Sponsored by Consumer Magazine · A Consumer.info + NDS publicationBuy this site · Make your best offer ↗
Data & Research

One family. Five versions of the data.

The first rule of reading the public file: rows are not always people.

One family beside five versions of the same family record
Ask · Measure · Learn

The Short Answer

The first rule of reading the public file: rows are not always people.

Missing answers need a careful approach

A family may not answer every question. The SCF uses multiple imputation: a way to fill gaps while keeping track of uncertainty. Each public family record appears in five completed versions called implicates.

Count families correctly

The 2025 public file contains 21,800 records for 4,360 families. Those records are not 21,800 separate family interviews. Seven completed interviews were removed from the public file for disclosure protection.

Use the weights

Researchers must account for survey weights, all five implicates, and the survey's methods for measuring uncertainty. A plain average of all rows is not a replacement for the required survey analysis.

What our download contains

Our free CSV copies selected published summary measures into tidy columns. It is not household microdata and does not recalculate the Federal Reserve's estimates.

Frequently asked questions

Are five implicates five different families?

No. They are five completed versions of one family’s public record.

How many records and families are in the public file?

The 2025 public file has 21,800 records representing 4,360 families.

Is this site’s free data pack the full SCF file?

No. It is a tidy table of selected published summary measures. The Federal Reserve provides the official public microdata and methods.

The Long Answer · A deeper look

Multiple imputation represents uncertainty in incomplete survey responses by creating several completed datasets. In the SCF public file, each family has five implicates. The 21,800 rows in the 2025 public file represent 4,360 families, not that many independent interviews. Treating them as independent observations can misstate sample size and understate uncertainty.

Researchers must follow the survey’s weighting and variance procedures and combine results across implicates appropriately. Disclosure protection also means the public family count differs from the completed interview count. Our cleaned summary pack avoids claiming a new household-level analysis: it records selected published aggregate estimates with their units, reference years, population labels, and source locations.

Follow the evidence

Federal Reserve Board · SCF data and documentation ↗

Our guide explains the source. Consumer Magazine and NDS did not conduct the official survey.

Make it my share link
A FREE perk for every reader

Share the story.
Pass a little knowledge on.

Choose a story, add your name or business link, and get a reusable share link. Download a graphic or copy a website link. No payment or email is needed.