Fraud has become quite politicized in the United States the last few years. We had a poorly-calibrated federal initiative led by a charismatic tech entrepreneur which believed it would unearth trillions of dollars of fraud that focused substantial effort on large programs which are comparatively fraud-resistant. Across the aisle, we have reflexive dismissal that fraud happens in social programs, which functions as air cover for scaled criminal operations which loot many varied social programs [0] and are sometimes run out of geopolitical adversaries of the U.S. including by ambiguously-retired members of their clandestine services.
[...]
Minnesota has suffered a decade-long campaign of industrial-scale fraud against several social programs. This is beyond intellectually serious dispute. The 2019 report from the Office of the Legislative Auditor (a non-partisan government body) makes for gripping reading. The scale of fraud documented and separately alleged in it staggers the imagination: the state’s own investigators believed that, over the past several years, greater than fifty percent of all reimbursements to daycare centers were fraudulent. (Separate officials took the… novel position that they were only required to recognize fraud had happened after securing a criminal conviction for it. Since they had only secured a few criminal convictions, there was no way that fraud was that high. Asked to put a number on it, repeatedly, they declined.)
[...]
Fraudsters are liars and will cheerfully mouth any words they believe will absolve them of their crimes. If an accusation of racism gets one a free pass to steal hundreds of millions of dollars, they will speciously sue you alleging racial discrimination. That empirically worked in Minnesota. The OLA takes explicit notice of this multiple times, a coordinator for the fraud operation is on record explicitly explaining the strategic logic of accusations of racism, and a judge was even moved to make an extraordinary statement to clarify that the bad-faith lawsuit alleging racism did not achieve success through the formal judicial process but rather through the voluntary compliance of governmental actors shamed by its allegations.
[...]
As mentioned, there is enormous visceral distaste for the conclusion that a particular fraud ring operates within a particular community. This is quite common. You should expect to find circumstances which rhyme with it when conducting effective fraud investigations. You should not abandon fraud investigation when you chance upon this.
People assume a level of ethical fraughtness here which is not warranted. You would, if doing ethnographic work on perfectly legitimate businesses across industries, routinely discover ethnic concentration rather than population-level representation everywhere you looked. The Patels run the motels. One doesn’t need to adopt grand theories about how certain groups are predisposed to becoming pharmacists or startup employees or line cooks; simple microeconomic reasoning explains reality easily. Firms hire the people they already know, like, and trust. That will routinely include friends and family, who are going to be much more like the founding team than they are like randomly drawn members of the population. This is the default outcome.
Fraudsters do have one structural factor here. Everyone wants to trust their coworkers. Fraudsters need to trust their coworkers will be loyal even upon threat of prison time. That necessarily selects for tighter bonds than the typical workplace. Madoff was a family affair, SBF was in an on-again off-again romantic relationship with a chief lieutenant, and neither of those facts is accidental or incidental.
That’s the other ethical dimension of being other-than-blind to concentration: so-called affinity frauds do not merely recruit fraudsters from affinity groups. They recruit victims from affinity groups. Madoff mobilized the social infrastructure of the Jewish community in New York and Palm Beach to find his marks. Community members certainly did not intend their charitable foundations to be looted by a fraudster. It was an emergent consequence of trust networks.
[...]
Responsible actors in civil society have a mandate to aggressively detect and interdict fraud. If they do not, they cede the field to irresponsible demagogues. They will not be careful in their conclusions. They will not be gentle in their proposals. They will not carefully weigh consequences upon the innocent. But they will be telling a truth that the great and the good are not.
The public will believe them, because the public believes its lying eyes.
Insightful article, as always, but I'm looking forward to reading the book mentioned, Lying for Money. I was really surprised to hear that the government doesn't have fraud analysts the way...
Insightful article, as always, but I'm looking forward to reading the book mentioned, Lying for Money.
I was really surprised to hear that the government doesn't have fraud analysts the way companies do! Back when I worked for a bank, I remember coming across one particular person that was clearly defrauding one of COVID relief programs and raised it internally, but it sounds like it's likely that was never actually communicated to the government or turned into anything?
Well, the government does, but it’s not a super controversial claim to argue that companies are much better at fraud detection and prevention than basically any government. There are basically two...
I was really surprised to hear that the government doesn't have fraud analysts the way companies do!
Well, the government does, but it’s not a super controversial claim to argue that companies are much better at fraud detection and prevention than basically any government. There are basically two issues here. The first is that as McKenzie (the original author) mentions, blacklisting people without concrete proof goes directly against the deeply-held convictions of the better sort of civil servant. McKenzie takes a pretty dim view of this —
An accusation is not a conviction, and absent clear authority to impose consequences in a new program, an actor convicted at enormous societal cost emerges to a new program officer as tabula rasa, equal in moral worth to any randomly chosen citizen. I will not argue that Mastercard has better moral intuitions than the Founding Fathers. I would, however, happily suggest that the government not assume that the Constitution contains emanating penumbras obligating it to be repeatedly taken advantage of by the same people in the same fashion. We are not forbidden object permanence.
— but I think he’s being a little too glib. Yes, the way the American government is structured (no blacklists without a conviction) makes it hypervulnerable to fraud, but there are massive upsides as well. It’s one of the miracles of western civilization that western governments generally work the way they do. In many developing countries, bureaucrats primarily exist to funnel state resources to their own in-groups. Sometimes the in-group is clan, sometimes it’s religious compatriots, sometimes literal kin, but what it almost never is is McKenzie’s tabula rasa. It’s actually pretty unusual, in the global context, for a government to view all its citizens equally. My point is that having a civil service which hesitates to blacklist anyone without a conviction is something to be treasured, not something to discard lightly.
The second challenge is that Mastercard et al. have real motives to prevent fraud: they lose money from fraud. The fraud investigators can get fired if they let too much fraud through. The company itself can be held liable by the American court system, both criminally and civilly, if they miss certain types of fraud.
None of that is true for bureaucrats and civil servants. The tap of money is either on or off, and their actions and bearing have zero impact on that. They won’t get fired - elected officials can often be held accountable by the public, compare how Minnesota Governor Walz just ended his re-election campaign - but the people who would actually need to do the work to identify fraud will never be touched by the scandal. And the government doesn’t hold itself liable. Ultimately, it’s an incentives problem.
(Large federal programs like Medicare/Medicaid do much better at fraud prevention than local programs, mostly due to economies of scale.)
I think I just fundamentally disagree with you on the first point. "Beyond a reasonable doubt" is a good threshold for incarceration, but is completely the wrong threshold for access to government...
I think I just fundamentally disagree with you on the first point. "Beyond a reasonable doubt" is a good threshold for incarceration, but is completely the wrong threshold for access to government contracts. For civil matters, we already accept a 50% threshold. After all, I assume that there are other folks in the same industry competing for those contracts, why award money to anyone but the least risky actors?
I'd completely agree with you for things like personal disability benefits or the like. It absolutely must be beyond a reasonable doubt in those cases.
Another great share. Quite long! A few thoughts: I found this pretty difficult to read. There were a lot of thoughts which included sort of tangential connections or were not particularly...
Another great share. Quite long!
A few thoughts:
I found this pretty difficult to read. There were a lot of thoughts which included sort of tangential connections or were not particularly well-explained. I don’t say that with malice; many of my own articles are (unfortunately) like this and it takes a lot of editing and revision to make them not that way. The author struck me as an extremely smart guy who’s often a step or two ahead of most people — sort of like one of the math professors who writes a complicated equation on the board and then says “proving this is, obviously, trivial” and you’re sitting in the back like “the fuck it is!” but it’s too late because the professor has already moved on and your TA is way smarter than you and could probably explain it but barely speaks your native language and now you’re going to fail the exam for sure.
Here’s a representatively confusing paragraph:
With a good machine learning practice, you can increase data ingested but decrease the burdensome formal application/etc requirements. This is in no small part because those data points are less probative (they are under the direct control of the attacker and announce that they will be scrutinized). But it bears a dividend: if you better control fraud, and can successfully demonstrate that to the public and legislators, you can decrease application burden and perhaps even widen eligibility criteria. Those are both in the direct interests of potential marginal beneficiaries. [emphasis in original]
What does he mean that the data points announce they will be scrutinized??? What is a potential marginal beneficiary???
Even with that complaint, I will note, this was still well worth reading.
In part this is worth reading because it starts so strong:
(As a sidenote: one has to be able to hold two thoughts simultaneously about fraudulent operations. They can be sophisticated with respect to exploiting sociopolitical cleavages in their targets while also being comically inept at faking evidence elsewhere, such as having a single person write dozens of adjacent rows in a sign-in sheet. This routinely surprises observers and it should not surprise them. The financial industry also has a division of labor in it. The person architecting the fraud department’s standard processes is well-paid, well-educated, and routinely brings crossdisciplinary expertise to bear. A Fraud Analyst I, on the other hand, bears a lot of similarity to a call center employee in terms of compensation, education, and permitted amounts of agency.)
This is the kind of thing that makes sense immediately after thinking about it, but which you would never come up with yourself without experience in the antifraud industry. The whole article - kinda confusing though it may be - is stuffed with interesting asides like this, and it more than makes up for any difficulty in reading.
And finally,
Fraud has become quite politicized in the United States the last few years. We had a poorly-calibrated federal initiative led by a charismatic tech entrepreneur which believed it would unearth trillions of dollars of fraud that focused substantial effort on large programs which are comparatively fraud-resistant. Across the aisle, we have reflexive dismissal that fraud happens in social programs, which functions as air cover for scaled criminal operations which loot many varied social programs [0] and are sometimes run out of geopolitical adversaries of the U.S. including by ambiguously-retired members of their clandestine services.
This strikes me as unambiguously true. I got in a pretty serious argument with some family members when news of the Minnesota fraud broke, because they viewed the entire thing as fake news created by US President Trump to try to undermine his political rivals. I think that’s a huge shame and emblematic of how the partisan divide in the United States is pulling the country apart more broadly: one should be able to think that President Trump is a nightmare and also think it’s bad that more than 50% of all daycare reimbursements in Minnesota went to children who didn’t exist. And accusing everything you don’t like of being Fake News is just as bad when democrats do it.
When you fill out a form, you the person who gave you the form to carefully read over your responses and scrutinize them. On the other hand, you don't have control over and may not even be aware...
What does he mean that the data points announce they will be scrutinized???
When you fill out a form, you the person who gave you the form to carefully read over your responses and scrutinize them. On the other hand, you don't have control over and may not even be aware of other data points being captured. For example, in an e-commerce context, how long does it take you to go from your first visit to a web store to checking out? I imagine the difference between fraud and legitimate customer is fairly wide.
What is a potential marginal beneficiary???
I understood this to mean someone who barely meets the requirements or is otherwise towards the bottom of the pile when ranked by most qualified. Since fraudulent actors will lie to appear more qualified, this means barely-qualified folks would have gotten unfairly excluded from consideration.
I definitely agree that his writing is a bit idiosyncratic, and I'd say it's targeted towards the audience he is familiar with. I personally don't have too much trouble with it, but I have read a lot of his writing, and I've worked in fintech, so I'm basically right in the middle of his target audience.
From the article:
[...]
[...]
[...]
[...]
Insightful article, as always, but I'm looking forward to reading the book mentioned, Lying for Money.
I was really surprised to hear that the government doesn't have fraud analysts the way companies do! Back when I worked for a bank, I remember coming across one particular person that was clearly defrauding one of COVID relief programs and raised it internally, but it sounds like it's likely that was never actually communicated to the government or turned into anything?
Well, the government does, but it’s not a super controversial claim to argue that companies are much better at fraud detection and prevention than basically any government. There are basically two issues here. The first is that as McKenzie (the original author) mentions, blacklisting people without concrete proof goes directly against the deeply-held convictions of the better sort of civil servant. McKenzie takes a pretty dim view of this —
— but I think he’s being a little too glib. Yes, the way the American government is structured (no blacklists without a conviction) makes it hypervulnerable to fraud, but there are massive upsides as well. It’s one of the miracles of western civilization that western governments generally work the way they do. In many developing countries, bureaucrats primarily exist to funnel state resources to their own in-groups. Sometimes the in-group is clan, sometimes it’s religious compatriots, sometimes literal kin, but what it almost never is is McKenzie’s tabula rasa. It’s actually pretty unusual, in the global context, for a government to view all its citizens equally. My point is that having a civil service which hesitates to blacklist anyone without a conviction is something to be treasured, not something to discard lightly.
The second challenge is that Mastercard et al. have real motives to prevent fraud: they lose money from fraud. The fraud investigators can get fired if they let too much fraud through. The company itself can be held liable by the American court system, both criminally and civilly, if they miss certain types of fraud.
None of that is true for bureaucrats and civil servants. The tap of money is either on or off, and their actions and bearing have zero impact on that. They won’t get fired - elected officials can often be held accountable by the public, compare how Minnesota Governor Walz just ended his re-election campaign - but the people who would actually need to do the work to identify fraud will never be touched by the scandal. And the government doesn’t hold itself liable. Ultimately, it’s an incentives problem.
(Large federal programs like Medicare/Medicaid do much better at fraud prevention than local programs, mostly due to economies of scale.)
I think I just fundamentally disagree with you on the first point. "Beyond a reasonable doubt" is a good threshold for incarceration, but is completely the wrong threshold for access to government contracts. For civil matters, we already accept a 50% threshold. After all, I assume that there are other folks in the same industry competing for those contracts, why award money to anyone but the least risky actors?
I'd completely agree with you for things like personal disability benefits or the like. It absolutely must be beyond a reasonable doubt in those cases.
Another great share. Quite long!
A few thoughts:
Here’s a representatively confusing paragraph:
What does he mean that the data points announce they will be scrutinized??? What is a potential marginal beneficiary???
Even with that complaint, I will note, this was still well worth reading.
This is the kind of thing that makes sense immediately after thinking about it, but which you would never come up with yourself without experience in the antifraud industry. The whole article - kinda confusing though it may be - is stuffed with interesting asides like this, and it more than makes up for any difficulty in reading.
This strikes me as unambiguously true. I got in a pretty serious argument with some family members when news of the Minnesota fraud broke, because they viewed the entire thing as fake news created by US President Trump to try to undermine his political rivals. I think that’s a huge shame and emblematic of how the partisan divide in the United States is pulling the country apart more broadly: one should be able to think that President Trump is a nightmare and also think it’s bad that more than 50% of all daycare reimbursements in Minnesota went to children who didn’t exist. And accusing everything you don’t like of being Fake News is just as bad when democrats do it.
When you fill out a form, you the person who gave you the form to carefully read over your responses and scrutinize them. On the other hand, you don't have control over and may not even be aware of other data points being captured. For example, in an e-commerce context, how long does it take you to go from your first visit to a web store to checking out? I imagine the difference between fraud and legitimate customer is fairly wide.
I understood this to mean someone who barely meets the requirements or is otherwise towards the bottom of the pile when ranked by most qualified. Since fraudulent actors will lie to appear more qualified, this means barely-qualified folks would have gotten unfairly excluded from consideration.
I definitely agree that his writing is a bit idiosyncratic, and I'd say it's targeted towards the audience he is familiar with. I personally don't have too much trouble with it, but I have read a lot of his writing, and I've worked in fintech, so I'm basically right in the middle of his target audience.