Praxis Data Brief · PDB-002

Income-to-Poverty Missingness

A careful reading of INDFMPIR in the August 2021-August 2023 NHANES demographics file

Audience
Analysts using socioeconomic variables in NHANES
Reviewed
Format
One-page PDF · US Letter

Guide overview

The essential idea

INDFMPIR is the ratio of reported family income to the applicable federal poverty guideline. In DEMO_L, 9,892 of 11,933 interviewed records have a released value and 2,041 are missing. The resulting 17.1 percent missing fraction is an unweighted file calculation. It is large enough that automatic complete-case exclusion can change the analytical sample and potentially its composition.

Practical framework

What to apply

9,892 available values

What does an observed INDFMPIR value represent?

  • INDFMPIR divides annual family or individual income by the poverty guideline applicable to survey year, family size, and location.
  • Observed values are available for 82.9 percent of interviewed DEMO_L records as an unweighted file ratio.
  • Some incomes were reported exactly; others were assigned the midpoint of a selected income range.
  • The measure is a ratio, not household income in dollars and not a direct measure of wealth.

2,041 missing values

Why should the 17.1 percent missing fraction be examined?

  • INDFMPIR is missing for 2,041 of the 11,933 interviewed records in the public codebook frequencies.
  • NCHS did not compute the ratio when the required income detail was unavailable or income data were missing.
  • Excluding these records can reduce precision and alter the distribution of age, race and Hispanic origin, health status, or other covariates.
  • The direction and importance of resulting bias cannot be inferred from the missing count alone.

Values at 5.00 are top-coded

What does the maximum public-use value mean?

  • Values at or above 5.00 are released as 5.00 because of disclosure concerns.
  • The codebook reports 2,149 records at the released value of 5.00.
  • A value of 5.00 therefore means 5.00 or greater, not an exact ratio of 5.00.
  • Modeling choices should respect this upper censoring and the study's scientific purpose.

Report the decision trail

How should INDFMPIR appear in an analysis and manuscript?

  • Show the number and weighted proportion missing within the actual eligible analytical population.
  • Compare key observed characteristics by INDFMPIR availability using survey-aware methods.
  • Justify complete-case analysis, categorization, imputation, missing-indicator use, or sensitivity analyses.
  • State the top-coding rule, chosen scale or categories, and any implications for interpretation.

Before you finish

A short quality review

  1. INDFMPIR is defined as a ratio to the applicable poverty guideline, not as dollar income.
  2. Missingness is quantified in the analysis-eligible population, not only the full file.
  3. Characteristics of records with and without INDFMPIR are examined where appropriate.
  4. The public-use top-code of 5.00 or greater is stated.
  5. The primary missing-data strategy is justified and sensitivity analyses are considered.
  6. Unweighted counts are distinguished from weighted estimates and population inference.

Missing is a result of the data-generating process. A complete-case model may be reasonable in some analyses, but it should be a documented scientific decision - not an unnoticed consequence of default software behavior.

Authoritative sources

Continue with the primary guidance