Comparing Violent Crime in Honolulu to Other U.S. Cities

A question everyone wants to know about where they live: is it safe? With crime in the news and in political speeches, it can be hard to know. In fact, it can be really misleading. Excellent sociological research has shown that political speeches and news coverage can convince people that crime is actually worse than it is. So it’s good to look at the data and figure it out. In this case, we want to know how is the violent crime rate in O‘ahu relative to other places.

For this post, my main source is the 2024 midyear report by the Major Cities Chiefs Association’s Violent Crime Survey. They include 69 cities’ counts of homicide, rape, robbery, and aggravated assaults. (They use the Uniform Crime Reports (UCR) definitions of those crimes and you can read the report for definitions of each of those crimes.) The numbers are preliminary and subject to change, but they probably won’t change by much. They list Honolulu’s crime data and it matches what was listed on HPD’s dashboard in July (that is, everything from January to June).

Before getting to the comparison, I want to offer another caveat which is that we’re just comparing crime snapshots. We’re looking at the crime counts for four categories of violent crime at one point in time. This is just a “back of the envelop” analysis to get a ballpark idea of what our crime rate looks like relative to other cities.

So what do we find when we compare Honolulu to these five cities? As expected, Cleveland and New Orleans have much higher homicide counts. On that front, Honolulu is actually more consistent with Orlando, Aurora, and Wichita. Cleveland and New Orleans also have much higher rape counts; here, Honolulu is more on par with the somewhat larger cities of Wichita and Aurora and has a larger count than Orlando. In terms of robberies, Cleveland is again much higher than all the other cities, but Honolulu is also the highest of the remaining cities. Finally, of all the cities, Honolulu has the lowest count of aggravated assaults, which seems to mostly track population size (but here Orlando has a higher rate than Honolulu or New Orleans). Taking these four categories of violent crime together, I’d say Honolulu is “in the distribution” of comparably sized cities, meaning it’s not really exceptionally high or low.

Now, if you’ve been paying close attention, you’ll see that I’ve already done a bait and switch: I wanted to compare crime rates on O‘ahu to other places, but I’m now comparing Honolulu to other comparably sized cities (because that’s what data I have available). The problem with this is that HPD reports crimes for all of O‘ahu. Unlike most other cities, the police department’s jurisdiction is the city and county of Honolulu, and the county is much bigger than the city. If the population of Honolulu is about 340,000, the population of O‘ahu in 2024 is more like 950,000 according to the World Population Review. So in comparing our crime rates to these other cities, we’re actually being unfair to Oahu. We’re taking a much bigger jurisdiction and comparing it to the crime rates of much smaller jurisdictions.

While it might be unpopular to call O‘ahu the “Greater Honolulu Area” that’s typically how we would label a comparable city. I’ve lived in Toronto, Canada, and Tallahassee, Florida, and in both cases, there was the “Greater Toronto Area” (the “GTA”) and the “Tallahassee Metropolitan Area”—both of which have greater populations than the city limits. The problem is my data source doesn’t give us data by the metropolitan area. The only one of the cities in my comparison that has a police department with a jurisdiction beyond the city limits is Charlotte, whose police department’s jurisdiction includes Mecklenberg County. So now Charlotte’s population is more like 1.1 million, so a bit bigger than Oahu, but a useful comparison.

Okay, so we’ve got our small cities that are comparably sized to Honolulu proper, but are not a fair comparison for HPD’s jurisdiction of the whole island. We’ve got our larger cities that are comparably sized to O‘ahu’s population, but still not a fair comparison because they’re going to be all city and not the mix of urban, suburban, and rural that chracterizes O‘ahu. And then we’ve got Charlotte-Mecklenberg, which has a larger population (although not by a huge amount) that might be a closer comparison to O‘ahu because it’s another city-county jurisdiction. So what do the data tell us?

In the above graph, I included Honolulu twice for comparison—once as the city population and once as the county population—but my main focus is Honolulu County.

Let’s start with our closest comparison of Charlotte-Mecklenberg: in homicides, robberies, and aggravated assaults, Charlotte-Mecklenberg has counts that are quite a bit higher than Honolulu County. In rapes, Honolulu County has more even though Charlotte-Mecklenberg has a higher population.

In terms of our more comparably sized cities (again, still an unfair comparison in that they will put Honolulu in a better light), Honolulu is generally at the lower end for all four cirm categories. Again, this is what we would expect given the mixed jurisdiction of Honolulu County. In a few categories, Honolulu County is on par with comparably sized cities (so much so that my marker labels are overlapping): Columbus and Honolulu County are basically on par when it comes to robberies and aggravated assaults while San Jose and Honolulu county are basically on par when it comes to homicides and rapes. The other cities all have higher counts as expected.

So where does that leave us? I did one final analysis. Although it’s not a good idea to run regression analyses on small datasets like this one where I have 12 data points, I was curious. A regression analysis is a statistical analysis that finds relationships in the data and gives us a sense of how meaningful differences are. For each of the four crime types, I ran a regression of the count on population size and a “dummy variable” for Honolulu County. The idea is to see whether, controlling for population size, Honolulu stands out from the crowd.

Population is a good control variable typically when looking at crime rates, with a few caveats. First, when we’re dealing with relatively small counts (as in homicide and rape), the data are a bit noisy and you can end up with big fluctuations. So just looking at the first row of plots in the last combination of graphs, you can see there isn’t a strong relationship between crime count and population. That relationship starts to show up as you get into crimes with larger counts: it’s sort of there in the robberies graph (with Cleveland as a big outlier) and more clearly in the aggravated assaults graph. The second caveat is that outliers can have a big impact on the calculation of the overall relation, but you generally want to be really careful about dismissing them because they still convey a lot of information. So I’m leaving them in.

In my four regression analyses (again, we have to be cautious because I’ve only included 12 data points), the population variable is only significant in one: aggravated assault. That’s consistent with an eye-ball inspection of the above graphs. You can actually see the relationship. In that regression as well, Honolulu is at about 1.88 “standard errors” from the average city included in the analysis. This is close to our generally recognized threshold of 2 standard errors, but not quite, so we would say Honolulu is not a significant variable, meaning it doesn’t stand out. And the aggravated assault regression is the most significant the Honolulu variable gets, meaning it doesn’t really stand out when we’re controlling for population. Now, again, this is only with 12 cities; with more data, we would probably find population significant and we might find that Honolulu is significant then, too. But for now, it’s in the distribution. But there’s one more point: the coefficient on the Honolulu variable in all four regressions is negative. So while Honolulu is not significantly different from the other cities, it does tend to have a lower crime rate than we would expect given its population size. That’s also what we would expect given this challenge we have in finding a comparable jurisdiction to Honolulu: Honolulu County’s counts are expected to be smaller than comparably sized cities because they tend to be more city like and thus have higher crime rates; we should expect Honolulu County’s crime rate to be sort of diluted by its more suburban and rural parts of its jurisdiction.

So my conclusion is that Honolulu is sort of a difficult jurisdiction to compare, but with all the caveats we’ve discussed, I’d say Honolulu’s crime rate is generally “in the distribution” of what we would expect, but if we’re going based on population, it tends to be a bit lower than we would expect.

A future analysis will compare all the cities we have data on and rerun the regressions to see what we can find with a somewhat larger dataset. Other analyses will add in more crime counts from other periods, starting with mid-year 2023, and we’ll update it when more reports come out from the Major Cities Chiefs Association. Finally, a future post will also look through the academic literature to see what experts on crime rates tell us about city-based comparisons that include Honolulu and see what they say.