Data & methods
The site combines survey evidence, a randomized information experiment, enacted legislation, and successive CBO projections. This page documents the samples, construction, and limits.
Survey samples
Participants were recruited in the United States through Prolific prescreeners. The labels below identify screened research samples, not probability samples of the full investing public or electorate.
U.S. residents recruited through Prolific’s government-bond ownership prescreener. Eligibility relies on self-reported ownership; the survey does not observe position size or trading activity.
Used for valuation, sustainability, portfolio responses, and the randomized information treatment.U.S. residents recruited through Prolific’s registered-voter prescreener. The sample is useful for studying the political relevance of fiscal beliefs but is younger and more educated than the U.S. adult population.
Used for valuation, sustainability, and reported voting decisions.Respondents whose highest completed degree is a master’s or doctoral degree in economics or finance.
Used as a trained comparison group for valuation and sustainability beliefs.Representativeness
Both samples are younger than the U.S. adult population; bond investors are also more male and more often employed full time.
| Characteristic | Bond investors | Registered voters | U.S. adults |
|---|---|---|---|
| SexMale | 57.4% | 44.4% | 48.5% |
| SexFemale | 42.6% | 55.6% | 51.5% |
| AgeUnder 35 | 38.4% | 38.4% | 28% |
| Age35–49 | 37.2% | 39.1% | 24% |
| Age50–64 | 19.1% | 17.1% | 25% |
| Age65+ | 5.4% | 5.5% | 23% |
| RaceWhite | 64.6% | 69.6% | 61% |
| RaceBlack | 18.2% | 14.6% | 13% |
| RaceAsian | 6.2% | 5.5% | 7% |
| RaceMixed or other | 11.1% | 10.3% | 19% |
| EmploymentEmployed full time | 59.9% | — | 50% |
| EmploymentEmployed part time | 16.4% | — | 10% |
| EmploymentNot in paid work | 7.6% | — | 37% |
| EmploymentUnemployed, seeking work | 9.0% | — | 2% |
| EmploymentOther | 7.0% | — | — |
| EducationLess than bachelor’s | — | 40.2% | 62% |
| Bachelor’s | — | 41.7% | 23% |
| Graduate degree | — | 18.2% | 15% |
Benchmark: U.S. adult comparisons reported in the working paper. Age, sex, and race use Census benchmarks; employment uses CPS benchmarks; education uses adults aged 25 and older. Race and employment categories are only approximate matches to Prolific’s fields. Values are rounded, so grouped shares may not sum exactly to 100. Dashes indicate unavailable comparisons.
The primary survey analyses report sample estimates rather than claiming population-representative prevalence.
For the 994 voters with complete fields, post-stratification weights rake age and sex to U.S. adult margins and education to the available margin for adults aged 25 and older. Valuation shares move by no more than 1.1 percentage points, and the ordering of sources is unchanged.
Observed-demographic weights cannot correct selection on unobserved characteristics, including attention to fiscal issues.
Recruitment, screening, and sequence
The survey screens for prior awareness of the government-debt question, records baseline beliefs and behavior, elicits valuation mechanisms, randomizes information treatments, and then measures sustainability beliefs and intended decisions.
Perceived debt/GDP, debt-sustainability concern, and sample-specific reports of past behavior.
Which sources support government debt value and the relative importance assigned to each.
Respondents are assigned to control, debt-trajectory information, safe-asset vignette, or rollover vignette groups.
Crisis beliefs, sustainable debt assessments, and prospective Treasury intentions where applicable.
Exclusions: most exclusions reflect incomplete responses. Criteria requiring judgment remove 1.9% of eligible investors and 0.8% of eligible voters. The debt/GDP cap removes 3 of 250 respondents with graduate training in economics or finance.
Fiscal expectations data
The project aligns enacted legislation with successive CBO projections to reconstruct the expected path of taxes, spending, primary deficits, and debt from each quarterly vantage point.
Revenue and outlay effects are assigned by fiscal year and budget component.
Identifies discrete fiscal news.Successive publications provide the contemporaneous budget path used at each forecast date.
Tracks how the outlook changes over time.Projected primary-deficit paths are compared with subsequently realized values at matched horizons.
Measures projection gaps, not a pure forecast error.The analysis links fiscal news to forecast revisions and Treasury-market valuation.
Connects fiscal news to beliefs and prices.Interpretive limits
Convenience samples describe the screened respondents; they are not population probability samples.
The 100-point decomposition applies the paper’s benchmark allocation rule and is not a direct portfolio valuation.
Past behavior and future intentions are self-reported and do not measure transaction quantities or realized trades.
Later legislation and economic or technical revisions enter the realized-versus-projected difference, so it is not a pure CBO forecast-error statistic.
Related findings