Bear in mind that social security fraud is a major problem in many countries and impossibly old people is a usual indicator. The famous "blue zones" that were once studied for their long lived people are now believed to mostly be an artifact of undetected pensions fraud.
If that's what he was trying to ask, then I have, as detailed above, some notes on the attempt. But the myth that Social Security numbers were ever designed to work even half-assedly well as a single national ID number will never die, I suppose.
> Bear in mind that social security fraud is a major problem in many countries and impossibly old people is a usual indicator. The famous "blue zones" that were once studied for their long lived people are now believed to mostly be an artifact of undetected pensions fraud.
I can't really evaluate Newman's preprint [1] with much confidence, but it's only been out since September, and the Ig Nobel isn't peer review. Given also that the US state with the largest count of "supercentennarians" in Newman's data, Iowa, has a whopping 37 [2] of them, you're stretching this piece of un-peer-reviewed research far past anything it could reasonably be asked to support - hence, I suspect, the "are now believed" passive-voice weaseling in your claim quoted above.
[1] https://www.biorxiv.org/content/10.1101/704080v3
[2] As given "1.25e-05" (p. 5, ibid.) or 12.5 per million, for a population just over three million. I've taken the liberty of rounding down on the assumption we are not meant to be counting fractional supercentennarians.
Peer review hardly means anything in these sorts of fields but seeing as you asked, the same guy has previously published peer-reviewed papers on fraud in Blue Zones research. Here's a press release with some references:
https://www.ucl.ac.uk/ioe/news/2024/sep/ucl-demographers-wor...
> Dr Newman has previously disproved a 2016 study published in Nature on extreme-age research that accidentally rounded off a substantial amount of its data to zero. His peer-reviewed paper demonstrated that if corrected, this error eliminated the core findings claiming that human lifespan had a defined limit. Then, Dr Newman also countered a 2018 paper which made the opposite claim and, in the process, demonstrated a theoretical result predicting that patterns in old-age data are likely to be dominated by errors.
In investigating this theory, Dr Newman demonstrated fundamental and comedic mismatches between longevity claims and observed patterns. In the process, Dr Newman revealed that the well-publicised “Blue Zones” claims for the secrets of longevity are infallibly flawed.
Dr Newman showed that the highest rates of achieving extreme old age are predicted by high poverty, the lack of birth certificates, and fewer 90-year-olds. Poverty and pressure to commit pension fraud were shown to be excellent indicators of reaching ages 100+ in a way that is ‘the opposite of rational expectations’.
I understood GP as saying that this data is the best, cleanest data available. If nothing better exists, then record what you have. If that means some fields are blank, well then they're blank - you'll have to deal with this for country-sized databases, one way or another.
Im genuinely curious about this, would you mind sharing references?
Maybe the data was there initially, but was deleted or lost for technical reasons, for example the date was incomplete, e.g. only the year of birth was known, but not the exact day and then was dropped to null as invalid during some kind of upgrade.
Maybe the data was not possible to enter into the system for technical reasons -- somebody didn't have access, something was down, the next shift was supposed to enter it once the system was back up. Maybe the same, but for process reasons, for example setting the date requires attaching the birth certificate and the birth certificate is lost.
Maybe the data was intended to be entered, but somebody forgot to actually do it, was waiting for birth certificate to be produced, but approved the benefits provisionally.
Maybe the birth certificate was shown to the official, but was of the wrong kind (handwritten extract, foreign document without consular legalization, damaged, illegible or incomplete date), so it was legit enough to start paying benefits but not legit enough to upload to the server. Maybe the date was conflicting with other records somehow. Maybe there was a court decision involved which determined eligibility for benefits but didn't determine exact date of birth.
The key point here -- the system allows for this kind of inconsistency (benefits are being paid, but the date is not present) and relies on a human process to eventually reach a consistent state, but the process failed and nobody really cares to rectify it as long as benefits are being paid.
That's assuming the benefits are actually paid to those people and the organization is not aware of this data issue and not investigating those people for fraud right now.
In an organization big enough there is always this kind of stuff happening (see "seeing like a bank"), which is tolerated as long as statutory task is not compromised.
It could also be that entries in the DB that are complete are not exactly correct or even belong to real people and of course there is fraud and corruption going on at least somewhere.
1. Run a complex process requiring human intervention to determine age where it is hard to determine
2. Enter the result of step 1 in the DB
And it's pretty obvious that 99.99% of the cost of this is coming from step 1. So I find it completely implausible that just tracking the results here, when a single error could potentially result in a loss of hundreds of thousands of dollars, is not cost effective.
It is not a major problem, it is a minor problem blown completely out of proportions by right-wing propaganda to cover much bigger problems caused by power-grabbing, tax-evading plutocracy.
> a standard process to locate and clean missing data through investigation
What would that process look like assuming you are e.g. trying to establish a DOB of a person born long time ago in a country that does not exist anymore under wartime conditions? Would that be cost effective?