Federal Data Users

 View Only
  • 1.  New Census Bureau noncitizen voting brief

    Posted 26 days ago

    Yesterday the Census Bureau released Noncitizen Voting in the 2020 Election, A Beginning Analysis, which links a commercial voter file to federal administrative records and reports ~24,000 noncitizen voters in the 2020 election, with 32 million records "yet to analyze."

    The brief frames this work in terms of a Census Bureau crime-statistics mandate dating back to 1902. The history is more complicated: the 1902 act authorized decennial statistics on "crime, pauperism, and benevolence, including prisoners, paupers, juvenile delinquents, and inmates of benevolent and reformatory institutions"-in other words, counting people already in institutions, not identifying potential lawbreakers. The Bureau discontinued its national Judicial Criminal Statistics program after 1945, and its prisoner-statistics work subsequently shifted to the Justice Department.

    I'm concerned about the potential chilling effect of this work. The brief notes that the analysts "did not have access to the names of the population analyzed or any other identifying information." And the Bureau is publishing aggregate statistics, not identifiable records. But this still sends a potentially troubling message about how administrative data shared with the Census Bureau might be used.

    The Bureau's own research found that households containing noncitizens were particularly sensitive to citizenship-related data collection. A 2024 Census study found reduced response, increased household-roster omissions and internet breakoffs, and an increased undercount among households with noncitizens. 

    It's telling that no statistician or researcher put their name on this report-unlike virtually every other Census Bureau publication. Mike Schneider has a nice piece on that if you want to learn more:https://countofcounting.substack.com/p/no-researchers-name-is-on-census

    Anyone have any other thoughts about this? I'd especially welcome perspectives from folks who know more about Census Bureau history than me. 



    ------------------------------
    Mark Mather
    Associate VP
    PRB
    ------------------------------


  • 2.  RE: New Census Bureau noncitizen voting brief

    Posted 26 days ago
    Re Mar's post:  "I'm concerned about the potential chilling effect of this work."

    All,

    I'm concerned too, and suspect that this forum could do good things to sort out the actual sequence of events, what happened when, why, how, etc., so we can think about responses and support for the bureau going forward.

    Mike Schneder, Count of Counting and Hansi Lo Wang, NPR have both posted more information this am.  If I understand their posts correctly, the president referenced a Census Bureau study on Truth Social saying he won the 2020 election, and subsequently, the Commerce Secretary posted the study on the bureau website.  I don't know the mechanics of that, but confirmation would be good.

    Who ordered the study and when?  I presume in the current administration, but did the impetus come from the White House, Commerce Department, request from another agency like Homeland Security?   Getting details on when it started, how bureau officials responded, would help.  Who is paying for it? Where in the bureau is the work being done? Did the bureau have a planned release schedule for results?  Was that preempted by the Truth Social post?

     

    The historical analogies are many, in addition to Mark's example.  The one that first came to mind for me was the 2003 Arab tabs controversy.  In 2002, Homeland Security asked for a tabulation from the 2000 census of "Arabs" by zipcode.  The bureau provided the tabulation using the 'ancestry' variable from the sampled 2000 census long form, but did not immediately release a public file on what would now be the special tabulations page on the website.  The Electronic Privacy Information Center filed a FOIA, and by August 2003, the issue was reported in the NYT.  There were differences from the current situation; the releases were tabular, sampled, not reports on matched microdata.  But the 'sensitive' nature of the release was obvious to the public and generated much heat. Let me stop for now, and if I have time, I'll add more.


    There are many threads to disentangle. 

    Margo Anderson








  • 3.  RE: New Census Bureau noncitizen voting brief

    Posted 26 days ago

    I agree with all of your concerns and is what I immediately thought about when I saw the post and review the "research."  I immediately wondered what would be next after they made a point about using CB for crime analysis.  Thank you for the additional historical context.

    The post was shared on the WH Facebook page.  So, it has had more widespread distribution than just Truth Social.



    ------------------------------
    Mike Cline
    State Demographer
    NC Office of State Budget and Management
    ------------------------------



  • 4.  RE: New Census Bureau noncitizen voting brief

    Posted 25 days ago

    Now that this report has been released at the national level we should be alert for similar instances being requested at the state level.  Whether this is possible to produce is unclear to me but perhaps with access to an RDC it might be.

    In the early 2000s the police commissioner in my city asked for a report about people of Arabs and other middle eastern descent, so local leaders and officials can and do take cues from federal actions such as this one.



    ------------------------------
    Cliff Cook
    Urban Planner and Local Demographer (retired)
    ------------------------------



  • 5.  RE: New Census Bureau noncitizen voting brief

    Posted 25 days ago

    The state-level numbers of noncitizen voters look suspiciously round...e.g, 500 for Maryland.  Could it be 499 or 501?  These are obviously rough estimates but the text (and I'm sure most news articles about the report) make them seem like counts.

    I'm not sure I understand the method well enough -- do the "noncitizen voters" include people for whom they couldn't find a matching administrative files?  Might there be a lot of married women listed under maiden names, and Hispanics listed under maternal instead of paternal names, and Van Dykes listed as Dykes, And asians listed as Abe Shinzo instead of Shinzo Abe, and all the usual problems when you try to match multiple files?



    ------------------------------
    John Haaga
    Maryland Commission on Aging
    ------------------------------



  • 6.  RE: New Census Bureau noncitizen voting brief

    Posted 25 days ago

    The Massive Data Institute has published this incredibly helpful "dissection" of this report. They lay out the checklist the Census Bureau uses before linking data, and this violates 6 of the checks that would have stopped this report from being undertaken were professionals at the Bureau the ones working on it.

    One of the most interesting factors they discuss (on page 6):

    "The number of alleged noncitizen voters found is indistinguishable from 0.

    "The number of matches found in "Beginning Analysis" report (just 24,000) is well below what we would expect based on false positive error rates.

    "As noted above, prior Census Bureau research into incorrect match rates demonstrates that false-positive linkages can exceed 6%. Rates above 1% are not uncommon. This means that the 24,000 is well under what we would typically expect to see based on linkage error, suggesting exactly the opposite of the headline: noncitizen voting simply has not, and cannot, be proven based on this analysis." (emphasis mine)



    ------------------------------
    Sarah Eisele-Dyrli
    Assistant Director
    CTData Collaborative
    ------------------------------



  • 7.  RE: New Census Bureau noncitizen voting brief

    Posted 24 days ago

    My understanding is that rounding is the new way of "protecting privacy."

    And absolutely false matches and unmatched records are a problem--that the Census Bureau would typically point out--in the analysis. I'm going to post a link below to a point-by-point critique of the report.



    ------------------------------
    Beth Jarosz
    ------------------------------



  • 8.  RE: New Census Bureau noncitizen voting brief

    Posted 24 days ago

    Massive Data Institute did a detailed read of the unsigned "Beginning Analysis" and compiled a point-by-point critique of both methodological and process concerns. https://mdi.georgetown.edu/response-to-voter-record-analysis/ 



    ------------------------------
    Beth Jarosz
    ------------------------------