That's a tricky number to tease out. One must ask how "actual amount of crime" is measured.

Oftentimes, one finds a feedback loop between arrest rate counting and instances of crime counting that means all kinds of confounding factors enter into the number (including distribution of police patrol resources and likelihood police will let an infringement off with a warning).

It's not very tricky at all. A random sampling of citizens are asked if they experienced violent crime (including instances not reported to police), and if so, to describe the offender (ethnicity being just one component).[1]

Year after year the random samples of victimization correlates with arrest rate by ethnicity.

[1]: https://www.bjs.gov/index.cfm?ty=dcdetail&iid=245#Methodolog...

I wonder if any of those surveyed included banking crimes in their list that affected them.

It’s a survey of violent crime. So if by banking crime you mean getting mugged, yes, but if you mean a guy from India calling your grandma and saying he’s the IRS and needs her banking info, then no.