Homicides are useful because if you use anything else the "nuh uh, it's impossible that (for example) NYC is safer than lots of other places" contingent will just point out (correctly!) that there are massive problems with crime stat collection and so the ones in the statistically-not-that-bad big cities must be faked [EDIT: or, at least, very inaccurate for other reasons] and the ones in the statistically-worse smaller cities and towns are simply not, or are less, faked (this part doesn't necessarily follow, but they're not wrong that it might be true)
Homicides are harder to hide, and are far less likely to go unreported, than other crimes.
The other useful route is victimization surveys. They introduce other issues, but the trade-off is well worth it for evading stats-juking or underreporting. Not a cure-all depending on the topic[1] but pretty good. I believe you'll also find NYC and some other perceived-by-many as especially-risky cities fall off the higher part of top-crime-city lists with this metric applied, when we lower the bar of what we're calling a "city" to include small and mid-sized ones.
Here's a fun one (check the study linked near the top if you want the source—the article's kinda ass, but the charts are handy) that attempts to use per-capita cost of crime to sort of normalize for overall severity of crime ("is double the overall crime rate really worse if 100% of those are pickpocketing and 100% of the 'lower'-crime city crimes are murder?", to illustrate the problem by exaggerated example) and does include smaller cities (broken out from the 300ish ones the study classifies as "large"):
(Some large cities are very dangerous and have crime problems! This source confirms that and I do not deny it! The goofy part is that, if you live in certain places and around certain sorts of people, they'll get super-worried about you when travel to ones that simply aren't that bad, and these same folks would never think to worry if you tell them you're visiting, say, Saginaw, MI, or Pine Bluff, AR, but, apparently, they should worry—hell, this whole illustrative exercise probably plays out in Saginaw regularly, I guarantee there are people there who tell their families "hey, cool news, I'm going to NYC for a week" and many of their family members get really worried about their safety)
[1] Picking up on sexual assaults with victimization surveys can be especially tricky. There are two main factors at play, one of which is simply that people are a less likely to report a sexual assault even in an anonymous poll than they are other crimes, and the other is that you'll get super-different results if you ask people whether they've been the victim of a sexual assault or rape, versus asking them whether they've experienced something that, as described, definitely legally qualifies as a sexual assault or rape—that is, a lot more people have factually experienced something that qualifies as SA or rape, than regard themselves as having been victims of those crimes, so... which thing are you trying to capture? Incidence rate of behavior that qualifies, or rate at which people regard themselves as being victims of that behavior? You need a different approach for each.
Homicides are harder to hide, and are far less likely to go unreported, than other crimes.
The other useful route is victimization surveys. They introduce other issues, but the trade-off is well worth it for evading stats-juking or underreporting. Not a cure-all depending on the topic[1] but pretty good. I believe you'll also find NYC and some other perceived-by-many as especially-risky cities fall off the higher part of top-crime-city lists with this metric applied, when we lower the bar of what we're calling a "city" to include small and mid-sized ones.
Here's a fun one (check the study linked near the top if you want the source—the article's kinda ass, but the charts are handy) that attempts to use per-capita cost of crime to sort of normalize for overall severity of crime ("is double the overall crime rate really worse if 100% of those are pickpocketing and 100% of the 'lower'-crime city crimes are murder?", to illustrate the problem by exaggerated example) and does include smaller cities (broken out from the 300ish ones the study classifies as "large"):
https://thehill.com/homenews/nexstar_media_wire/4366668-thes...
(Some large cities are very dangerous and have crime problems! This source confirms that and I do not deny it! The goofy part is that, if you live in certain places and around certain sorts of people, they'll get super-worried about you when travel to ones that simply aren't that bad, and these same folks would never think to worry if you tell them you're visiting, say, Saginaw, MI, or Pine Bluff, AR, but, apparently, they should worry—hell, this whole illustrative exercise probably plays out in Saginaw regularly, I guarantee there are people there who tell their families "hey, cool news, I'm going to NYC for a week" and many of their family members get really worried about their safety)
[1] Picking up on sexual assaults with victimization surveys can be especially tricky. There are two main factors at play, one of which is simply that people are a less likely to report a sexual assault even in an anonymous poll than they are other crimes, and the other is that you'll get super-different results if you ask people whether they've been the victim of a sexual assault or rape, versus asking them whether they've experienced something that, as described, definitely legally qualifies as a sexual assault or rape—that is, a lot more people have factually experienced something that qualifies as SA or rape, than regard themselves as having been victims of those crimes, so... which thing are you trying to capture? Incidence rate of behavior that qualifies, or rate at which people regard themselves as being victims of that behavior? You need a different approach for each.