 percent decrease from control group mean).
  - Effect insignificant among respondents with "low" debt priors.
  - Interpretation: treatment operates through individuals with high initial qualitative debt priors who lower tax increase expectations when informed that debt has been stable.
- Debt-increased sample:
  - Respondents who perceive debt as "not very high" increase expectations of spending cuts by 0.06 - 0.07 standard deviations (2.6 - 3.1 percent increase relative to control mean).
  - Treatment has no significant impact on respondents with initial priors that debt is "very high".
  - Interpretation: treatment operates through respondents who learn debt-to-GDP is much higher than their priors, raising probability assigned to spending cuts.

#### Role of numerical priors and misperception direction
- Treatment effects primarily driven by respondents who report numerical debt priors; respondents without numerical priors tend to exhibit insignificant treatment effects.
- Treatment effects strongest among:
  - Debt-stable sample respondents who initially overestimate their country’s debt levels.
  - Debt-increased sample respondents who underestimate their debt levels.
- Overall interpretation:
  - People update expectations of fiscal adjustment when informed about the level and trajectory of debt.
  - Direction of adjustment depends on prior error: overestimators revise down expectations of tax increases in stable-debt countries; underestimators revise up expectations of spending cuts in rising-debt countries.
- Mechanism interpretation:
  - Because effects are driven by those surprised by information (low prior accuracy), results interpreted as Bayesian-style updating rather than emotional priming.

*Source: wpiea2025197-print-pdf - 6.3  Channels of treatment effects (IMF).*

### 10 & 13  Spending cuts, tax increases, regressions and robustness (selected empirical findings)

#### Data sources, model specification, and inference
- Sources: World Economic Outlook database, official government sources, and IMF staff calculations based on IMF-YouGov Survey.
- Figures reference control-group respondents for perceived incidence measures.
- Confidence intervals reported are 95% confidence intervals computed using robust standard errors.
- Linear probability model regression coefficients reported for likelihood and time horizon of fiscal adjustment (separate analyses for spending cuts and tax increases).
- Models include full set of socioeconomic controls and country fixed effects; some specifications additionally include prior knowledge and beliefs.

#### Sample sizes and representativeness (exact reported sample sizes)
- United Kingdom NA 2,338; United States NA 2,013; Poland NA 2,018; Hungary NA 2,040; Canada NA 2,065; Netherlands NA 2,131; Italy NA 2,064; Germany NA 2,182; France NA 2,101; Brazil NA 2,127; Argentina NA 2,026; Australia NA 2,036; Japan NA 2,061.

#### Key regression magnitudes (selected exact coefficients from Tables)
- Taxes (Observations 16,889; R-squareds by column reported):
  - Female: -0.02 (Taxes on middle income), 0.05*** (Taxes on high income), 0.10*** (Taxes on corporations), 0.08*** (Sales taxes).
  - High income: -0.10*** (Taxes on middle income), 0.15*** (Taxes on high income), -0.12*** (Taxes on all households), 0.13*** (Taxes on corporations), -0.09*** (Sales taxes).
  - Political orientation Right: 0.10***, 0.24***, 0.10***, 0.20***, 0.15***, 0.22***, 0.12*** across columns (1)–(7).
  - Pays more in taxes than receives in benefits: 0.16*** (Taxes on middle income), 0.08*** (Taxes on all households), 0.08*** (Taxes on small businesses), -0.07*** (Taxes on corporations), 0.09*** (Sales taxes).
- Spending cuts (Observations 9,570; R-squareds by column reported):
  - Female: 0.06*** (Pensions), 0.04* (Defense).
  - Age 55+: -0.23*** (Education), -0.07* (Pensions), -0.12*** (Defense), -0.22*** (Infrastructure), -0.10*** (Climate).
  - Political orientation Right: -0.10*** (Education), -0.10*** (Social programs), 0.14*** (Defense), 0.04** (Infrastructure).
  - Self-reported awareness economic issues: 0.11*** (Education), 0.07*** (Pensions), 0.09*** (Social programs), 0.08*** (Infrastructure), 0.10*** (Climate).

#### Past consolidation and perceptions/expectations (selected exact coefficients)
- Current perceptions regressions (Observations 25,960):
  - Selected aggregated effects of Past experience of fiscal consolidation reported include -0.061***, -0.014, -0.082***, -0.018, -0.042***, 0.027*, 0.083***, 0.081***, -0.104***, -0.126*** across columns.
- Expectations regressions (Observations 25,960):
  - Past experience of fiscal consolidation coefficients: 0.021*** (Expectation of tax increases, column 1), 0.007 (col 2), 0.012 (col 3), -0.080*** (col 4).
- Respondent beliefs and mechanisms (selected coefficients from Table 7):
  - Past experience associated with negative coefficients on "Debt will stabilize/decrease" and "Respondent will be better off": -0.102***, -0.072***, -0.088***, -0.052***.
  - Positive coefficients on "Higher debt will have to be paid off", "Debt harmful for future taxpayers", "Inflation may have to be higher", "Trust in government": 0.063***, 0.029*, 0.037***, 0.038**, 0.117***, 0.039** (exact arrangement preserved in Table 7).

#### Treatment predictability and heterogeneous treatment effects (selected exact results)
- Predictability tests (Table 8): Control group F-statistic 0.77 p-value 0.67; Treatment 1 F-statistic 1.16 p-value 0.31; Treatment 2 F-statistic 1.28 p-value 0.23; Treatment 3 F-statistic 0.78 p-value 0.66 (tests against control and across groups reported).
- Treatment effects by debt-sample (Table 9; Observations 8,371 and 18,831):
  - Treatment 1 on Expectations of tax increases: -0.11*** (Debt stable), -0.02 (Debt increased).
  - Treatment 1 on Expectations of spending cuts: -0.02 (Debt stable), 0.04** (Debt increased).
  - Treatment 2: -0.06** (taxes, Debt stable); -0.05* (spending, Debt stable); 0.04** (spending, Debt increased).
  - Treatment 3: -0.07** (taxes, Debt stable); 0.04** (spending, Debt increased).
- Treatment effects by qualitative priors (Table 10; Observations across cells 2,611; 5,760; 11,900; 6,931):
  - Example: Treatment 1: -0.06 (Debt not high, Debt stable), -0.14*** (Debt high, Debt increased) on Expectations of tax increases.
  - Treatment 1 on Expectations of spending cuts: -0.04 (Debt not high, Debt stable), 0.07*** (Debt high, Debt increased).
- Interaction with past consolidation experience (Table 11):
  - Received information treatment coefficient on Expectations of tax increases: -0.075** (Debt stable), 0.003 (Debt increased).
  - Above median experience of fiscal consolidation: 0.125** (Debt stable, expectations of tax increases), -0.094*** (Debt increased, expectations of spending cuts).
  - Interaction Received information treatment * Above median experience: -0.028 (Taxes, Debt stable), 0.082** (Spending, Debt increased).

#### Heterogeneity by knowledge and priors (selected exact findings)
- Knowledge index interactions (Table A.14): Treatment 2 main effect 0.05 (Debt stable) and 0.05 (Debt increased) on Expectations of tax increases; Knowledge index 0.02 (Debt stable) and 0.07*** (Debt increased); interactions not significant.
- Treatment effects among respondents with numerical priors (Table A.16):
  - Treatment 1 on Expectations of tax increases: -0.14*** (Has numerical priors, Debt stable), -0.09** (No numerical priors, Debt stable).
- Treatment effects by over/underestimate of prior (Table A.17):
  - Treatment 1 on Expectations of tax increases: -0.17** (Overestimate, Debt stable), -0.13** (Underestimate, Debt stable).

#### Knowledge and numerical prior regressions (selected exact coefficients)
- Knowledge index correlates (Table A.2):
  - Female: -0.20*** (All) and negative, significant across countries (examples: -0.26*** UK, -0.23*** US).
  - College education: 0.21*** (All) and positive, significant across countries.
  - Savings/checking account: 0.24*** (All); country examples include 0.47*** (NLD).
- Numerical prior bias for debt (Table A.3):
  - Outcome Current Debt-to-GDP: Constant 73.77***; Female coefficient -7.66***; Stocks/shares/funds/bonds/retirement products 5.99***; Observations 12,964; R-squared 0.05.
  - Deviation of Belief from Actual (Absolute): Constant 61.06***; Age 55+ coefficient -7.34**; Observations 12,964; R-squared 0.18.

#### Fiscal consolidation episodes (Annex examples)
- United Kingdom: 1969 - 1971, 1977 - 1982, 1994 - 1999, 2010 - 2015.
- United States: 1969 - 1971, 1976 - 1977, 1978, 1980 - 1981, 1985 - 1986, 1988, 1990 - 1998, 2011 - 2013.
- Japan: 1972 - 1974, 1979 - 1983, 1997 - 1998, 2003 - 2007, 2014 - 2015, 2019 - 2020.

#### Notes on methodology and inference
- Outcome variables in expectation regressions standardized using the mean and standard deviation of the control group.
- Regressions include socio-demographic characteristics and country fixed effects; standard errors clustered at the country level unless noted.
- Significance notation: *** p<0.01, ** p<0.05, * p<0.1.

*Source: IMF staff calculations based on IMF-YouGov Survey and World Economic Outlook database; content unit: wpiea2025197-print-pdf.*

### 2.1  Survey data collection and sample

### 2.1  Survey data collection and sample

### Data collection
- Survey administered between April 2024 and May 2024 using the survey company YouGov.
- Respondents: at least 18 years of age, surveyed in their native language, selected from pre-profiled online panelists.
- YouGov panel note: access to a panel of over 22 million registered members across more than 40 countries; respondents who fully complete the survey receive compensation of varying amounts and forms, including cash and gift cards.
- Quality procedures:
  - Sampling with replacement to minimize non-response bias; each respondent could only take the survey once.
  - Questionnaire administered individually and focused solely on study-related questions to reduce cognitive fatigue.
  - Median completion time was 19 minutes (interquartile range: 20 minutes).
  - Only respondents who completed the survey were included; exclusions: respondents finishing in less than 1 minute (speeders) and inattentive respondents who failed attention checks.
- Final sample: 27,202 respondents (ranging from 2,013 to 2,338 per country).

### Sample
- Countries surveyed: Argentina, Australia, Brazil, Canada, France, Germany, Hungary, Italy, Japan, the Netherlands, Poland, the UK, and the US.
- Samples selected to be nationally representative by age, gender, and region.
- Additional variables used in some countries to ensure representativeness: socioeconomic class (UK) and race/education (US).
- Sampling weights constructed to ensure representativeness of the final sample.

### Questionnaire
- Annex B contains the full questionnaire (note in source).
- Background information collected:
  - Gender, age, income, education level, employment status, marital status, number of children, residence, financial constraints, savings, and portfolio choices.
  - Political leanings, social class, receipt of pensions or need-based government transfers (e.g., childcare, unemployment support), and self-identified net beneficiary status (whether paying more in taxes than receiving in benefits).
  - Main sources of economic news, engagement with economic issues, trust in government, and views on the government’s role in public service provision.
- Knowledge and beliefs module:
  - Factual knowledge questions on relationships among government spending, tax revenues, budget deficits, and debt.
  - Randomized framing: half the respondents in each country asked about tax revenue–deficit relationships in the context of increased taxes; the other half about decreasing taxes. Similar randomizations for spending–deficit and deficit–debt relationships.
  - Questions on perceptions of country total debt as a share of GDP and expectations for debt over the next five years.
  - Questions on whether current levels of taxes, spending, and debt are high or low, and expectations for government debt over the next five years.
- Treatment:
  - Information experiment designed to evaluate how informing respondents about the level and trajectory of debt influences beliefs about fiscal policy direction and respondent behavior.
  - Detailed descriptions of information treatments provided in Section 6 (note in source).
- Policy views and own behavior:
  - After randomized information intervention, respondents asked probabilistic expectations that the government will increase taxes or cut government spending.
  - Elicited perceptions about time horizon for tax increases and/or spending cuts and beliefs about incidence (likelihood to impact specific tax and spending categories).
- Beliefs regarding efficacy and mechanisms:
  - Views on impact of public debt on household, whether higher public debt today will be paid by higher taxes and/or lower government spending in the future, distributional impact of higher debt (who will win or lose), and whether inflation will have to increase to lower debt in the future.

### Data on past macroeconomic experiences
- Primary source for fiscal consolidations: Adler et al. (2024) compiling fiscal consolidation data for 17 OECD economies (1978–2020) and Latin American and Caribbean economies (1989–2020).
- Extended temporal coverage and addressed data gaps using Escolano et al. (2018), Alesina and Ardagna (2010), and relevant IMF country reports.
- Fiscal consolidation years for each country reported in Annex Table 1 (note in source).
- Recent data on debt, real GDP growth and inflation sourced from the IMF’s World Economic Outlook.
- Historical inflation data sources: OECD Consumer Prices Indices COICOP 1999 database; national data sources via Haver Analytics; Australian Bureau of Statistics; Instituto Nacional de Estadistica y Censos de la Republica Argentina.
- Historical data on public debt retrieved from the International Monetary Fund’s database on Public Finances in Modern History.

### Final sample and data cleaning notes
- Debt variables winsorized by country at the 98 percent level to reduce impact of outliers.
- Subsample analyses use both qualitative and quantitative assessments of debt where available.

*Source: wpiea2025197-print-pdf - 2.1  Survey data collection and sample (IMF).*

### 3  Descriptive Statistics: Knowledge and Prior Beliefs

#### 3.1  Knowledge of fiscal variables — key findings
- Overall knowledge of factual relationships among taxes, spending, deficits, and debt: described as weak with significant cross-country variation.
- Country examples:
  - Netherlands: nearly 60 percent accurately understand the relationship between changes in tax revenues and budget deficits.
  - Argentina: 39 percent accurately understand that relationship.
- Asymmetry in understanding:
  - 58 percent recognize that higher government spending leads to higher budget deficits.
  - 42 percent believe that lower spending reduces budget deficits.
  - Nearly half understand that lower tax revenues increase budget deficits; only 43 percent correctly identify the relationship between tax revenue increases and budget deficits.
- Knowledge index construction (methodology):
  - Transform underlying knowledge variables into z-scores (subtract sample mean, divide by sample standard deviation).
  - Average z-scores across items, then standardize that average (subtract mean, divide by standard deviation).
  - Regress knowledge index on socioeconomic characteristics with country fixed effects (Figure 3.1 / Annex Table A2).
- Correlates of greater accuracy:
  - Possessing financial assets (savings/checking accounts, investments in stocks and bonds, or home ownership) is significantly correlated with more accurate knowledge in most countries.
  - College degree and higher income are significantly associated with accuracy (country variation exists).
  - Effect of age generally positive, with heterogeneity across countries.
  - Reliance on traditional media sources and valuing being informed about economic policies associated with greater accuracy.
- Machine learning (random forests) results:
  - Primary predictors of knowledge of fiscal relationships: owning financial assets and being aged 55 and above.
  - Random forest feature importance used to evaluate predictors’ contribution to prediction accuracy.

#### 3.2  Beliefs regarding debt levels — key findings
- Perceptions of current and five-year-ahead debt:
  - Respondents systematically under-estimate public debt in countries with debt-to-GDP ratios exceeding 100 percent of GDP in 2023 (US, UK, Canada, Japan, France, and Italy).
  - Extent of under- or over-estimation influenced by initial country debt levels; people more likely to underestimate debt in high-debt countries.
- Correlates of perceived debt-to-GDP ratios:
  - Respondents with financial assets and older respondents more likely to report higher estimates of the debt-to-GDP ratio; results statistically significant (Annex Table A3).
  - Misperception defined as absolute deviation of prior beliefs from the true value as of 2023; errors lower among older respondents, respondents receiving old age benefits, and respondents who value being informed about economic policy.
- Expectations of future debt:
  - Strong positive association between perceptions of current debt and expectations of future debt.
  - Respondents more likely to have inaccurate perceptions of future debt levels relative to current debt levels.
  - In countries where debt is projected to rise, 74 percent of respondents underestimating future debt versus 50 percent in countries where debt is not expected to rise.
- Qualitative priors:
  - 72 percent of respondents perceive debt to be somewhat high or very high.
  - A higher share of respondents believe tax levels are high compared to government spending.
  - Nearly half the respondents in Argentina believe that debt will decrease in five years.
- Quantitative–qualitative correlation:
  - Median and mean quantitative estimates of debt are strictly increasing for respondents reporting debt as 'Somewhat high' or 'Very high'.
  - Example medians: respondents reporting 'Very high' median debt of 75 percent of GDP; 'Somewhat high' median 45 percent of GDP; not high median 30 percent of GDP.
  - Subjective qualitative beliefs used as a proxy for perceptions of debt levels, particularly because half the sample did not provide a quantitative estimate.

*Source: wpiea2025197-print-pdf - 2.1  Survey data collection and sample (IMF).*

### 4  Policy Beliefs

#### 4.1  Likelihood of fiscal adjustment — stylized facts
- Analysis based on control group sample to avoid treatment effects; regressions control for information treatments for full-sample analyses.
- Perceived likelihood of adjustment:
  - Majority of respondents perceive a high probability of tax increases (50-80 percent).
  - Only one-third report a high likelihood of spending cuts.
  - US and Brazil: approximately 30 percent of respondents assign a probability of less than 25 percent to spending cuts.
  - Argentina: 24 percent of respondents assign a probability of greater than 75 percent to spending cuts.
- Perceived time-horizon:
  - Most respondents who anticipate tax increases believe these will occur within two years.
  - Perceived time horizon for spending cuts varies more across countries.
  - Argentina and Italy: 77 and 69 percent, respectively, believe spending cuts are imminent.
  - Belief that spending cuts are imminent is much less prevalent in Brazil and Japan.

#### 4.1.1  Perceptions across different groups
- Method: ordinal responses converted into z-scores relative to control group mean and standard deviation; regressions include socioeconomic characteristics and country fixed effects (Figure 10; Annex Tables A4–A7).
- Correlates of expectations:
  - Ownership of financial assets correlated with a higher probability of expecting spending cuts; no significant relationship with expectation of tax increases.
  - Gender:
    - Women tend to expect tax increases more than men.
    - Women are less likely to assign a high probability to spending cuts.
    - Women believe tax adjustments will occur in the near term.
  - Age:
    - Significance of age varies across policy views and countries.
    - Older respondents are generally less likely to expect spending cuts (except in the Netherlands where older respondents are more likely to expect cuts).
    - Older respondents demonstrate higher expectations for tax increases in the U.S., Germany, and Argentina.
    - Prime-age and older respondents are less likely to believe policy changes are imminent.
  - Recipients of need-based transfers and respondents who self-report staying informed about economic policy show heightened expectations for both tax increases and spending cuts.
  - Self-identified net tax payers (paying more in taxes than receiving benefits) exhibit higher expectations of tax increases and lower likelihood assigned to spending cuts.
  - Partisan differences:
    - Right-wing orientation generally associated with higher expectations of tax increases and lower expectations of spending cuts (exceptions: Argentina, Hungary, and Japan).

#### 4.1.2  Incidence of adjustment
- Among respondents reporting over a 50 percent probability of tax increases or spending cuts:
  - Over 70 percent in all countries predict higher taxes on middle-income households.
  - Expected spending cuts by category:
    - Social programs: 58 percent expect cuts.
    - Public infrastructure: 61 percent expect cuts.
    - Defense spending: 38 percent expect cuts.
- Correlations with socioeconomic characteristics:
  - Higher-income respondents:
    - More likely to expect tax increases on wealthy households, wealth/estate taxes, and taxes on large corporations (controlling for income and asset ownership).
    - Less likely to expect tax increases on middle-income households, taxes on small businesses, and sales taxes.
  - Prime-age respondents:
    - More likely to anticipate sales tax increases and taxes affecting middle-income households and small businesses.
  - Expectation of expenditure cuts tends to align with direct benefit receipt:
    - Recipients of need-based transfers more likely to anticipate cuts to education spending.
    - Lower-income respondents more likely to expect cuts to education spending relative to higher-income respondents.
    - Individuals aged 55 and above and the retired are less likely to anticipate cuts to pensions, but respondents who report receiving old age benefits or retirement benefits are more likely to expect cuts to pensions.
  - Higher-income respondents more likely to assign a lower likelihood to reductions in social programs than lower-income counterparts.
- Overall interpretation:
  - Individuals generally expect the burden of tax increases and spending cuts to fall on themselves, consistent with prior evidence that people perceive themselves as more directly affected and disadvantaged by policy changes.

*Source: wpiea2025197-print-pdf - 2.1  Survey data collection and sample (IMF).*

### 4.2  Role of beliefs and knowledge

### 4.2  Role of beliefs and knowledge

### Key predictors of expectations about fiscal policy
- Beliefs that taxes, spending, and debt levels are high, along with expectations of future debt increases, correlate with higher expectations of tax increases.
- Greater trust in government is associated with higher expectations of expenditure cuts and lower expectation of tax increases.
- Random forest analysis identifies the key predictors of tax increase likelihood as:
  - perceptions of government debt levels and trajectory,
  - prior beliefs about tax levels,
  - trust in government,
  - knowledge of fiscal relationships.
- Country fixed effects are the primary predictors of expectations regarding spending cuts, implying strong country-specific influences (social contract, institutions, composition and trajectory of social spending).
- Trust in government and perceptions of government debt and spending levels and knowledge are more important predictors of expectations of spending cuts than demographic and socioeconomic characteristics.

### Measure of lifetime exposure to fiscal consolidation
- Lifetime exposure measure Ai defined as:
  - Ai = (Σ_{k=1}^{age_i−1} w_i(k,λ) C_{age_i−k}) / (Σ_{k=1}^{age_i−1} w_i(k,λ)), where w_i(k,λ) = ((age_i − k)/age_i)^λ and C_{(age_i−k)} indicates a consolidation episode in the k-th year prior to the survey.
- Recency bias implemented with λ = 1 so weights decrease linearly from the year before the survey to zero at age 20; consolidation episodes experienced prior to age 20 are weighted as 0.
- Robustness: parallel measures for experiences of inflation and sovereign debt (historical cross-country data on inflation level and public debt as percentage of GDP); alternative measure is simple sum of consolidation episodes.

### Past consolidation experiences and prior beliefs (Section 5.1)
- Respondents with greater experience of consolidation are less likely to perceive existing debt, spending, or tax levels in their country as high.
- These results generally remain robust after controlling for contemporaneous macroeconomic conditions (inflation, real GDP per capita, average debt over the past five years), but lose significance when country fixed effects are included.
- Past experiences shape beliefs about future debt outcomes:
  - A one standard deviation increase in exposure to fiscal consolidation predicts a 0.08 standard deviation increase in the belief that debt will rise in the future, representing a 2.8 percent increase relative to the control group mean.
- Expectations of future tax increases and spending cuts:
  - Greater lifetime exposure to fiscal consolidation is associated with higher expectations of future tax increases, but this finding is insignificant after including country fixed effects.
  - Greater lifetime exposure to consolidation is significantly less likely to predict expectations of spending cuts in respondents’ country.
  - Results remain robust to inclusion of existing policy priors, beliefs, and knowledge; greater experience of consolidation continues to predict a lower likelihood of respondents expecting spending cuts.

### Beliefs about efficacy of fiscal adjustment and underlying mechanisms (Section 5.2)
- Among respondents reporting over a 50 percent probability of tax increases or spending cuts, those with a history of fiscal consolidation demonstrate significantly lower optimism about the government’s ability to reduce or stabilize debt levels.
- Greater exposure to past fiscal consolidations correlates with:
  - Less optimism about anticipated economic policy changes (belief they will not be better off because of these changes).
  - Greater likelihood of perceiving current high debt levels as detrimental to future taxpayers:
    - A one standard deviation increase in past exposure to consolidation increases belief that debt will be harmful for future taxpayers by 0.04 standard deviation, or a 1 percent increase relative to the control group mean.
  - Stronger belief that current high debt levels will necessitate future tax increases or spending cuts:
    - A one standard deviation increase in past exposure to consolidation increases belief that higher debt today will have to be paid off in the future by 0.03 standard deviations, or a 1 percent increase relative to the control group mean.
  - Greater inclination to believe that inflation may rise in the future due to elevated public debt levels today.
- Experienced fiscal consolidation is strongly and negatively associated with trust in government both across and within countries.
- Effects vary by type of consolidation: impacts on beliefs about efficacy and mechanisms are driven entirely by respondents who experienced substantial spending-based consolidations; effects of tax-based consolidations are small and statistically insignificant across outcomes examined.

### Causal effect of information about government debt (Section 6)
- Three randomized information treatments (layered):
  1. Treatment 1: Random quarter informed about actual debt-to-GDP ratio in their country in 2023, shown a figure contrasting 2023 with pre-pandemic average, with a short script explaining debt-to-GDP and its interpretation.
  2. Treatment 2: Treatment 1 information plus an equation relating future and current debt: "Debt in the future = Debt issued today + interest paid on today’s debt + government spending today − taxes collected by the government today."
  3. Treatment 3: Treatments 1 and 2 plus explicit statement that "Economic forecasters predict that government debt in the US will remain high relative to historical standards in the coming years." (applied analogously in other countries’ scripts)
- Analysis separates countries by debt trajectory information:
  - Debt-stable sub-sample: countries where 2023 debt decreased or remained stable relative to historical average (Netherlands, Germany, Poland, and Hungary).
  - Debt-increased sub-sample: countries where 2023 debt increased relative to historical average (the remaining nine countries).
- Treatment effects on expectations of fiscal adjustment:
  - In the debt-stable country sample, all three information treatments lower expectations of tax increases by 0.6 - 0.11 standard deviations relative to the control group (1.7 - 3.2 percent decrease relative to the control group mean).
  - In the debt-increased country sample, all three information treatments increase respondents’ expectations of spending cuts by 0.04 standard deviations relative to the control group (1.7 percent increase relative to the control group mean).
  - In the debt-stable sample, no consistent statistically significant effect on spending cuts was found.
- Comparing treatments:
  - No significant differences between treatments 1 and 2, even after accounting for prior knowledge.
  - Treatment 3 (information on forecasters’ predictions) increases likelihood of expecting tax increases relative to those only informed of current debt levels; this effect is driven by respondents who learn their country’s future debt levels are expected to remain high.
- Interpretation:
  - In countries failing to control debt, agents may be more prone to anticipate cuts to benefits.
  - In countries more successful at keeping debt under control, taxation appears the more relevant margin of adjustment.
  - Differences may reflect agents’ perceptions about whether fiscal issues stem from excessive spending or historical patterns where taxation has been the adjustment margin.

*Source: wpiea2025197-print-pdf - 4.2  Role of beliefs and knowledge*

### 6.3  Channels of treatment effects: heterogeneous debt priors

### 6.3  Channels of treatment effects: heterogeneous debt priors

### Treatment effects by qualitative debt priors
- In the debt-stable country sample:
  - Treatment effects reduce expectations of tax increases by 0.09 - 0.14 standard deviations relative to the control group for respondents who perceive debt as "high" or "somewhat high".
  - This corresponds to a 2.6 - 4 percent decrease from the control group mean.
  - The effect is insignificant among respondents with "low" debt priors.
  - Interpretation: the treatment operates through individuals with high initial qualitative debt priors who lower their tax increase expectations when informed that debt has been stable.

- In the debt-increased country sample:
  - People in treatment groups with qualitative priors that debt is "not very high" increase expectations of spending cuts by 0.06 - 0.07 standard deviations relative to the control group.
  - This corresponds to a 2.6 - 3.1 percent increase relative to the control group mean.
  - The treatment has no significant impact on respondents with initial priors that debt is "very high".
  - Interpretation: the treatment operates through respondents who learn that the debt-to-GDP ratio is much higher than their priors, leading them to assign a higher probability to spending cuts.

### Role of numerical priors and direction of misperception
- Annex Table A16:
  - Shows that in both the debt-stable and debt-increased samples, the treatment effects reported above are primarily driven by respondents who report numerical debt priors.
  - Respondents without numerical priors tend to exhibit insignificant treatment effects.
- Annex Table A17:
  - Reveals treatment effects are strongest among:
    - Respondents in the debt-stable sample who initially overestimate their country’s debt levels.
    - Respondents in debt-increasing countries who underestimate their debt levels.
- Overall interpretation:
  - People update expectations of fiscal adjustment when informed about the level and trajectory of debt.
  - The direction of adjustment depends on prior error: overestimators revise down expectations of tax increases in stable-debt countries; underestimators revise up expectations of spending cuts in rising-debt countries.

### Differential responses across fiscal instruments and priors
- Countries with stable or decreasing debt:
  - Information reduces expectations of tax increases and spending cuts.
  - Result is driven by respondents with high qualitative and quantitative initial debt priors who revise their expectations of tax increases down.
- Countries with high or rising debt:
  - Information increases expectations of spending cuts, but does not change expectations regarding tax increases (which start with high baseline priors).
  - This pattern is consistent with Roth et al. (2022) for the US.
  - The effect is primarily driven by people with low initial qualitative and quantitative debt priors (i.e., surprised by high debt).
- Implication about mechanisms:
  - Because treatment effects are driven by individuals with lower prior belief accuracy (i.e., those who are surprised), the results are interpreted as true Bayesian-style updating of beliefs rather than emotional responses or priming.

*Source: 6.3 Channels of treatment effects: heterogeneous debt priors, wpiea2025197-print-pdf*

### 10.1 Spending Cuts

### 10.1 Spending Cuts

### Data sources and sample
- Sources: World Economic Outlook database, official government sources, and IMF staff calculations based on IMF-YouGov Survey.
- Figures reference control-group respondents for perceived incidence measures.
- Confidence intervals reported are 95% confidence intervals computed using robust standard errors.

### Empirical approach and model specification
- Linear probability model regression coefficients are reported for:
  - Dependent variables: likelihood of fiscal adjustment and time horizon of fiscal adjustment.
  - Separate analyses for spending cuts (Panel 1) and tax increases (Panel 2).
- Models include:
  - The full set of socioeconomic controls.
  - Country fixed effects.
  - In the analysis shown in Figure 12, prior knowledge and beliefs are additionally included as regressors.

### Perceived incidence across policy categories
- Figure 11: Perceived Incidence across Tax Types and Spending Categories (Percent of control group respondents)
  - Top panel question: "What do you think is the likelihood that the government will cut the following categories of government spending?"
  - Bottom panel question: "What do you think is the likelihood that the government will raise the following categories of taxes?"
  - Shares reported reflect responses for the control group only.

### Fiscal adjustment and prior beliefs
- Figure 12: Fiscal Adjustment and Prior Beliefs (Regression Coefficients & 95% Confidence Intervals)
  - Presents regression coefficients and 95% confidence intervals for:
    - Panel 1: Spending Cuts.
    - Panel 2: Tax Increases.
  - Models control for socioeconomic characteristics, country fixed effects, and prior knowledge and beliefs (where indicated).

### Predictors of expectations
- Figure 13: Importance of Predictors of Expectations of Fiscal Policy Changes
  - Reports relative importance of predictors used to explain expectations about fiscal policy changes (specific predictor magnitudes are presented in the figure).

*Source: IMF staff calculations based on IMF-YouGov Survey and World Economic Outlook database; content from the referenced chapter section.*

### 13.1 Spending Cuts13.2 Tax Increases

### 13.1 Spending Cuts13.2 Tax Increases

### Sample representativeness
- Sample sizes: United Kingdom NA 2,338; United States NA 2,013; Poland NA 2,018; Hungary NA 2,040; Canada NA 2,065; Netherlands NA 2,131; Italy NA 2,064; Germany NA 2,182; France NA 2,101; Brazil NA 2,127; Argentina NA 2,026; Australia NA 2,036; Japan NA 2,061. (Tables 1–2)
- Key sample demographic proportions (examples preserved exactly as reported):
  - Female: 0.51 (United Kingdom sample), 0.51 (United States sample), 0.52 (Poland sample), 0.52 (Italy sample), 0.52 (Germany sample), 0.52 (France sample), 0.52 (Japan sample). (Tables 1–2)
  - Tertiary education: 0.51 (United Kingdom population/sample), 0.29 (United States sample), 0.50 (Poland sample), 0.54 (Hungary sample), 0.34 (Canada sample), 0.41 (Netherlands sample), 0.29 (Italy sample). (Tables 1–2)

### Socioeconomic correlates of expectations (Taxes)
- Regression sample size: Observations 16,889 for all seven tax-category regressions. R-squared reported by column:
  - Taxes on middle income: R-squared 0.03
  - Taxes on high income: R-squared 0.06
  - Taxes on all households: R-squared 0.02
  - Wealth/estate taxes: R-squared 0.06
  - Taxes on small businesses: R-squared 0.07
  - Taxes on corporations: R-squared 0.04
  - Sales taxes: R-squared 0.03
- Selected coefficient magnitudes and significance (exact values and significance markers preserved):
  - Female: -0.02 (Taxes on middle income), 0.05*** (Taxes on high income), 0.00 (Taxes on all households), 0.05*** (Wealth/estate taxes), 0.01 (Taxes on small businesses), 0.10*** (Taxes on corporations), 0.08*** (Sales taxes). (Table 3)
  - High income: -0.10*** (Taxes on middle income), 0.15*** (Taxes on high income), -0.12*** (Taxes on all households), 0.08*** (Wealth/estate taxes), -0.09*** (Taxes on small businesses), 0.13*** (Taxes on corporations), -0.09*** (Sales taxes). (Table 3)
  - Political orientation: Right: 0.10***, 0.24***, 0.10***, 0.20***, 0.15***, 0.22***, 0.12*** across columns (1)–(7). (Table 3)
  - Pays more in taxes than receives in benefits: 0.16*** (Taxes on middle income), -0.03 (Taxes on high income), 0.08*** (Taxes on all households), 0.00 (Wealth/estate taxes), 0.08*** (Taxes on small businesses), -0.07*** (Taxes on corporations), 0.09*** (Sales taxes). (Table 3)

### Socioeconomic correlates of expectations (Spending cuts)
- Regression sample size: Observations 9,570 for spending-category regressions. R-squared by column:
  - Education: 0.08; Pensions: 0.05; Social programs: 0.10; Defense: 0.07; Infrastructure: 0.06; Climate change-related needs: 0.05
- Selected coefficient magnitudes and significance:
  - Female: -0.01 (Education), 0.06*** (Pensions), -0.01 (Social programs), 0.04* (Defense), -0.00 (Infrastructure), -0.01 (Climate). (Table 4)
  - Age 55+: -0.23*** (Education), -0.07* (Pensions), -0.00 (Social programs), -0.12*** (Defense), -0.22*** (Infrastructure), -0.10*** (Climate). (Table 4)
  - Political orientation: Right: -0.10*** (Education), -0.03 (Pensions), -0.10*** (Social programs), 0.14*** (Defense), 0.04** (Infrastructure), -0.03 (Climate). (Table 4)
  - Self-reported awareness economic issues: 0.11*** (Education), 0.07*** (Pensions), 0.09*** (Social programs), 0.02 (Defense), 0.08*** (Infrastructure), 0.10*** (Climate). (Table 4)

### Relationship between past experiences of fiscal consolidation and perceptions/expectations
- Current perceptions regressions (Table 5):
  - Observations 25,960 (columns vary with country fixed effects). Selected reported effects of Past experience of fiscal consolidation (exactly as table shows, aggregated across columns): examples include -0.061***, -0.014, -0.082***, -0.018, -0.042***, 0.027*, 0.083***, 0.081***, -0.104***, -0.126*** (Table 5).
  - R-squared reported across columns: 0.082, 0.142, 0.079, 0.148, 0.057, 0.127, 0.018, 0.070, 0.049, 0.070 (Table 5).
- Expectations regressions (Table 6):
  - Observations 25,960 across four specifications.
  - Past experience of fiscal consolidation coefficients: 0.021*** (Expectation of tax increases, column 1), 0.007 (col 2), 0.012 (col 3), -0.080*** (col 4). R-squared 0.03, 0.041, 0.016, 0.100. (Table 6)
- Respondent beliefs and mechanisms (Table 7):
  - Observations vary by column (e.g., 17,128; 21,879; 21,458; 22,208; 19,381).
  - Past experience of fiscal consolidation associated with negative coefficients on beliefs like "Debt will stabilize/decrease" and "Respondent will be better off": -0.102***, -0.072***, -0.088***, -0.052*** respectively; and positive coefficients on "Higher debt will have to be paid off", "Debt harmful for future taxpayers", "Inflation may have to be higher", "Trust in government": 0.063***, 0.029*, 0.037***, 0.038**, 0.117***, 0.039** (exact arrangement preserved from Table 7).

### Treatment and information experiments (predictability, heterogeneous treatment effects)
- Predictability tests (Table 8):
  - F-statistics and p-values for joint significance: Control group F-statistic 0.77 p-value 0.67; Treatment 1 F-statistic 1.16 p-value 0.31 (test against all other groups), test against control group 0.92 p-value 0.52; Treatment 2 F-statistic 1.28 p-value 0.23, test against control group 0.98 p-value 0.47; Treatment 3 F-statistic 0.78 p-value 0.66, test against control group 0.76 p-value 0.68. (Table 8)
- Treatment effects by debt-sample (Table 9):
  - Debt stable sample (Observations 8,371) vs Debt increased sample (Observations 18,831).
  - Treatment 1 on Expectations of tax increases: -0.11*** (Debt stable), -0.02 (Debt increased). Treatment 1 on Expectations of spending cuts: -0.02 (Debt stable), 0.04** (Debt increased). (Table 9)
  - Treatment 2: -0.06** (taxes, Debt stable), -0.03 (taxes, Debt increased); -0.05* (spending, Debt stable), 0.04** (spending, Debt increased). (Table 9)
  - Treatment 3: -0.07** (taxes, Debt stable), 0.01 (taxes, Debt increased); -0.04 (spending, Debt stable), 0.04** (spending, Debt increased). (Table 9)
  - R-squareds: taxes columns 0.03/0.04; spending columns 0.09/0.10. (Table 9)
- Treatment effects by qualitative priors (Table 10):
  - Subsamples and columns report Treatment 1/2/3 effects conditional on respondents’ qualitative priors (Debt not high / Debt high / Debt not very high / Debt very high) with Observations 2,611; 5,760; 11,900; 6,931 across cells. Examples:
    - Treatment 1: -0.06 (Debt not high, Debt stable sample), -0.14*** (Debt high, Debt increased sample) on Expectations of tax increases. (Table 10)
    - Treatment 1 on Expectations of spending cuts: -0.04 (Debt not high, Debt stable), 0.07*** (Debt high, Debt increased). (Table 10)
  - R-squared values by column reported (examples): 0.04, 0.04, 0.03, 0.06, 0.10, 0.09, 0.06, 0.17. (Table 10)
- Interaction of treatment with past consolidation experience (Table 11):
  - Debt stable sample (Observations 8,371) and Debt increased sample (Observations 18,831).
  - Received information treatment coefficient on Expectations of tax increases: -0.075** (Debt stable), 0.003 (Debt increased).
  - Above median experience of fiscal consolidation: 0.125** (Debt stable, expectations of tax increases), -0.094*** (Debt increased, expectations of spending cuts).
  - Interaction Received information treatment * Above median experience: -0.028 (Taxes, Debt stable), 0.082** (Spending, Debt increased). R-squared 0.041 and 0.105. (Table 11)

### Heterogeneity by knowledge, numerical priors, and over/underestimation
- Knowledge index interactions (Table A.14):
  - Treatment 2 main effect: 0.05 (Debt stable sample), 0.05 (Debt increased sample) on Expectations of tax increases; Knowledge index 0.02 (Debt stable) and 0.07*** (Debt increased) reported. Interaction Treatment 2 * Knowledge index: -0.02 and -0.01 (not significant). Observations 4,188 (tax columns) and 9,446 (spending columns). R-squared reported 0.04/0.04/0.10/0.11. (Table A.14)
- Treatment effects among respondents with numerical priors (Table A.16):
  - Debt stable and debt increased samples split by Has numerical priors vs No numerical priors.
  - Treatment 1 on Expectations of tax increases: -0.14*** (Has numerical priors, Debt stable), -0.09** (No numerical priors, Debt stable); Treatment 2: -0.10** (Has numerical priors) etc. Observations across cells e.g., 3,881; 4,490; 9,083; 9,748. (Table A.16)
- Treatment effects by over/underestimate of prior (Table A.17):
  - For Overestimate vs Underestimate subgroups, Treatment 1 on Expectations of tax increases: -0.17** (Overestimate, Debt stable), -0.13** (Underestimate, Debt stable). Observations 1,682 (Overestimate) and 2,199 (Underestimate) in Debt stable sample; 2,912 and 6,171 in Debt increased sample. (Table A.17)

### Knowledge, priors, and correlations with characteristics
- Knowledge index correlates (Table A.2, selected exact coefficients):
  - Female: -0.20*** (All), -0.26*** (UK), -0.23*** (US), -0.25*** (POL), -0.19*** (HUN), -0.24*** (CAN), -0.13*** (NLD), -0.22*** (ITA), -0.22*** (DEU), -0.20*** (FRA), -0.16*** (BRA), -0.18*** (ARG), -0.24*** (AUS), -0.17*** (JPN). (Table A.2)
  - College education: 0.21*** (All) and positive, significant across countries (examples: 0.26*** UK, 0.22*** US, 0.26*** POL). (Table A.2)
  - Savings/checking account: 0.24*** (All), 0.25*** (UK), 0.25*** (US), 0.28*** (POL), 0.16*** (HUN), 0.23*** (CAN), 0.47*** (NLD), 0.20*** (ITA), 0.31*** (DEU), 0.21*** (FRA), 0.11** (BRA), 0.21*** (ARG), 0.15*** (AUS), 0.21*** (JPN). (Table A.2)
- Numerical prior bias for debt (Table A.3):
  - Outcome variable Current Debt-to-GDP: Constant 73.77***; Female coefficient -7.66***; Stocks/shares/funds/bonds/retirement products 5.99***; Observations 12,964; R-squared 0.05. (Table A.3)
  - Deviation of Belief from Actual (Absolute): Constant 61.06***; Age 55+ coefficient -7.34**; Residence -3.01*; Pays more in taxes than receives in benefits -2.26; Observations 12,964; R-squared 0.18. (Table A.3)

### Perceived effectiveness, trust, and prior beliefs (select findings)
- Perceived likelihood that debt will stabilize/decrease (Table A.8 Panel A):
  - Female 0.08***; Age 25-55 -0.07***; Age 55+ -0.07***; Savings/checking account 0.06***; Correct knowledge on relationships between spending, taxes, deficit, and debt 0.15***; Has numerical estimate for debt 0.03*. Observations 17,937; R-squared 0.18. (Table A.8)
- Beliefs linking priors to expectations (Table A.9):
  - Belief: Debt is very high associated with higher expectations of tax increases: 0.38*** (column 1). Correct knowledge on relationships between spending, taxes, deficit, and debt associated with lower expectations of tax increases: -0.03*** (column 1). Observations 27,202; R-squared 0.13. (Table A.9)

### Fiscal consolidation episodes and country-level measures (Annex)
- Fiscal consolidation episodes listed by country (Table A.1). Examples preserved exactly:
  - United Kingdom: 1969 - 1971, 1977 - 1982, 1994 - 1999, 2010 - 2015
  - United States: 1969 - 1971, 1976 - 1977, 1978, 1980 - 1981, 1985 - 1986, 1988, 1990 - 1998, 2011 - 2013
  - Japan: 1972 - 1974, 1979 - 1983, 1997 - 1998, 2003 - 2007, 2014 - 2015, 2019 - 2020 (Table A.1)

### Notes and methodology details (as reported)
- Outcome variables in expectation regressions are standardized using the mean and standard deviation of the control group. Regressions include socio-demographic characteristics and country fixed effects; standard errors are clustered at the country level unless noted. Significance notation: *** p<0.01, ** p<0.05, * p<0.1. (Multiple tables)
- Information treatments described in survey instrument:
  - Treatment 1: shows debt levels and estimates for this year and states "debt as a share of GDP has increased relative to recent years."
  - Treatment 2: same as Treatment 1 plus identity "Debt in the future = Debt issued today + interest paid on today’s debt + government spending today – taxes collected by the government today."
  - Treatment 3: includes Treatment 2 content and adds "Economic forecasters predict that government debt in the UK will remain high relative to historical standards in the coming years." (Survey Annex)

*Source: IMF staff calculations based on IMF-YouGov Survey (content unit: wpiea2025197-print-pdf - 13.1 Spending Cuts13.2 Tax Increases).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025197-print-pdf.pdf_