When 0.5% of all counties, with 0.8% of the people, account for 25% of the homicides, you must pay attention.

When those same counties, on average, have five times the murder rate than the average of all other counties, you must pay attention to details.

When those 15 counties are also known to be dominated by a single city with a reputation for gang violence, you must pay attention to why people in those cities are different than the rest of the country.

But, what makes those 15 counties uniquely violent?

Major Takeaways

• 1% of counties with any homicides have 24.9% of the homicides, but only 8.4% of the population in counties where gun homicides occur.
• The states with these Hot 15 Counties have evenly distributed gun law scores ranging from A- to F by a gun control advocacy group.
• A number of factors (including population density, teen-age poverty, and more) help explain their disparity.

It is a fast fall-off

Every year we see that homicides are geographically concentrated. In the year of this study, 2018 ¹, a mere 15 counties out of 3,124 accounted for just under 25% of all murders.

But some counties have no homicides, while others have lots of dead people. What we explore here includes:

• Are some counties more deadly than their populations alone would predict?
• What is different about those counties that might explain their murderous ways?

The answers are at times surprising, unsurprising, and a bit frustrating.

One reality is that counties with big metropolitan hubs (i.e., huge cities) tend to be at the top of the list. All but one of the Hot 15 are classified as “Large Central Metro” counties, which means they are “counties in metropolitan statistical areas (MSA) of 1 million or more population that contain the entire population of the largest principal city of the MSA or are completely contained in the largest principal city of the MSA.”

Cook County, Illinois sits atop the Hot 15 counties for an excess of their share of homicides minus their share of the population (herein, we call this the “gap,” which expresses excess homicides over what population alone would predict). Looking at just the state of Illinois, we see Cook County – which includes Chicago – having nearly 73% of the of the state’s homicides but only 41% of the state’s population.

Cook County has a lot of people, so you expect more murders there. When we look at the gap between their share of the state’s firearm murders and population, we see the disturbing reality. Most counties adjoining Cook have negative gaps (meaning their firearm homicide rate is lower than their population would predict) while Cook is amazingly high. Chicago bleeds into the rest of Cook County, but not beyond.

The same applies for Baltimore (though they are an independent city, not within the jurisdiction of a county), Philadelphia within Philadelphia County in Pennsylvania, Detroit within Wayne County Michigan… and on down the list. 15 big cities – from Chicago down to Dallas – have large core cities, outsized homicide issues, and 25% of America’s murders.

But it levels off pretty fast after that.

 

click for a larger, sharable version

Looking at the homicides-population share “gap” rates, we see that soon after the Hot 15 counties, there is a profound leveling-off of the gap rate that carries through to most of the rest of the nation. One oddity in the data is that there are a few “large” population counties with low homicide rates, which causes a short plunge in the gap rate on the right side of the chart (here’s looking at you, Wake County, North Carolina and San Diego County, California).

To make this a bit more vivid, let’s zoom in on the on the left side of that chart and look at the troublemakers. The degree of leveling is transparent. Between the top and bottom of the Hot 15 counties is a four-fold drop in the homicide-population gap. We blow past another 39 counties before seeing another four-fold drop, and after that 94 counties until the next four-fold decrease.

In short, there is something specifically different about the Hot 15 that must explain why their share of homicides is far above both their share of the population and the gap nearly every other county in the country has.

Why that is, is confounding.

Confound Those Variables!

In a perfect world we would have pristine data on all demographic, sociographic, and criminology variables that might possibly explain why Cook County, Illinois (1.7% of the country’s population and 4.3% of its firearm homicides) is different from Jasper County, Texas (halfway down the list, 0.01% of the population and 0.01% of the gun homicides).

But we live in an imperfect world.

We pulled data from the Census Bureau, the CDC, and the FBI to gather 24 variables, mostly common (poverty, race, population density) and some not so common (number of police per 1,000 residents, single mother households). We also started a side project to estimate the number of street gang members in each county. In this analysis we looked first at ratios between the Hot 15 counties and all the others that reported any homicide at all (about half of the counties in the United States report no homicides in a typical year).

Police Staffing

	Hot 15	All Others	Difference
Population (average)	2,116,839	174,703	12.1
Density pop/sqkm (average)	2,447	156	15.7
Officers/1K Population (average)	4.9	2.0	2.4
Homicide Clearence Rate	54%	68%	21%

The depth of policing in the Hot 15 is low given both population and population density. The Hot 15 counties have more than double the number of officers per capita of their population than the other counties ² but their population is 12 times higher, and their density is 15 times the average of the others. In short, they are under-policed by one half or one third depending on which variable you use. ³

If all other factors (e.g., poverty, education, etc.) were the same (which they are not) then the Hot 15 would be savagely behind in terms of putting cops on the street for their population alone. Since population density basically multiplies the availability of criminal opportunities per district, police staffing should be exponentially higher. Combined, the police shortage in the Hot 15 borders on astonishing.

And that was before the misguided “defund the police” movement.

Clearing Homicides – a Police Staffing Issue?

“Clearances” are the criminology term for having resolution to an investigation of a crime. Typically, this is when a suspect is arrested.

In the Hot 15 counties, 21% fewer murder cases are cleared than in the rest of the counties that have any murders at all. This is significant when paired with the fact that most street gang membership and activity occur in Large Central Metro areas, and all but one of the Hot 15 are such. Uncaught gang bangers continue to bang. Some police agencies have publicly stated that a gang member who is being held on a murder charge is the lead suspect in 2-3 other homicides. Newish thinking in inner-city policing is to hunt the apex predators in gangs, the ones who commit most of the murders.

But this appears to not be the mindset in the Hot 15.

Poverty

Before going too far, we’ll note that the Gun Facts project managed to break the Census Bureau’s data explorer tool. We attempted, in many ways, to export the combination of race and poverty rates for all counties and each attempt caused the Census website to reply “Something went wrong. Sorry, we’re having some issues trying to get the table you’re looking for.”

That said, we do see that the poverty rate for households with people under age 18 are 31% higher in the Hot 15 counties than all the others. Combined with the population data, we see a pattern of more poor people packed tightly into key counties.

Covariates are increased rates of households that received food assistance (20% higher) and single-mother households with offspring under age 18 (30% higher). The latter comes in a bit later, as there is an ugly intersection of fatherless homes, homicides and public policy.

Poverty is not an excuse for crime, especially homicides. But cultures and cultural attitudes about violence develop within environments. We reviewed a sociological survey of three such neighborhoods which noted that the level of trust therein is low to nonexistent (about 26% felt they could trust their neighbors, but only 9% trusted the police). “I’m always on point because you just gotta be.” Other literature notes that in poor, urban neighborhoods “everybody will rip you off.” Poverty can motivate someone into criminal activity for profit; and where it goes unchecked (by police [see above], elders [absent fathers], etc.) it can feed upon itself. Street gangs, a magnet for young males who have a tendency to violent behavior anyway, provide opportunities for “profitable” criminal activity.

Start with poverty, abandon sufficient professional policing, feed new members to gangs… the cycle begins and continues until the mechanics changes the culture.

Where this gets intersectional, is between poverty and…

Race

At the Gun Facts project, we detest talking about race. We firmly believe there is no genetic predisposition to criminality.

But statistically, there is a correlation.

The Hot 15 counties have larger shares of blacks and Hispanics (3.2 and 1.8-fold) over the rest of the counties. Homicide rates for those groups are appreciably higher, though not proportionate. Were the Census Bureau’s data exploration tool more bulletproof, we might be more precise by reporting specific rates of poverty for each race in each county. Should they fix their tool, we’ll update this report.

Key here is that poverty and minorities are a significantly larger presence in the Hot 15 than the average for all other counties. As such, this combination – especially when paired with understaffed police departments – leaves streets exposed to larger groups of economically “desperate” people.

Street gangs

There is an explainable but frustratingly incomplete set of data on where gangs exist, and how many gang members there likely are.

But do know:

• They are primarily in metropolitan areas.
• They are primarily staffed by young males (which are also the people most often shot dead in America).
• They are outspokenly violent, willing to kill over trivial reasons.

But there is no precise county-level street gang membership data. Local law enforcement makes public statements from time to time, but they are guesses. Some agencies report “known” gang members (typically ones who have been arrested and confess their association) but these numbers are a fraction of the total membership. And since criminals rarely answer surveys, there is no way to directly measure.

The closest was the National Youth Gang Survey of 2012 (slightly outdated) which surveyed 2,500 of the nation’s 36,489 law enforcement agencies. Even if we had access to this data (which is locked away and not generally available), the picture would be incomplete.

Urbanization	Total Gang %	Youth Gang %
Large Central Metro	1.0%	5.6%
Large Fringe Metro	0.4%	2.5%
Medium Metro	0.4%	4.3%
Small Metro	1.3%	1.3%

Gun Facts has started a street gang prediction model, but as of this report, the number of data points is too small to accurately guess the number of gang members in any given county. That said, the data shows large and small metro regions have higher gang activity, especially “youth gangs” which is a synonym for “street gangs.”

But a correlation of sorts does exist. Excess homicides (gap) has twice the correlation with their share of gang age people (ages 14 to 24) for the Hot 15 counties than all the others. Stated more simply, when a major metropolitan area has a higher percentage of gang age people, the degree of excess homicides rises. This is not proof of gang activity, but it is a mighty interesting indirect indicator.

All but one of the Hot 15 counties are Large Central Metro areas and for blacks, gang-age youths are present at a 50% higher rate than other counties. Contrarywise, Hispanics youths are about 20% fewer in the Hot 15, which may contribute to their lower homicide rate. Stated differently, black youth are more frequent in the Hot 15 counties, are thus more exposed to gang activity, and thus commit more gang-related or nexus gun homicides that Hispanic youths.

For now, we can’t say anything more definitive about gangs. In previous Gun Facts research (derived indirectly from the locked-up National Youth Gang Survey data), we concluded that street gangs are likely the source of 85% of gun homicides. A strong, but alas incomplete indicator for the Hot 15.

Multiple variables

When we look at the variables that are strongly associated with the Hot 15 we see divergence between the Hot 15 and all the other counties on a few key variables.

MAJOR LEAGURE WARNING:

With only 15 counties in the first half of this analysis, we have too few data points for any statistical analysis to be considered “robust.” Take everything below with a saltshaker.

For the numbers nuts (everyone else can skip ahead): We did a multivariate regression on the key 11 remaining variables against the “gap” rate (share of gun homicides minus share of population) where there was a significant gap between the Hot 15 counties and all the others. We ran separate regressions for the Hot 15 and then for the rest. Divergence between the two is what we explore.

The BIG take-away is that for the selected variables, the correlation (Multiple R²) is a huge 0.97, and even the adjusted R² is 0.56 (a score of 1.0 means perfect correlation). Without the homicide clearance variable, the score drops to 0.87. In short, policing and the ability of local agencies to find, charge and prosecute murders contributes to more murders and thus the larger gap between a county’s share of murders and their share of population.

Disturbingly, the concentration of black youths is highly correlated within the Hot 15 counties, but not for blacks of all ages in the Hot 15, or for either total black population or black youth population elsewhere. The R² regression was 0.64 in the Hot 15 (but due to the limited sample size, the p values were high). Indeed, the only other race that shows correlation in the Hot 15 counties are Asians, and they present a negative correlation.

Past that, households receiving cash assistance had a moderate correlation (0.25 with a moderately stronger p value). Single-mother households were not correlated, though other researchers had concluded just the opposite when measured against all counties.

Oddly, despite other predictors, total population did mildly correlate with the homicide–population gap (0.13), but population density did not.

Policing (actually, the lack thereof) did contribute, but with one really weird gotcha. New York County is in the Hot 15 but has a “badged officer” headcount more than six times that of all the other hot counties. This one datapoint skews the correlation measure of police staffing levels having an effect on gun homicide. Removing this one county and changing its staffing level to the average of all the other Hot 15 counties changes the correlation measure from 0.01 to 0.12. Given that this is a huge outlier, we are going with the later measurement.

An important note is that of all the important variables we tested, none were significant in not-hot counties. Things that made a difference in 15 counties counted for nothing elsewhere.

Sadly, all this leaves out…

Local culture, local politics

Culture tends to go bottom-up. In environments where bad culture is not controlled, it thrives.

There is no means to measure how local attitudes and political priorities affect gun crime. Likewise, given how political law enforcement appointments can be in some places, police chief talking points are often reduced to mere echoes of elected officials.

Here is a comparison.

Chicago police chief after a bloody weekend: “Our heart goes out to the families. We just need to do better. Try a little harder.”

Brevard County sheriff after killing suspect, to ambushed officers: “Evil can never be dead enough.”

Culture derives from communities. In places and times before modern police forces, communities defined how law was enforced (and it was imperfectly practiced, especially in the South). But it reflected cultural attitudes about crime.

In the early twentieth century, communities insisted on a strong police presence – a cop on every corner. Any mayor who did not “keep the streets clean, keep the streets safe” became unemployed.

In the modern era, some localities have notions about crime tolerance that appears shaped by political ideology ( “defund the police”) or the abandonment of government responsibility (“society is to blame”).

Therein may be two impossible-to-measure variables: current political unwillingness to control crime via developing large and professional police departments, and long-term effects by a succession of politicians who are content with crime as long as it is confined to poor, minority race neighborhoods.

	Police/1K Density	Pop Density
Hot 15: Average w/o NY	2.0	1,820
Not hot: Average other LCM	1.4	1,889

Let’s ponder this. In the Hot 15 counties, sans the New York outlier, the average number of cops per their population density (density divided by 1,000 to make easy numbers) is 2.0. For all not-hot counties that are also Large Central Metro (LCM) areas, the rate is 1.4. These counties are thus achieving lower gun homicide rates with fewer police both per capita and per capita/density.

This means something outside of common demographic variables is at play. Chicago, Baltimore, Detroit, and DC fit one perception. Baton Rouge, New Orleans fits another. Dallas is different than Houston despite being in the same state, but they hold #1 and #2 slots in terms of population for Texas.

Nowhere does this present itself more vividly than at the intersection of policing and gangs.

This somewhat complex chart shows that counties with insufficient policing in terms of clearing (arresting) murder suspects and a higher potential gang population have a higher gap (the Hot 15 on the left side). Contrarywise, counties that have sufficient police and clear homicide cases have a lower (often negative) homicide-population gaps despite having many potential youth gang members (and, yes, there are other variables, but none as statistically clear as this).

Data, hygiene, methodologies and other disasters

Despite having some of the most robust crime data collection processes in the world, American crime data is occationally messy.

In preparing this review, we have to suffer our share of outrageous fortunes.

FBI’s Supplementary Homicide Records

The core data comes from the FBI’s Supplementary Homicide Records (SHR), which details many things about murder, from the age, sex and race of perpetrators and victims.

Except for Florida and the Dakotas.

Participation in filing SHR data is completely voluntary, and Florida perpetually opts out. Fortunately, they annually post their version of SHR online and after a couple of hours’ work, we managed to normalize their data with the FBI’s and include Florida in this report. We did not do that for the Dakotas due to them having comparatively small populations (North Dakota’s most populous county has less than 1% the number of people in Los Angeles). Likewise, they lack major metropolitan areas, which years of data and research (as well as this study) show are violent turf.

Dearth of Gang Dope

No, not the narcotic type. Raw data.

click for a larger, sharable version

Officers per 1K Pop ! Homicide Clearance Rate ! Pop Density (per sqkm) ! % Less than HS – % HS Diploma – % AS degree or Some College – % Bachelor’s or More – % People in Poverty – % Ages 0-17 in Poverty ! Median Household Income – Unemployment Rate – % White – % White Youths – % Black ! % Black Youths ! % Hispanic ! % Hispanic Youths ! % Indian/Alaska Native – % Asian – % Pacific Islander – % Two or More – % Households Receiving Cash Assist or Food Stamps !

We won’t belabor this (again). The National Gang Center confirms that most gang members are found in the major metropolitan areas, but that is not granular enough for robust analysis.

We can opine with a casual observation. We are attempting to construct a gang membership rate predictive model by gathering local law enforcement estimates of the number of gangs in their counties. To date, we have 50 such reports – about 1.6% of all counties, spanning four urbanization measures and four regions. Too little data across too many model points.

That said, we see that Large Central Metro counties have gang participation rates 2.8 times larger than either Large Fringe Metro areas or Medium Metro areas. All but one of the Hot 15 counties are of the Large Central caste.

Variables That Matter

Police presence linked to population density and homicide clearance rates are a key measurable.

The other variables stand alone.