DebtData uses a source-first methodology. Each report begins by defining the exact question being measured, identifying the population and time period, and selecting the most appropriate available dataset. Values from different sources are not combined unless their definitions, populations, units, and periods are compatible.
Source Selection
Primary administrative and statistical sources are preferred whenever they directly answer the research question. Common source organizations include Federal Student Aid and the U.S. Department of Education, the National Center for Education Statistics, the Federal Reserve and Federal Reserve Banks, the U.S. Census Bureau, and the Bureau of Labor Statistics.
When a government dataset does not contain the required measure, DebtData may use established research institutions, peer-reviewed studies, or other documented datasets. Secondary publishers are not used as the final authority when the original underlying source is available.
Measure Definition
Before a value is used, DebtData identifies its unit, geography, population, reporting period, and measurement definition. This step is especially important for student loan data because several commonly cited figures answer different questions.
- Outstanding federal loan balance measures the federal loan portfolio.
- Consumer student loan balance can include a broader or differently defined credit population.
- Debt at graduation describes a defined graduating cohort.
- Average debt among borrowers excludes people who did not borrow.
- Median debt is not interchangeable with average debt.
- Delinquency and default measure different stages of nonpayment.
Time Periods
The period attached to a statistic refers to the period measured by the source. A dataset period, a source publication date, and a DebtData page publication or modification date are separate concepts. DebtData does not change a data period merely because a report was edited later.
Historical series are kept on a consistent definition whenever possible. When an agency changes methodology or breaks comparability, the change is described instead of silently joining incompatible periods.
Calculations
DebtData may calculate values such as percentage change, absolute change, share of total, ranking, average balance from compatible totals and counts, or inflation-adjusted comparisons. Calculated values are based on published inputs and are identified as DebtData calculations when that distinction matters.
Precision follows the source. Rounded source values are not converted into falsely precise results. A calculation based on rounded billions and rounded millions is reported as approximate even when a calculator produces more decimal places.
Tables and Charts
Tables preserve the underlying data and are treated as the primary presentation layer. Charts are added when a visual comparison makes a trend, ranking, distribution, or composition easier to understand. Important values are not intentionally placed only inside a chart.
Charts and tables should represent the same underlying series when they are paired. Labels, units, periods, and populations are kept consistent between the two formats.
Handling Conflicting Numbers
Two reliable sources can publish different student loan totals without either source being wrong. Differences may result from coverage, account type, borrower definition, timing, sampling, or reporting rules. DebtData identifies the reason for the difference before presenting the values together.
Revisions and Updates
Reports may be updated when a source releases new data, revises an earlier value, changes a methodology, or when DebtData identifies a material error. Historical values are not rewritten solely to make them agree with a newer but differently defined series.
Limits of Interpretation
Descriptive statistics can show that two measures move together, but they do not by themselves establish cause. Claims about economic or behavioral effects require evidence that is appropriate to the question. DebtData distinguishes descriptive, correlational, and causal evidence when the distinction affects interpretation.