Earlier quoted context omitted.
I'd be really interested to know whether a significant amount of fraud and fraud attempts involve devices with root or non-stock operating systems. This has always struck me as a matter of checkbox compliance rather than a commonly-exploited attack vector, though I'll grant that's partially because few people actually use such devices.
Intuitively I'd say no, there's no way it's a significant amount of fraud. Number one because, as you said, it's rare, but number two because you just don't need a rooted phone to scam someone. You can very easily scam people on perfectly legitimate phones and with perfectly legitimate apps.
Keyloggers would be considered a form of fraud, right? Customers can be protected by not allowing rooted phones which may contain malware and steal credentials, but then again Windows is a nightmare for security and nobody is banning banking from Windows.