They got their data from the Ashley Maddison hack and then cross referenced that with police records of people who got caught. I'm not sure this is a very valid study.
Using the Ashley Madison data is basically what we would call a "reduced form estimate". The idea is that someone in the Ashley Madison data is more likely (or at least not less likely) than a comparable individual not in the Ashley Madison data to cheat on their spouse. They would like to measure whether or not people actually cheat, but they can't, so they measure this instead. This kind of reduced form estimate is a fairly common thing when you can't measure actual treatment compliance, but you can measure an "invitation to treat". We, in fact, know that Ashley Madison wasn't actually used for cheating because there were no (zero) real women on the site. Extensive analyses of the data leak suggest basically that men signed up and were faced with phony bots. But the idea should still be that the choice to sign up on Ashley Madison reveals an attempt to cheat.
The other major challenge for the design is that the control group of CEOs need to have what we would think are baseline similar propensity to commit fraud etc in all respects except for the choice to cheat. So, at the very least I would want to do a matching, weighting, or propensity score design that accounts for differential company characteristics (sector, size, age, any characteristics we would a priori affect propensity to engage in financial malfeasance.) If what we learn is that people who run shady payday loan companies also cheat on their spouses, then this is maybe not interesting. But if what we learn is that among F2000 blue chip companies in similar sectors, cheating spouses are more likely to be cheating CEOs, this is more interesting, right?
I agree with you that it's possible that the degree of difficulty / competence factor might be a confounder. Probably they could compensate for this by looking at something other that rate of conviction for malfeasance, rather by modelling types / degrees of malfeasance and degree of difficulty for getting caught.
Finally, the casual inference here is impossible. So it's possible that they have a good design to answer a question they're not asking. I would assume the default assumption is not that cheating on one's spouse causes one to cheat on financial things, but rather that both are outcomes which flow from an underlying propensity for deceit. In this case, the main takeaway is that early indicators of deceit may allow us to head off or catch financial malfeasance. That's interesting, but maybe not the exact question they're asking.
I'd be really interested in reading the final thing because I could see it going either way. I think it's an interesting proposal for a design. Would love to see a pre-print
There were numerous bots and fake accounts to attempt to make it seem as though there were more women on it, and I'm sure the gender ratio was absurd, but I seriously doubt the number was 0.
Wikipedia discussion of the bots : https://en.wikipedia.org/wiki/Ashley_Madison#Fake_female_bot...
This is my recollection. I can cite no source for evidence.
There is no person that is so intelligent that he/she can cheat for long and not eventually get caught. If people want to know the truth about a person badly enough, they will.
How do you know? Wouldn't the very best never get caught and therefore we'd never know?
How do you explain all of the unsolved crimes?
https://www.npr.org/2015/03/30/395069137/open-cases-why-one-...
Getting “caught” so far has not mattered and only resulted in further consolidation of power. Correlating that so far with infidelity, the only takeaway is that you should put all your mistresses under an NDA, which also wouldn’t have come up if he didn’t run for office resulting in the feds raiding his lawyer’s office and indicting the lawyer due to conduct that became part of the national spotlight! These are things that don’t happen, so all you have to assume is that the President wasn't the genius that invented these tactics and merely used tools available.
The main point I am making here is that this is enough information to rationalize proving an absence:
There are many many people like him at least in an even distribution throughout society. Let alone people with a network of enablers.
In fact, cheating has little to do with how intelligent you are and how well you are at containerizing different parts of your life so that they never cross. It is ridiculously easy to cheat if you travel a lot on your own.
See also psychopathy and sociopathy.
Cheating is not associated with low intelligence but with a corrupt character.
From previous experience consuming public record data sources programmatically, a proof of concept would take no more than a full day of work. I am not suggesting one do this, only that pandora's box is already open.
In fact, the latter ones are so ubiquitous, I think you would be hard-pressed to find a person in a major US city in their 20s who was single for a period of time in the past 5 years and haven't used one of those apps.
And soon, it will the premise of everyday life in the US. Whoopee!
Or said in one word: blackmail.
There's no sugar coating it. You'd be collecting data points and analyzing them to pursue your own justice without due process. That's quite a dangerous thing to claim to want to do in the name of "insider risk management".
I don't support blackmail in any form, full stop. I do want to know if you're going to embezzle millions of dollars before you do, as do shareholders or customers (if you're managing assets) of the organization.