That's why for such studies, the crime rate is assessed by victim surveys (free of police bias) and the difficult-to-fudge homicide rate. The results remain consistent.

Edit: For an example of a victim survey, see https://www.bjs.gov/content/pub/pdf/cv18.pdf Table 12. That's violent crime, but I believe not including homicide, as it's based on surveyed victim responses. Note that the data is collected independent of whether the crime was reported to police (https://www.bjs.gov/index.cfm?ty=dcdetail&iid=245#Methodolog...)