The Homicide Rate, Race and Poverty
In my last post, I noted a positive correlation between the homicide rate in a state and killings by the police in the same state. In states where the risk of homicide is higher, police killings also tend to be higher. But there is a mitigating race component, and one which (not surprisingly for those who care about data) goes against conventional wisdom:
for the same state homicide rate, people are less likely to be shot by cops in states where Black people make up 10% or more of the population than in states where people make less than 10% of the population.
Looking at homicides, and accounting for race, it seems there are different dynamics at play among different population groups:
(Click to embiggen.)
I can’t find murder rates (whether victimization or offender) by race at the state level, but I do note that the data appears to show a clear relationship between the overall murder rate and the ethnic makeup of a given state:
Relative to the rest of the population, there is an elevated homicide rate (both offense and victimization) in our Black population. Thus, if we want to reduce the homicide rate, perhaps the opportunity is greatest in understanding why the homicide rate is as high as it is in the Black community.
Poverty is often mentioned as a factor driving crimes in general, and sometimes homicides in particular. To examine whether that is the case here, the next graph shows the percentage of a state’s population that is Black on one axis, and the homicide (victimization) rate on the other axis. States with fewer than 3 million than people are omitted. Additionally, states with a Black poverty rate in excess of 22% (which is approximately the median poverty rate for the Black population, measured by state) are colored orange:
While the homicide rate is lower in states with less poverty (median of 42 murders per million v. 58), there is no clear pattern that would indicate that poverty rates the Black community are a primary driver of the murder rates according to the above graph. If that isn’t clear to you, the graph below includes the same information but from a different perspective.
So the data seems to indicate that reducing poverty in the Black community, while a laudable goal for any number of reasons, is probably not going to have a strong influence on the homicide rate.
This post is getting long so it’s time to wrap it up. I would note that it goes without saying that it would be tremendously useful to find ways to reduce the homicide rate and reduce the number of victims, particularly in the Black community.
One final note… if anyone knows where I can get data on the homicide rate by different population groups by state please let me know.
—
Update, 7/31/2017, 5:21 PM PST
As should be obvious from the wording of the post, I was concerned that someone might misconstrue the second to the last graph. To make the homicide relationship more obvious, I included the last graph as a companion piece. However, a habitually rude offender claims to believe that the states with a relatively low level of Black poverty (i.e., the blue points on the second to the last graph) have a different relationship between Black poverty and homicide rates than the states with relatively high levels of Black poverty, and all of that is being somehow masked by the existence of Maryland. So… below please find the same graph, redone, including only the states with low levels of Black poverty, but leaving out Maryland. (Note that the slope of the graph below is actually steeper than the slope of the graph with a broader set of data though the fit isn’t as high.)
Two things to look at in the statistics – in the first table, how is it that under 2% of homicide victims are of an unknown race but over 30% offenders are ? That makes it seem like the numbers have been manipulated.
Also, on the poverty charts – which poverty ? Costs of living vary by over 100% within states depending on County & distance to a high cost area. The same income is poverty in Howard County Maryland vs. living great 100 miles out.
Pete:
Welcome to AB. 1st comments always go to moderation.
Pete,
Murderers have an incentive to prevent the authorities from connecting them to their crime.
Also… as to the definition of poverty, the census does not adjust for differences in state cost of living, except insofar as (I believe) the figures reported by state are for non-rural areas.
But… that would imply there are more truly poor people in NY, MA and CT which are all states with lower homicide rates than the trend line in the second to last graph. That if anything strengthens the argument that poverty, even a real measure adjusted for cost of living, is not driving the homicide numbers.
Mr Kimel
You use deception again.
Your chart (with “orange: Black poverty rate greater then 22%) shows a regression line independent of poverty rates… e..g homicide Rate as a function of Black Proportions.
In that Chart the solo point with Black poverty rates less than 22% with a high proportion of Blacks in the State is Maryland (MD on your chart).
Mayrland’s homicide rate is nearly all and only in Baltimore which has Black Population = 63% of Baltimore’s population. To top it off, poverty rate in Baltimore is over twice the rate of Maryland’s. Then the gross difference — the Maryland Homicide rate without Baltimore is just 2.6 / 100k while Baltimore’s Homicide Rate is 52.5 / 100k.
So the Mayland “blue” dot in that chart is a huge outlier among the blue dots… for the reasons cited above. Omitting that blue dot and regressing the blue dot’s separately from the orange dots gives a huge difference: Approximately, (eyeballing only) the already shown red regression line would remain the same if it was applied only tot the high poverty rate data (orange dots). But the blue dot regression line would be nearly horizontal. In other words homicide rates in states with high poverty is positively correlated with high black proportions, while there’s no correlation (or very low correlation) of homicide rates to black proportions in lower poverty States.
You noticed immediately in that chart what any chartsmanship author would notice… the Md data point is a huge outlier and the reason would be generally known by anybody who understands Baltimore’s relation to Maryland in both Black proportions and crime rates — this was true in 1969 as well as it is today.
So you have intentionally misled the readers. Poverty plays a huge role in homicide rates as every academic study has shown forever.
Mr Kimel,
Here’s the comparison of Maryland to Baltimore
Maryland
Population (2016) 5.774 million
– % African American 30.7% (2016), 29.4% (2010)
-Poverty 9.7%
– Homicide Rate 8.6 / 100k (2015)
…- Murders 493 (2015)
….. – Murders Excluding Baltimore 149
….- Homicide Rate Excluding Baltimore 2.6 / 100k
Baltimore
Population (2016) 0.614 million
– % African American Only 63% (2010)
– Poverty 23.7%
– Homicide Rate 52.5 / 100k (2015)
..- Homicides 344
– Black Homicide Rate 54.5 / 100k pop (2015)
…- Black Homicides 339
– Black Homicide Rate / 100 k Black pop 86.5 (2015)
…..(based on blacks at 63% of pop).
Longtooth,
OK. So I’ve added an update. The new bit of deception is a graph essentially the same as the one to which you object but omitting the orange states and omitting Maryland. And guess what it shows? Not what you claim at all. The slope of the line is pretty similar to the slope with all the blue and orange dots.
That said, you have a point. If we get rid of not just Maryland, but also IL and TN and OK, then, as you put it, “the blue dot regression line would be nearly horizontal.” In other words, if we labeled a third of the blue dots outliers for no reason at all except that we think it would change the relationship to something we want to see, we could, in fact, get the results we want to see. I am embarrassed you saw through my deception so easily.
No it’s not the same chart with MD excluded.
Do the same chart, exclude MD, then do regression on the Bllue and Orange States. You purposefully mislead and misled againg.
To be explicit, do a regression on the blue states excluding MD, (or include MD but with excluding Baltimore’s rates) AND a regression on the orange States The Orange States will have approx. the same slope as the slope shown on the original chart, but he blue state regression line will be essentially flat.
I honestly cannot believe how much of this oafchuckery I am tolerating. Perhaps I am getting sick or something.
Longtooth,
Step 1. Press Ctrl-F
Step 2: Copy this into the search box:
Update, 7/31/2017, 5:21 PM PST
Step 3. Hit enter
Step 4. Read the paragraph that follows those words. Look at the graph. See the equation for the trend line pasted into the graph.
Step 5. Realize that I’ve dealt with this first in the original post, and again in the update from a few days ago.
For extra credit, explain why you think my refusal to chuck a specific set of points equal to one third of the blue sample for no apparent reason amounts to deception.
Mr. Kimel,
I’ve pointed out the deceptive charts and suggested remedy by either omitting MD or omitting Baltimore’s homicide rate from MD in general … a factor 56/2.6 = 21x (per 100k population) and running two separate regressions on two separate populations. You’ve chosen not to do so.
I’ll leave others to see and realize the deceptive chart and conclusion.
And just btw, if you compare Illinois with Chicago you’ll find a similarly huge differential as is the case with MD and Baltimore… If you want to use states as comparatives, then you are ignoring the major metro area’s relationships as being important and significant aspects of black poverty and crime, homicide rates not less so.
From Mr. Kimel’s 2nd Chart:
States Counts:
Segment the vertical axis at just above TN.
Omit MD as before for reasons previously cited
Of States with Poverty Rates > 22% (Orange dots):
– 5 of 16 (~ 31%) have Homicide rates >75/million population
– 4 of 16 (25%) have black populations >25%
Of States with Poverty Rates 75/million population
-0/12 (0%) have black populatioins > ~ 18%
Homicide rates > 75 / milllion population are ~4x greater (32/8) in States with Poverty rates > 22%
Black Populations in States with Poverty Rates > 22% are at least 4x more likely to have black populations > 18%.
—————————————————————————————————
Segment the Horizontal Axis just to the right of NY (18% Black Population)
Omit MD as before for reasons previously cited
Of States with Poverty Rates > 22% (Orange Dots)
– 6 of 16 (~38%) have black populations > 18%
– 4 of 16 (25%) have black populations > ~27%
– 4 of 16 (25%) have Homicide rates > 70 / million population
Of States with Poverty Rates 18%
– 1/12 (~ 8%) have Homicide rates > 70/million population
States with Poverty Rates > 22% have at least 38x the likelyhood of having Black Populations > 18%
States with Poverty Rates > 22% have at least 25x the likelyhood of having Black Populations > 27%
States with Poverty Rates > 22% have 4x the likelyhood of having Homicide Rates > 70/million population.
————————————————————————————————-
Of States with > 22% Poverty Rates AND >18% Black Populations AND > 70 Homicides / milllion population, ALL (4) are former Confederate secessionist Slave States
Of States with > 22% Poverty Rates AND > 18% Black Population, ALL (6) are former Confederate secessionist Slave States
Of States with Homicide rates > 70 / million population 89% (8/9) are former Confederate secessionist Slave States (MO is included as secessionist because it’s gov’t voted to succeed and went into exile)
Longtooth,
Upthread you demanded I redo the graph for blue dots only but omit Maryland. That was done in the update. I posted that graph in response to your comments several days ago. Do you see Maryland on that graph? So I did as you insisted. On 7/31.
Now, what happened to the slope of the blue dots? You said, in your comment, that with MD removed the blue dots would have a flat slope. Quite the opposite happened.
Now you keep insisting that I remove MD and that I am being deceptive by not removing MD. There is not a damn thing I can do. The graph that was updated several days already has MD removed. Every point is labeled. Every single one. So you can compare the updated graph with the earlier version and see that all the blue dots are in the same place except MD which is no longer there. (Yes, the scale changed.) If that isn’t enough, I noted above the graph, and in the graph itself, that MD was removed.
To give you the benefit of the doubt, I just checked the post again, both on my laptop and my phone. The updated graph with just blue dots, leaving out MD, is there in both. This isn’t a case of somehow the update not being visible to you.
So you can leave it to the readers to decide if I am being deceptive, and I will leave it to the readers to decide what is going on with you. This goes well beyond your insistence on defending bacha bazi and FGM.
Mr. Kimel,
then where’s the regression for just the Orange dots Your chart # 2 is a single regression for the combined organ and blue dots? You continue to deceive.
And I don’t insist you do anything… I suggest what you may do for a more accurate analysis of the data you use and point out why what you do is an inaccurate analysis and thus also misleading..
Mr. Kimel,
It occurs to me that you may also be deficient in your Excel charting skills. I say this only as a conjecture only because you can’t seem to show two separate regressions, with regressions stats on the same chart, which is the simplest and far faster way to do it.
Mr. Kimel,
Sorry to bother you again, but the only reason I take issue with your deceptions is that all academic objective sociological studies of poverty, v race v crime rates show that rural (low density) populations show no difference in black v white crime rates, but at the same time all studies show poverty is the most significant variable in metro area’s and that adjusting for poverty by race shows no significant difference in metro (high density) environments. These studies aren’t new. They have been widely published and available for decades.
I have previously suggested you spend more time reading and understanding the objective academic sociological studies on the subject rather than trying to roll your own using dubious and selective data sets.
The fact that you don’t post about the massive sociological studies and their conclusions only gives me more reason to conclude you are intentionally misleading both in the data you use and the conclusions you draw from it.
The only reason I post my comments is to point out the deficiencies in what you post. I am a strong believer in public forums but at the same time it then becomes incumbent on participants in those forums to point out falsehoods (lies), misleading and/or selective use of information supplied, if they are aware of them.
If you chose to persist in ignoring these things that’s your prerogative, but it doesn’t make your analysis any more accurate or any less misleading..
Longtooth,
This one has the broader set:
http://angrybearblog.strategydemo.com/wp-content/uploads/2017/07/Figure-3-header-changed.jpg
This is the smaller set (just blue dots less MD):
http://angrybearblog.strategydemo.com/wp-content/uploads/2017/07/Figure-5-for-update.jpg
The regression line for the two (and they are in the two graphs) are, respectively:
Y = 200.8x + 27.365
Y = 210.9x + 27.103
I don’t feel much of an urge to do the orange dots by themselves since anyone who doesn’t feel that the second line is flat, as you insisted it was going to be, can estimate from the two regression lines pretty easily approximately what the slope and intercepts would be for that data set.
I appreciate your concern for my Excel skills. Rest assured I can stick multiple regressions into a graph but it didn’t strike me as useful in this exercise. I still don’t. I am just curious how far you are planning on pushing your questions though to be honest, I am getting less curios fast.
As to why I don’t spend my time doing an exhaustive search of the sociology literature, I follow George Bernard Shaw’s prerogative: one doesn’t have to eat a whole egg to know its rotten.
My experience is that whenever sociology papers touch on race at all, the field’s usefulness approaches zero. Sadly, I agree with many of the definitions you will find here.
Now, I wouldn’t be surprised if you were a sociologist at one point. You certainly fit the bill. You have spent many pixels defending the rape of little boys and the mutilation of little girls because as equally valid to any other culture. You also refuse to accept what the data shows, to the extent that you will continue to harp that I am being deceptive for not putting up a graph with MD left off of it literally days after I already produced that graph at your request.
1) It helps to remember that, historically, killing a black person was not considered a criminal matter. Whether a black or white was the killer, the police wouldn’t consider it something they had to deal with until relatively recently. Even in NYC, whole districts, like Harlem in the 1930s, were declared free fire zones, as long as white outsiders weren’t harmed.
2) The American south has long had a relaxed attitude towards murder. In the north, it was a big deal, but in the south, it was just part of being a man, whether one was the victim or perpetrator. A lot of blacks absorbed this southern attitude and took it north in the latter half of the 20th century. The whole black thug culture is just the white thug culture relocated.