The case of the missing US data

Summary

More or Less – "The Case of the Missing US Data"

Date: September 27, 2025
Host: Tim Harford (BBC Radio 4)
Guests: Maggie Levenstein (Director, ICPSR, University of Michigan), John Cubali (Research Assistant Professor, Epidemiology, University of Michigan)

---

Episode Overview

This episode investigates a sudden, sweeping disappearance of US federal government data from the public domain in early 2025. Tim Harford and his expert guests unpack how and why decades of datasets vanished from major agencies’ websites, the impact on research and public services, and the ongoing uncertainty surrounding the restoration and integrity of crucial information.

---

Key Discussion Points & Insights

1. The Disappearance of Data

• Sudden Data Blackout ([00:24]–[01:12])

  • In February 2025, extensive government data and agency websites (e.g., CDC, NIH, Census Bureau) went offline—8,000 pages and 3,000 data sets disappeared, some never returning.
  • Executive orders from President Trump’s second term were the catalyst.
    • "Decades of data began disappearing from web pages [...] Some 8,000 government pages and 3,000 data sets had been taken down." – Tim Harford [00:25]

• Political Motives and Executive Orders ([01:12]–[01:58])

  • Orders particularly targeted diversity, equality, climate change, gender, sexuality, and race-related content.
  • Many agencies, uncertain if content complied with new policies, chose to take entire websites down.
    • "The order targeted websites over language about minorities, indigenous peoples, race, gender, sexuality and disability. Climate change was another banned topic." – Tim Harford [01:34]
    • "They just brought the entire websites down, and that meant that data [...] became inaccessible." – Maggie Levenstein [01:58]

2. Budget Cuts and Human Impact

• Reduction in Personnel & Budget ([02:11]–[02:41])
  • Major layoffs of data-related personnel severely impeded both data production and access.
    • "Lots and lots of people were fired, and those people are often people who produce data or provide access to data." – Maggie Levenstein [02:22]
  • Efforts for massive deficit cuts affected ongoing and future data collection.
    • "There also were budget cuts that were affecting data access and data collection going forward." – Maggie Levenstein [02:41]

3. Restoration and Alteration of Data

• Partial Recovery, With Compromises ([02:41]–[03:21])
  • Some data sets restored, some discontinued, others reappeared in altered forms, especially in areas considered politically sensitive.
    • "There were actual decisions for particular data sets to alter them to obscure information about gender identity or about vaccines or about sexual behavior or about climate." – Maggie Levenstein [02:55]

4. Real-World Consequences

• Loss of Key Information for Vulnerable Groups ([03:21]–[04:11])
  • Removing categories like non-binary gender is not merely symbolic; it severs links to vital public health information.
    • "Teenagers who identify as non binary are more likely to be more vulnerable to certain risks than their peers. That's something we couldn't have known without collecting the data in the past." – Tim Harford [03:21]
    • "When we don't have that information, it's much harder to provide the resources that people need to live full and healthy lives." – Maggie Levenstein [03:47]

5. Impact Across Society and Economy

• Loss of Economic Indicators ([04:11]–[04:48])

  • Example: Bureau of Labour Statistics stopped collecting CPI data from three cities—affecting business planning and economic analysis.
    • "Businesses rely on these and they rely on them to be high quality measures [...] businesses are kind of much more likely to be flying blind." – Maggie Levenstein [04:28]

• Government-Mandated Disclaimers and Doubt ([04:48]–[05:31])

  • New disclaimers on restored websites cast doubt on data integrity, e.g., CDC HIV surveillance page includes language rejecting gender diversity.
    • "Any information on this page promoting gender ideology is extremely inaccurate and disconnected from the immutable biological reality that there are two sexes, male and female." – Quoted disclaimer [05:06], read by Speaker C

6. Case Study: PRAMS Dataset

• PRAMS – A Critical Maternal and Child Health Dataset at Risk ([05:31]–[08:03])
  • Pregnancy Risk Assessment Monitoring System (PRAMS) used to track maternal and child health—vital as US lags behind peers in infant outcomes.
    • "PRAMS [...] is kind of the gold standard for assessing and monitoring maternal and child health in the US." – John Cubali [05:38]
    • "As an infant is around three times more likely to die in the US than in Finland, Japan or Sweden, this seems like a pretty important data set." – Tim Harford [05:55]
  • Since mid-February, PRAMS access requests have gone unanswered; recent data (since 2016) is effectively unavailable.
    • "I submitted multiple requests myself [...] essentially any of the data from the last 10 years, so 2016 or later, that's not publicly available." – John Cubali [07:05]
  • Future of PRAMS is uncertain—office closed in April 2025, despite previous assurances collection would resume.
    • "April 1, the entire PRAMS office was essentially fired. And so now it's really uncertain what this means for this data collection moving forward." – John Cubali [07:32]

7. Ongoing Uncertainty and Data Salvage Efforts

• Challenges in Data Preservation ([08:03]–[08:36])
  • Efforts to recover data are hampered by incomplete archives and lack of metadata/context.
    • "There's some data that you can't get to at all. There's some data that we don't know is gone yet [...] And you also don't have the context, what we call metadata." – Maggie Levenstein [08:15]

---

Notable Quotes & Memorable Moments

• On disappearing data:

  • "Decades of data began disappearing from web pages for agencies such as the Centers for Disease Control and Pre, the National Institutes of Health and the Census Bureau."
    — Tim Harford [00:25]

• On specific censorship:

  • "There were actual decisions for particular data sets to alter them to obscure information about gender identity or about vaccines or about sexual behavior or about climate."
    — Maggie Levenstein [02:55]

• On government disclaimers:

  • "Any information on this page promoting gender ideology is extremely inaccurate and disconnected from the immutable biological reality that there are two sexes, male and female."
    — Government-mandated disclaimer via Speaker C [05:06]

• On research impact:

  • "I submitted multiple requests myself. I reached out to that office directly. You know, I never heard any response."
    — John Cubali [07:05]

• On metadata and context:

  • "You also don't have the context, what we call metadata, the description of the data, all the things that allow you to make inference from the data."
    — Maggie Levenstein [08:15]

---

Timestamps for Key Segments

• [00:24] – Data begins disappearing from federal sites
• [01:12] – Executive orders and rationale for removals
• [02:41] – Restoration, alteration, and erasure of datasets
• [03:21] – Real-world impacts (e.g., non-binary youth, social support)
• [04:11] – Example: CPI data for cities erased
• [04:48] – Government-imposed disclaimers challenge data credibility
• [05:31] – PRAMS dataset described as “gold standard” under threat
• [07:05] – Access to PRAMS data cut off; future collection uncertain
• [08:15] – Difficulties in recovering or validating missing data

---

Conclusion

"The Case of the Missing US Data" exposes the profound vulnerability of public data to political shifts and bureaucratic uncertainty. The episode illustrates evidence-based policy, research, public health, and economic decision-making are all imperiled when critical data are censored, altered, or erased. It concludes with the case still unresolved, as researchers and advocates scramble to salvage, verify, and explain what remains.

---

Transcript

[A]

Hello, and thanks for downloading the More or less podcast. We're the programme that delves into data, steeps itself in statistics and mulls over mathematical mysteries. And I'm Tim Harford. In early February 2025, something strange started happening across US government websites.

[B]

Error.

[A]

Decades of data began disappearing from web pages for agencies such as the Centers for Disease Control and Pre, the National Institutes of Health and the Census Bureau. In many cases, the entire website went dark within a few days. Some 8,000 government pages and 3,000 data sets had been taken down. Since then, many have been reinstated, but some have not. A curious case indeed. As data sleuths, we have questions. Why was this data taken away? And why does any of it matter? Let us delve into the case of the missing US data.

[B]

Lots and lots of data sets disappeared from access because entire websites for agencies of the government came down because of executive orders from the president questioning what could be made available. Maggie. I'm Maggie Levenstein. I'm a professor at the University of Michigan and director of icpsr, the data consortium that's based here.

[A]

At the start of President Trump's second term, he signed a flurry of executive orders, for example, eradicating measures promoting diversity, equality and inclusion in federal government. The order targeted websites over language about minorities, indigenous peoples, race, gender, sexuality and disability. Climate change was another banned topic because

[B]

agencies weren't sure that their websites would be consistent with the policies. They just brought the entire websites down, and that meant that data that had been collected and was accessed through those websites became inaccessible.

[A]

While this was happening, the Department of Government Efficiency was plowing on with their goal of cutting $2 trillion from government spending. They haven't got anywhere close to those savings, by the way.

[B]

Lots and lots of people were fired, and those people are often people who produce data or people who provide access to data. And without those people in place, access was impossible. And there also were budget cuts that were affecting data access and data collection going forward.

[A]

Now, most of the government websites did eventually reappear. Some data sets were fully restored, but some data sets were discontinued and some reappeared, although researchers soon noticed they'd been changed.

[B]

There were actual decisions for particular data sets to alter them to obscure information about gender identity or about vaccines or about sexual behavior or about climate, that those were policy areas where there's a lot of data that the administration didn't want people to have access to that data about those questions. And so those data were removed or were altered.

[A]

When data goes missing or is altered, it can have very real life consequences. For example, some might see the eradication of the non binary gender category as purely a symbolic move. But we know that teenagers who identify as non binary are more likely to be more vulnerable to certain risks than their peers. That's something we couldn't have known without collecting the data in the past. And without the data now, are we

[B]

providing the appropriate social supports to them? It's very hard to know without the data that was being collected about the mental health of young people and their gender identity and their risk and their sexual behaviors. When we don't have that information, it's much harder to provide the resources that people need to live full and healthy lives.

[A]

The data at risk spans one areas of life and society. Recently, the Bureau of Labour Statistics announced that when calculating cpi, the Consumer Prices Index, they would stop collecting the data from three cities due to budgetary cuts.

[B]

Businesses rely on these and they rely on them to be high quality measures. When we reduce spending and the resources that the BLS has that we get lower quality measures and as a result businesses are kind of much more likely to be flying blind.

[A]

Many websites also now have government mandated disclaimers on their homepage throwing doubt on the data. For example, if you go on the cdc, the center for Disease Control and Prevention's homepage for national HIV Surveillance systems, you're met with this Per a court

[C]

order, HHS is required to restore this website as of 11:59pm Eastern Time, February 11, 2025. Any information on this page promoting gender ideology is extremely inaccurate and disconnected from the immutable biological reality that there are two sexes, male and female.

[A]

The fate of other data sets is uncertain. For example, a decades old data set that monitors birth outcomes in the US

[D]

prams prams, or the Pregnancy Risk Assessment Monitoring System, is kind of the gold standard for assessing and monitoring maternal and child health in the us. That's an area where the US has consistently lagged behind other high income countries.

[A]

That's John Cubali, a research assistant professor at the University of Michigan who specialises in epidemiology. The Department of Health and Human Services states that the purpose of prams is to reduce infant morbidity and mortality by influencing programs and policies aimed at reducing health problems among mothers and infants. As an infant is around three times more likely to die in the US than in Finland, Japan or Sweden, this seems like a pretty important data set.

[D]

It's something that is used both by policymakers and by researchers really to understand how are different aspects of maternal and child health changing over time and what populations are most impacted that's really important information that you need to have if you're going to be considering how to design interventions to address some of those challenges.

[A]

This data is collected by local states and governments and then sent to the CDC to be merged and for population weighting to be added to access the data, researchers needed to apply for access and sign a declaration stating they wouldn't share the data.

[D]

I started getting contacted by researchers and essentially these requests were just going unanswered. I submitted multiple requests myself. I reached out to that office directly. You know, I never heard any response. And so essentially any of the data from the last 10 years, so 2016 or later, that's not publicly available. And so if you did not have access to that already, you're kind of out of luck at this point.

[A]

Do we know whether they're still collecting the data?

[D]

Yeah, there is some uncertainty around that. There was an announcement, I believe, in early to mid February saying, oh, we're going to discontinue this data collection. And there was a massive backlash and outcry and there was kind of a clarification, oh, no, we're not going to discontinue this, but we're going to kind of push it back until April 2025 and then we'll start it up again. But April 1, the entire prams office was essentially fired. And so now it's really uncertain what this means for this data collection moving forward.

[A]

Since February, there's been a scramble to try and safeguard as much data as possible. Both Maggie and John are involved in this process, but snapshots of old web pages do not a full data set make.

[B]

So there's some data that you can't get to at all. There's some data that we don't know is gone yet. There's also data that's been preserved, but we don't know for sure that it hasn't been changed. And you also don't have the context, what we call metadata, the description of the data, all the things that allow you to make inference from the data.

[A]

The case remains open. Thanks to Maggie Levenstein and John Cubali. That is all we have time for this week, but if you have any questions or comments, please email in to more or lessbc.com co UK until next week. Goodbye.