What's Going on with SSA's "Death Master File"?

The Social Security Administration's "Death Master File" (DMF) has long been one source used by financial institutions to check whether the PII associated with an application corresponds to a legitimate, living person.

It has been reported that the newly-created Department of Government Efficiency (DOGE) has staff members working at the Social Security Administration in search of dead people who were previously missing from the DMF. News media have also reported on the potential addition of thousands of living immigrants to the DMF.  

The DMF is an important source of data for many FIs, so SentiLink is investigating the DMF to better understand what changes are being made.

What is the DMF and how is it used?

The DMF is a list of the identities of deceased people that is maintained by the SSA. At least until recently (more on this later), it included only identities that have been issued an SSN and that the SSA believes are deceased based on, for example, filings by state vital records offices.

Despite the "master" name, the DMF is not a one-stop-shop for determining whether an identity is living or deceased, and misses many deceased individuals. Many companies (including SentiLink) collect deceased data from a wide variety of sources including the DMF, obituaries, probate records, death reports to credit bureaus, etc. Even so, the DMF is an important source of data for verifying whether the identity on a checking account application (for example) is alive or deceased.

Of note: A 2023 audit by the Office of the Inspector General (PDF link) found that while there were millions of likely-deceased people (ages 100 or older) whose identities did not appear in the DMF, the vast majority of them (98%) did not receive Social Security benefits. 

The SSA publishes weekly updates to the DMF. Each week typically sees between 8,000 and 13,000 new identities added to the file. But in March, that changed dramatically.

A big spike in additions to the DMF

Beginning the week of March 14, 2025, SentiLink has observed a significant increase in the number of records added to the DMF. This increase is also visible in the file metadata published on the web by the National Technical Information Service. Prior to March 14, no week in 2025 had seen the addition of more than 13,000 identities to the DMF. However, the next five DMF update files published have all seen significant increases:

• 3/14 file: 960,010 records added.
• 3/21 file: 2,520,648 records added.
• 3/28 file: 3,840,839 records added.
• 4/4 file: 2,980,684 records added.
• 4/11 file: 25,646 records added.

Digging deeper into this spike, SentiLink examined the DMF data, looking at the March-April spike, and found that the vast majority of the spike comes from records of individuals with ages over 120. There was also a smaller bump in records added with an age of 0-9 years old: 

SentiLink's Fraud Intelligence Team checked identities in both groups and found that they mostly appear to be legitimate, and legitimately deceased, people. The spike in records with an age of zero appears to be due, in nearly all circumstances, to a data issue: "decedent age" defaults to zero in the case of a missing DOB, and most of these records had missing DOBs. However, despite the missing DOBs, the SSNs associated with these individuals would correspond to older individuals based on SSN issuance.

We also found that more than 10 million records in the March cohort had a listed date of death in March 2025 (compared to an average ~35k for other months). This further confirms that there was a batch data addition. 

In conclusion, the DMF data clearly points to a significant backfill of data to the DMF, consisting primarily of legitimate deceased individuals born 120+ years ago, and that the death date for many of these was entered as one of these four Sundays in March:

Have living immigrants been added to the DMF?

Last week, several news outlets reported that DOGE has added or is planning to add the identities of non-deceased immigrants to the DMF.  While the unusually large 4/11 file (25k records compared to the typical 8-13k/week) is suggestive of that, we don’t yet have confirmation either way.