## Public Debt and Household Inflation Expectations — Section 1–3 (excerpt)

## Source details

**Canonical URL:** [Public Debt and Household Inflation Expectations — Section 1–3 (excerpt)](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023066-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023066-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023066-print-pdf.pdf.json)

---

### Abstract and key findings
- The paper uses randomized controlled trials in the US, UK, and Brazil to examine the causal effect of public debt on household inflation expectations.
- People on average largely underestimate the level of public debt and revise inflation expectations upwards when informed about the correct levels.
- The extent of the revisions to inflation expectations is proportional to the size of the information surprise: people with lower prior beliefs about public debt increase inflation expectations more strongly once informed about the correct debt level.
- Quantitative effect: an increase in public debt by 10 percent of GDP leads to an increase in one-year ahead inflation expectations by about 0.6 percent.
- Confidence in the central bank considerably reduces the sensitivity of inflation expectations to public debt; people more confident in the central bank’s determination to fight inflation or more knowledgeable about the inflation target revise expectations much more modestly.
- Cognitive channel evidence: respondents perceive high public debt as bad news for the economic outlook, leading to both higher inflation and unemployment expectations—interpreted as stagflationary (negative supply shock) effects.
- No evidence that information about public debt increases the likelihood that respondents expect monetary finance by the central bank relative to the control group.
- Heterogeneity: women increase inflation expectations on average twice as much as men when informed about debt levels; lower-income households also show higher sensitivity. These patterns are partly explained by weaker confidence in the central bank and less knowledge about its inflation target among women and lower-income respondents.

### Motivation and contribution
- Motivation:
  - Post-COVID-19 recovery featured a sharp increase in inflation against record-high public debt levels and raised concerns that high public debt may weaken central bank resolve to fight inflation (fiscal dominance/monetary finance).
- Contribution:
  - Provides clean causal identification using an information provision experiment that elicits prior beliefs about debt levels, allows belief updates, and measures revisions in inflation expectations relative to a control group.
  - Extends analysis beyond the US to the UK and Brazil to test robustness across different economic and institutional environments.
  - Documents the moderating role of central bank credibility and explores cognitive transmission channels and demographic heterogeneity.

### Survey design, sample, and data cleaning
- Data source and timing:
  - Survey data collected by YouGov, administered online.
  - Survey administered between November 23rd and December 2nd, 2022.
- Sample:
  - 2,000 individuals in each country (US, UK, and Brazil); participants selected to ensure national representativeness; sample weights provided and used throughout.
- Questionnaire structure:
  - Total of 8 substantive questions besides demographic and socioeconomic characteristics.
  - Five pre-treatment questions: expected inflation rate over the next 12 months; the central bank’s inflation target; the level of public debt in percent of GDP; two confidence questions (central bank preventing large increases in inflation; government preventing large increases in government debt) on a scale from 1 to 5.
  - Treatment: informed about the correct level of public debt in percent of GDP—equal to 122, 87, and 88 percent in the US, UK, and Brazil, respectively. No information provided to the control group.
  - Three post-treatment questions: opportunity to revise inflation expectations (after reminder of prior answer); expected unemployment rate in 12 months; likelihood that the central bank would print money to finance government spending or repay government debt.
- Data cleaning:
  - Dropped 158 respondents out of 6,000 reporting inflation expectations in excess of 100 percent or debt levels above 200 percent of GDP or below 0.
  - Treated as missing about a dozen observations reporting beliefs about the inflation target above 100 percent and expected unemployment rates above 100 percent or below zero.
  - Table B.1 confirms that group assignment is not predictable by individual characteristics.

### Empirical identification and core estimates
- Identification strategy:
  - Exogenous variation delivered by information treatment (told correct debt level after eliciting prior belief).
  - Revisions measure: ∆Eiπ = E_postiπ − E_preiπ.
  - Baseline regression: ∆Eiπ = α + βTi + ξXi + εi, where Xi controls age, gender, geographical region, employment status, education level, income group, and whether they have kids. Parameter β measures whether treatment-group revisions exceed control-group revisions.
- Core quantitative findings:
  - Average pre-treatment public debt beliefs:
    - US: 54 percent of GDP
    - UK: 63 percent of GDP
    - Brazil: 36 percent of GDP
  - Correct values at end of 2022:
    - US: 122 percent of GDP
    - UK: 87 percent of GDP
    - Brazil: 88 percent of GDP
  - Treatment raises one-year-ahead inflation expectations; effect proportional to information shock.
  - Reported δ coefficients relating prior debt belief to treatment effect:
    - UK: 0.062
    - Brazil: 0.061
    - US: 0.045
  - Quantitative translation: a 10 percent increase in the debt-to-GDP ratio generates an upward revision in inflation expectations of about 0.6 percent.
  - Selected treatment estimates from heterogeneity regressions (Table 4 excerpts):
    - Column (4) All: Treatment = 4.25***
    - Column (8) All (with additional controls): Treatment = 4.69***
    - Column (4) All: T × women = 4.91***
    - Column (8) All (with additional controls): T × women = 4.34***
    - Column (8) T × high income = -4.47***
  - Pre-treatment and post-treatment sample means (selected):
    - Pre-treatment expected inflation (All): 19.1
    - Pre-treatment public debt beliefs (All): 54.0
    - Post-treatment expected inflation (All): 22.2
    - Post-treatment expected unemployment (All): 18.6
    - Post-treatment likelihood of monetary finance (All): 3.2

### Role of credibility (central bank and government)
- Measures used:
  - Degree of confidence in the central bank (scale 1–5 that the central bank will prevent large increases in inflation).
  - Absolute gap between people’s beliefs about the inflation target and the correct value (IT gap).
  - Degree of confidence in the government to prevent large increases in government debt.
- Findings:
  - Greater confidence in the central bank → smaller upward revision in inflation expectations when informed about debt.
  - People with lowest central bank confidence increase inflation expectations three times as much as people with highest central bank confidence (panel specification).
  - Interaction T × IT gap positive: larger IT gap (less knowledge of inflation target) → stronger revisions; coefficients significant for the US and panel regressions.
  - Government credibility: stronger confidence in government associated with smaller revisions in the US and Brazil; no statistically significant effects in the UK or panel sample. Government credibility effects weaken when controlling for central bank credibility.
- Conclusion: central bank credibility plays the predominant role in anchoring inflation expectations and shaping sensitivity to debt-level news.

### Cognitive transmission channels and mechanism evidence
- Outcomes elicited at end of survey:
  - Likelihood that the central bank “will print money in the future to finance government spending or repay government debt” (scale 1–5).
  - Expected unemployment rate in the near future.
- IV strategy:
  - First stage: E_postiπ = α + βTi + νE_preiπ + ξXi + εi.
  - Outcome equation: Y_posti = α + βE_preiπ + γE_postiπ + ξXi + εi, where E_post instrumented by the first stage.
- Findings:
  - Likelihood of monetary finance (Table 3 Columns (1)–(4)): no evidence that post-treatment inflation expectations correlate with higher perceived likelihood of monetary finance; null result holds even for college-educated subsample.
  - Expected unemployment (Table 3 Columns (5)–(8)): strong evidence that post-treatment inflation expectations are correlated with higher expected unemployment.
    - Quantitative magnitude: a one percent increase in inflation expectations due to the debt information treatment leads to an increase in unemployment expectations by 0.34–0.35 percent.
  - Interpretation: respondents tend to interpret high public debt as adverse supply shocks implying stagflationary effects (higher inflation and higher unemployment), rather than expecting central bank monetary financing.

### Heterogeneity and behavioral patterns
- Gender:
  - Women revise inflation expectations substantially more than men when informed about high public debt—about twice as much on average; robust across countries and after controls.
  - This stronger responsiveness partly explained by lower confidence in the central bank and less knowledge about the inflation target among women.
- Income and education:
  - Higher-education and higher-income individuals revise inflation expectations less strongly.
  - High-income men are largely unresponsive to debt information; linked to higher confidence in the central bank and better knowledge of the inflation target.
- Other demographics:
  - Treatment effects decline with higher prior debt beliefs: participants whose prior beliefs are close to correct show no statistical difference between treatment and control.

### Policy implications and recommendations
- Visibility of fiscal conditions:
  - Because households substantially underestimate public debt, events that draw public attention to fiscal positions (for example, market pressures during monetary tightening) could raise household inflation expectations.
- Communication and targeting:
  - Targeted communication campaigns informing specific population segments (notably women and lower-income people) about central banks’ mandates and track records could strengthen anchoring of inflation expectations.
  - Strengthening public knowledge of the inflation target and central bank credibility can moderate the pass-through from public debt perceptions to inflation expectations.

*Prepared by Francesco Grigoli and Damiano Sandri, March 2023 — IMF Working Paper WP/23/66*

### Section 1

### Public Debt and Household Inflation Expectations — Section 1

### Abstract and key findings
- The paper uses randomized controlled trials in the US, UK, and Brazil to examine the causal effect of public debt on household inflation expectations.
- People on average largely underestimate the level of public debt and revise inflation expectations upwards when informed about the correct levels.
- The extent of the revisions to inflation expectations is proportional to the size of the information surprise: people with lower prior beliefs about public debt increase inflation expectations more strongly once informed about the correct debt level.
- Quantitative effect: an increase in public debt by 10 percent of GDP leads to an increase in one-year ahead inflation expectations by about 0.6 percent.
- Confidence in the central bank considerably reduces the sensitivity of inflation expectations to public debt; people more confident in the central bank’s determination to fight inflation or more knowledgeable about the inflation target revise expectations much more modestly.
- Cognitive channel evidence: respondents perceive high public debt as bad news for the economic outlook, leading to both higher inflation and unemployment expectations—interpreted as stagflationary (negative supply shock) effects.
- No evidence that information about public debt increases the likelihood that respondents expect monetary finance by the central bank relative to the control group.
- Heterogeneity: women increase inflation expectations on average twice as much as men when informed about debt levels; lower-income households also show higher sensitivity. These patterns are partly explained by weaker confidence in the central bank and less knowledge about its inflation target among women and lower-income respondents.

### Motivation and contribution
- Motivation:
  - Post-COVID-19 recovery featured a sharp increase in inflation against record-high public debt levels.
  - Concerns that high public debt may weaken central bank resolve to fight inflation (fiscal dominance/monetary finance).
  - Broader debate on fiscal policy’s role in determination of the price level (references: Leeper and Leith (2016), Cochrane (2023) as cited).
- Contribution:
  - Provides clean causal identification using an information provision experiment that elicits prior beliefs about debt levels, allows belief updates, and measures revisions in inflation expectations relative to a control group.
  - Extends analysis beyond the US to the UK and Brazil to test robustness across different economic and institutional environments.
  - Documents the moderating role of central bank credibility and explores cognitive transmission channels and demographic heterogeneity.

### Survey design and experimental setup
- Data source: survey data collected by YouGov, administered online.
- Sample:
  - 2,000 individuals in each country (US, UK, and Brazil).
  - Survey administered between November 23rd and December 2nd, 2022.
  - Participants selected by YouGov to ensure national representativeness; sample weights provided and used throughout the analysis.
- Randomization and groups:
  - Participants were randomly allocated to a control and a treatment group of equal size.
  - All survey questions were common to both groups to control for priming effects.
- Questionnaire structure:
  - Total of 8 questions besides demographic and socioeconomic characteristics.
  - Five pre-treatment questions: expected inflation rate over the next 12 months; the central bank’s inflation target; the level of public debt in percent of GDP; two confidence questions (central bank preventing large increases in inflation; government preventing large increases in government debt) answered on a scale from 1 to 5 with higher values denoting stronger confidence.
  - Treatment: informed about the correct level of public debt in percent of GDP—equal to 122, 87, and 88 percent in the US, UK, and Brazil, respectively. No information provided to the control group.
  - Three post-treatment questions: opportunity to revise inflation expectations (after reminder of prior answer); expected unemployment rate in 12 months; likelihood that the central bank would print money to finance government spending or repay government debt.
- Design considerations:
  - Prior beliefs elicited to measure the direction and size of the information shock and isolate genuine belief-updates from priming.
  - Confidence in monetary and fiscal institutions elicited to assess anchoring effects on inflation expectations.
  - Questions on likelihood of monetary finance and future unemployment used to shed light on cognitive channels.
  - Demographic and socioeconomic information collected to analyze heterogeneity.

### Empirical identification and interpretation
- Identification:
  - The information treatment is the source of exogenous variation in respondents’ information sets about public debt levels.
  - Causal effect measured by comparing revisions in inflation expectations between treatment and control groups.
- Interpretation of effects:
  - The proportionality of revisions to the information shock supports a causal link between perceived public debt levels and household inflation expectations.
  - The attenuation of the effect among respondents with higher confidence in the central bank indicates institutional credibility anchors expectations.
  - The absence of treatment-induced increases in expected monetary finance suggests respondents instead interpret higher debt as associated with worse economic outlook (stagflationary negative supply shock) leading to simultaneous increases in expected inflation and unemployment.

### Heterogeneity and behavioral patterns
- Gender:
  - Women increase inflation expectations much more strongly than men when informed about debt levels—on average twice as much—even after controlling for additional individual characteristics.
- Income:
  - Lower-income households display higher sensitivity of inflation expectations to public debt levels.
- Mechanisms:
  - Differential sensitivity is partly explained by lower confidence in the central bank and less knowledge about the inflation target among women and lower-income respondents.

### Related literature (overview as presented)
- Builds on information provision experiments studying household inflation expectations (e.g., Armantier et al. (2016); Cavallo, Cruces and Perez-Truglia (2017)).
- Complements literature on central bank communication and household expectations (e.g., Haldane and McMahon (2018); D’Acunto et al. (2020); Coibion, Gorodnichenko and Weber (2022); Coibion et al. (2023b, 2023c)).
- Related to Coibion, Gorodnichenko and Weber (2021) who find US household inflation expectations respond to news about future debt levels; current paper differs by eliciting prior debt beliefs, extending country coverage (UK and Brazil), and emphasizing the moderating role of central bank credibility and cognitive channels.
- Consistent with Roth, Settele and Wohlfart (2022) on widespread underestimation of public debt; complements Andre et al. (2022) on heterogeneous responses to fiscal shock vignettes.

*Prepared by Francesco Grigoli and Damiano Sandri, March 2023 — IMF Working Paper WP/23/66*

### Section 2

### Section 2

### Data cleaning and sample
- Dropped 158 respondents out of 6,000 that report inflation expectations in excess of 100 percent or debt levels above 200 percent of GDP or below 0.
- Treated as missing a dozen observations that report beliefs about the inflation target above 100 percent and expected unemployment rates above 100 percent or below zero.
- Table B.1 confirms that the group assignment is not predictable by individual characteristics.

### Baseline finding: public debt is substantially underestimated
- On average, people report public debt levels equal to:
  - US: 54 percent of GDP
  - UK: 63 percent of GDP
  - Brazil: 36 percent of GDP
- Correct values at the end of 2022 were:
  - US: 122 percent of GDP
  - UK: 87 percent of GDP
  - Brazil: 88 percent of GDP
- Result consistent with Roth, Settele and Wohlfart (2022): median response 60 percent vs correct 105 percent in their US 2016 comparison.

### Identification strategy: information provision treatment
- Generate exogenous shift by informing treatment-group participants about the correct level of public debt in their country; control group receives no information.
- Survey elicits inflation expectations before (E_pre) and after (E_post) the information treatment; revise measure defined as ∆Eiπ = E_postiπ − E_preiπ.
- Baseline regression estimated for each country and pooled with country-fixed effects:
  - ∆Eiπ = α + βTi + ξXi + εi
  - Xi controls: age, gender, geographical region, employment status, education level, income group, and whether they have kids.
- Primary parameter of interest: β measuring whether treatment-group revisions exceed control-group revisions.

### Treatment effects and heterogeneity by prior beliefs
- Columns (1)–(4) of Table 1: treatment group substantially increases inflation expectations relative to control; effects "highly statistically significant and economically sizeable", upward revisions by several percentage points.
- To relate revisions to the size of the information shock, exploit pre-treatment debt belief Di with:
  - ∆Eiπ = α + γDi + (β + δDi)Ti + ξXi + εi
- Interpretation: δ negative implies larger prior underestimation (lower Di) → larger upward revision when informed.
- Table 1 Columns (5)–(8): δ negative and statistically significant across all countries.
- Quantitative interpretation:
  - δ implies an increase in public debt by 10 percent of GDP generates an upward revision in inflation expectations by about 0.6 percent.
  - Reported δ coefficients:
    - UK: 0.062
    - Brazil: 0.061
    - US: 0.045 (modestly smaller)

- Figure 1 (binscatter): treatment-group participants with lowest initial debt perceptions revise inflation expectations upwards the most; effect declines with higher prior debt beliefs; no statistical difference between treatment and control for people whose debt beliefs were close to correct values.

### The role of credibility (central bank and government)
- Survey measures:
  - Degree of confidence in central bank (scale 1 to 5 that central bank will prevent large increases in inflation).
  - Absolute gap between people’s beliefs about the inflation target and the correct value (IT gap).
  - Degree of confidence in the government to prevent large increases in government debt.
- Regression:
  - ∆Eiπ = α + βTi + β̄ Ti × Ci + ᾱ Ci + ξXi + εi, where Ci includes CB confidence, IT gap, and Govt confidence.
- Table 2 results — central bank credibility (Columns (1)–(4)):
  - Greater confidence in the central bank → smaller upward revision in inflation expectations when informed about debt.
  - Panel specification example: people with lowest CB confidence increase inflation expectations three times as much as people with highest CB confidence.
  - Interaction T×IT gap positive; people with larger IT gap (less knowledge of inflation target) revise more strongly. Coefficients significant for the US and panel regression.
- Table 2 results — government credibility (Columns (5)–(8)):
  - Stronger confidence in government associated with smaller revisions in the US and Brazil.
  - No statistically significant government credibility effects in the UK or panel sample.
  - Role of government credibility weakens when controlling for central bank credibility.
- Conclusion: central bank credibility plays the predominant role in anchoring inflation expectations and shaping sensitivity to debt-level news.

### Cognitive transmission channels and evidence on mechanisms
- Survey questions (asked once at end of survey, post-treatment for treatment group):
  - Likelihood that the central bank “will print money in the future to finance government spending or repay government debt” (scale 1 to 5).
  - Expectations about the unemployment rate in the near future.
- IV strategy to link revisions in inflation expectations to these outcomes:
  - Stage: E_postiπ = α + βTi + νE_preiπ + ξXi + εi
  - Outcome equation: Y_posti = α + βE_preiπ + γE_postiπ + ξXi + εi, where Y_post is likelihood of money printing or expected unemployment; E_post instrumented as above.
- Table 3 findings — likelihood of monetary finance (Columns (1)–(4)):
  - No evidence that post-treatment inflation expectations are correlated with likelihood of monetary finance.
  - No correlation detected even when limiting sample to people with a college education.
  - Conclusion: people do not associate higher public debt with a stronger propensity of the central bank to expand the money supply.
- Table 3 findings — expected unemployment (Columns (5)–(8)):
  - Strong evidence that post-treatment inflation expectations are correlated with higher expected unemployment.
  - Quantitative magnitude: a one percent increase in inflation expectations due to the debt information treatment leads to an increase in unemployment expectations by 0.34–0.35 percent (range 0.34 to 0.35 reported across countries).
  - Interpretation: people tend to interpret high public debt as negative supply shocks, implying stagflationary effects (weaker activity and higher inflation).
  - Consistency with prior work: Coibion, Gorodnichenko and Kumar (2018) and Coibion et al. (2023a) show firms and households associate high inflation with worse economic outcomes.

### Heterogeneity and concluding setup for next analysis
- The section concludes by stating the analysis will explore heterogeneity in effects of public debt on inflation expectations depending on demographic and socioeconomic characteristics of survey participants.

*Source: wpiea2023066-print-pdf - Section 2*

### Section 3

### Section 3 — Public Debt and Household Inflation Expectations (excerpt)

### Main experimental design and identification
- Randomized controlled trial: treatment group informed about the correct level of domestic public debt after eliciting pre-treatment beliefs; revisions of inflation expectations compared across treatment and control groups.
- Treatment interaction regression estimated:
  ∆E_i π = α + β T_i + ̄β T_i × W_i + ̄α W_i + ξ X_i + ε_i
  - W_i includes dummies for women, age 46+, children, college education, high income, employed.
- Surveys run in the US, UK, and Brazil.

### Core quantitative findings
- People substantially underestimate public debt levels; correct public debt levels provided to treatment group: US = 122 percent of GDP, UK = 87 percent of GDP, Brazil = 88 percent of GDP (end of 2022, October 2022 WEO database).
- Information about high public debt raises one-year-ahead inflation expectations:
  - A 10 percent increase in the debt-to-GDP ratio raises one-year-ahead inflation expectations by about 0.6 percent.
- Treatment effects from Table 4 (heterogeneity analysis):
  - Treatment (T) effects (selected columns reported in the source):
    - Column (4) All: Treatment = 4.25***
    - Column (8) All (with additional controls): Treatment = 4.69***
  - Treatment × women (selected columns reported in the source):
    - Column (4) All: T × women = 4.91***
    - Column (8) All (with additional controls): T × women = 4.34***
  - Treatment × high income (selected column reported in the source):
    - Column (8) T × high income = -4.47***
- Pre-treatment and post-treatment sample means (selected, from Table C.1):
  - Pre-treatment expected inflation (All): 19.1
  - Pre-treatment public debt beliefs (All): 54.0
  - Post-treatment expected inflation (All): 22.2
  - Post-treatment expected unemployment (All): 18.6
  - Post-treatment likelihood of monetary finance (All): 3.2

### Heterogeneity and mechanisms
- Gender:
  - Women revise inflation expectations substantially more than men. The paper documents that when informed about high public debt levels, women revise inflation expectations upwards by about twice as much as men (robust across countries).
  - The stronger responsiveness of women is partly explained by lower confidence in the central bank and less knowledge about the central bank inflation target (Appendix D). No systematic gender differences in prior debt beliefs were found.
- Income and education:
  - Higher-education and higher-income individuals tend to revise inflation expectations less strongly.
  - High-income men are largely unresponsive to information about public debt. This is linked to higher confidence in the central bank and better knowledge of the inflation target among higher-income individuals.
- Cognitive channels:
  - No evidence that high public debt leads respondents to expect monetary finance (i.e., central bank printing money to finance the government).
  - Instead, respondents interpret high public debt levels as adverse supply shocks that imply both higher inflation and higher unemployment. This suggests increases in inflation expectations due to debt are unlikely to stimulate private demand because respondents also expect a deterioration in the economic outlook.
- Central bank role:
  - People who are more confident in the central bank’s determination to keep inflation low or who are more knowledgeable about the central bank’s inflation target revise expectations much less when informed about public debt levels.
  - Confidence in the central bank matters more than confidence in the government’s fiscal rectitude for anchoring inflation expectations.

### Policy implications and recommendations
- Risk from visibility of fiscal conditions:
  - Because households substantially underestimate public debt, events that draw public attention to fiscal positions (for example, market pressures during monetary tightening) could raise household inflation expectations.
- Communication and targeting:
  - Targeted communication campaigns informing specific population segments (notably women and lower-income people) about central banks’ mandates and track records could strengthen anchoring of inflation expectations.
  - Strengthening public knowledge of the inflation target and central bank credibility can moderate the pass-through from public debt perceptions to inflation expectations.

_Italic: Source — Public Debt and Household Inflation Expectations, Working Paper No. WP/2023/066 (Section 3)._

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023066-print-pdf.pdf_