About . . . . Classes . . . . Books . . . . Vita . . . . Blog. . . . Podcast

by Peter Moskos

January 20, 2019

To call 911 or not to call 911?

May 17 of last year the NYPD issued an obscure order concerning "aided reports" -- that's when a cop responds to a 911 call for a sick person waiting for an ambulance (a "bus," as they say here) -- requiring the officer to enter the person's information into their phone. This looks all technical and boring.

When you put the "aided card" into the phone, it automatically goes and queries the warrant system. This means that if Uncle Pedro has a heart attack and he or you is wanted, cops will take you away (after medical treatment, but still). Thanks, technology! That'll teach you to call 911!

Say your Uncle Pedro has chest pains. Or is ODing. Should you call 911? Of course! Right? But what if you don't know if he's wanted? What if you don't know if you're wanted? Should you still call 911? You know cops might also respond because, well, why not? Maybe cops can do some good before the ambulance arrives. (Though generally, as a former cop, when it comes to medical care, are you serious?) Or keep the peace. But should you be debating all this before deciding to call 911? While you're discussing the pros and cons, Uncle Pedro just stopped breathing. 

The NYPD has spent time and dollars trying to build relations with all communities. We want people to call for help. The goal has always been to bring people into the system, not make them afraid of it. The Neighborhood Coordination Officer (NCO) philosophy is just the latest serious effort. All this will be for naught if people are afraid to call 911 or 311 even for non-police matters.

We don't need people thinking EMS are the bad guys. And we for sure don't need people fighting unarmed EMTs because they're worried that they the EMTs and paramedics are going to call the police and get them arrested. That's not good public policy.

Cops do not have discretion when somebody comes back wanted. A warrant is a warrant. And arguably for good reason. A judge hath spoken. But there are wanted people out there, and an entire undocumented population, for instance, whom we still want to call 911 when A) there's a fire, B) they witness a crime C) they victims of a crime, and D) when they need medical care. Needless to say, this is not an inclusive list.

Perhaps for minor violations, when you know the person’s name and addresses, just give the guy something like a "must appear notice," like the one he never got because it was mailed to his address from three years ago. Then bounce it to a detective for follow up. A surprisingly large percentage of people who have warrants simply do not know they are wanted. Give them 60 days or something. Why is the only part of the criminal justice system that moves quickly the one in which somebody wanted is taken in?

It's in everybody's best interests to have people turn themselves in at a more convenient time. This can be the difference between staying employed or being fired. Most warrants are not over urgent matters. And often staying employed can make all the difference in the world.

Could it become common in NYC hospitals (and not the hospitals serving rich white people) for police to run the names of visitors and patients while they are waiting around? For some, their injury or presence might constitute grounds for a probation or parole violation. This is exactly what Alice Goffman said was happening in Philadelphia. (It's not clear it actually was happening, but people thought it was, and that's bad enough.)

New York City has an estimated undocumented population (aka illegal immigrant) population of 560,000. Even in a sanctuary city, people -- more than half a million New Yorkers -- are afraid. Currently NYPD doesn't share this information with ICE. But that could change overnight. Recently I had an immigrant student whose boyfriend was hit by a car. He was hurt. The driver stopped, but the boyfriend didn't want to exchange information. A guy hit by a car through no fault of his own was afraid to get the driver's information or go to the hospital. This is not good.

What problem is this solution supposed to fix? "We want the cops to put an aided card into the phone on the scene and it to automatically query the warrant system." It is bad policy to routinely run warrant checks on people seeking medical care.

I know it's not in the public's interest to have wanted people running around. It's one thing for police to run somebody because they have suspicion. It's another to do so because they called for help. It's not in the public's interest to have people afraid to seek medical care or see EMTs and paramedics and the FDNY as part of law enforcement. Let's base a policy decision based on evidence rather than, "hey, cops now have smart phones linked to the warrant system!"

One interesting (at least to me) thing I learned in talking to somebody about this, cops in New York did not routinely run (check for warrants) every time they 250d (stopped) somebody. In Baltimore, we ran basically everybody we stopped. This is a big difference in police behavior, and I've never heard anybody discuss or even be aware of this. But even in Baltimore we didn't routinely run people on medical calls. In part we didn't want to know. Because if the person is wanted and going to the hospital, guess who gets to babysit the patient until they're released? Not a good use of patrol resources. And the next shift will really hate you, too.

Maybe we ran more people in Baltimore than they did in NYC because more people were wanted. But it probably had more to do with an unrelated technological issue. One radio channel in Baltimore covers one district (aka precinct) with 1 dispatcher for 15 (often fewer) patrol units. One radio channel in New York covers multiple precincts and has perhaps 10(?) times as many officers. It takes precious air time to run a 10-29 (Balto code for, check warrants). And air time in New York is more precious. Many stops, even car stops, weren't called in. That's not safe or good policy. Something as simple as how many units are on one radio channel, can change police culture more than any formal debate or informed policy. Maybe it shouldn't be that way.

And if it is good policy to check people in medical crisis for warrants, and I don't think it is, they hey, why not go all-out and front-end it to the 911 and 311 operators. Let them be part of the system, too. At least it would be honest. "Thank for calling 911. This call is being recorded and you are being checked for any felony warrants. Now, what is your emergency?"

January 6, 2019

What's Up With Crime Being Down in Camden?

Let me start by saying I don't know much about Camden, New Jersey. So if you know more, help me figure things out.

The city of Camden is just across the river from Philadelphia. It's part of Camden County. The city has a declining population of about 75,000. Camden is about half black and half hispanic. It is, by any quantifiable measure, a "struggling" place. I wrote a post about violence in Camden back in 2015.

In 2011 the city and police department were in crisis and announced plans to abolish the police department and start fresh, with a new police department. In May 2013, the city police department was abolished (in part to break the police union, which has since re-formed). Anybody that wanted to stay on had to re-apply for the job. Since then, the new Camden County Police Department covers Camden City (and only Camden City) while in Camden County, I guess some other agencies (I presume local agencies and/or the sheriff) do the work.

This all makes data gathering a bit confusing. But I (painstakingly) went through the UCR's arrest numbers for Camden City from 2009 to 2016 (the last available year).

Nice chart, if I do say so myself.

If history is a guide, and they say it is, when people blame an institution for human problems and tear it down and start new, after a few years you end up with pretty much the same situation and problems. Police are as much a product of their environment as anybody else. There are still occasional problems of corruption and brutality in Camden. Cops still get attacked. And Camden is still mired in poverty (a 37% poverty rate). But poverty is declining and money is being invested."

Meanwhile, across the river in Philadelphia, murders are up 25 percent (2016-2018). I do presume "underlying social conditions" haven't gone that drastically in opposite directions in these neighboring cities just since 2016. So what if -- crazy idea -- police (and prosecution) actually matter. Maybe a lot. And even more than the so-called "root causes."

I mention this because the new Camden County Police, policing Camden City, have become the progressive reformers' dream team (despite being founded, in part, in a fit of Republican union busting). Since 2013 there have been a lot of positive press, but here is one example that presses all the feel-good buttons like "strategic shift toward community policing" and "rebuilding trust between the community and their officers" and "being mentors in the community" and "a showroom for community policing techniques" and "nothing stops a bullet like a job." OK, but all that sets off my BS alarm.

In terms of crime, the proof is in the pudding. Give credit where credit is due. And here's the thing: violence really is down. A lot!

Last year there 22 murders and the year before 23, down from 67(!) in 2012. Shootings also been cut in half. Maybe police culture really did change for the better. Or training. Or technology. Or strategies. Or maybe police are now simply funded at the proper level they had not been. Or maybe we're getting more for less. I don't know.

But I do know, despite what is often reported, it hasn't been just kumbaya with carnivals and free ice cream. Those gimmicks can be part of building trust, but they're not crime prevention strategies. Non-criminals need more positive casual interactions with police. Criminals need more interactions, too Perhaps not all so positive, but still professional and respectful. (The person you arrest today can be your source or even save your ass tomorrow.) As Chief Thomson says), "Nothing builds trust like human contact."

And speaking of human contact, reported use-of-force -- usually something reformers want to reduce -- increased dramatically with the new police department. That could mean cops are now more brutal, but more likely cops are policing more, and some of that leads to justified use of force. Camden is being lauded by reformers for bringing down crime with exactly the form of pro-active policing loathed by the same reformers!


Force went up. Arrests went up. Crime went down. But what about the idea, very popular among reformers who don't live in high-crime neighborhoods, that arrests are bad, and people in dangerous neighborhoods hate police because police are arresting (or shooting) people of color for no good reason. 

If you decriminalize minor offenses, goes the hope, police "legitimacy" will increase, which along with leading to less incarceration means more solved crimes and many other wonderful things. It sounds good, especially if you think police are the problem and your neighbors aren't.

Based on UCR arrest numbers, arrests went up with the new police department. Camden cops are arresting more people, and crime is down. There may have also been better policing, but there was also just more policing. And the kind of arrests that increased -- low-level discretionary arrests -- would indicate that police focused on quality-of-life issues and Broken Windows. This is not the reformers' party line. 

Caveat: I really hate using arrests as a metric for anything, much less good policing. But arrest data is available. And huge changes in arrest numbers tell you something is going on. Arrests can be a proxy for pro-active policing: cops stopping suspicious people, chasing and catching the bad guys, cops less afraid of making an honest mistake. My inquiry into Camden was inspired by this article from March 2017 (that I just read) saying drug arrests were way down. Except that seems not to be true.

In 2011, Camden cops may have lost a little of whatever go-getter spirit they still had. This was Camden's Ferguson Effect (pre-Ferguson). Cops were told they were no good and their job was on the line. Arrested dropped 50 percent, from 11,000 in 2009 to 5,348 in 2011. Along with Camden, Baltimore and Chicago also saw similarly quick and drastic decreases in quantifiable policing. And in all three violence shot way up. Yes, correlation that is also causation [thunder clap]. At the start of 2012 Camden laid off 45 percent of the police force. Murders went up to 67 (which is a shocking number for a city of 75,000).

Let's compare 2012 and 2014 Camden, when murders went down from 67 to 33:
  • Drug arrests up 79% to 3,052.
  • Marijuana possession arrests in particular up 467% to 488.
  • Curfew and loitering violations up 34% to 1,128.
  • "All other offenses (not traffic)" up 50% to 3,352 (This most minor category is probably something catch-all like disorderly conduct, trespassing, loitering).
  • DUI arrests up 483% to 175 (an indicator of more policing).
  • Non-felony (ie: discretionary) assaults up 57% (to 754).
  • And murder arrests -- because there were fewer or them -- down 23% (to 20).
Policing get "better," but what does that mean? Maybe police officers have better manners. That matters. But what brings down crime is focusing on repeated violent offenders, usually young men, who commit the vast majority of violence crime.

There's irony here in that this little department so loved by progressives, has achieved success, in part, by arresting more minorities. And you know what kind of arrests increased the most? All the little ones that reformers want to stop in the name of social justice But those progressive reformers don't live in Camden. If you do live in Camden, you probably support anything that works.

For a small city, 9,000 arrests is a very large number. Scaled up to the population of New York City, for instance, this would be over one  million arrests a year (compared with the 240,000 arrests in NYC last year). One arrest for every 8 people is similar to the arrest rate that Baltimore had in the early 2000s (when I was there and violence was going down). This is the same arrest rate people (stupid people, mind you) blamed for Baltimore's riots a decade later (arrests and crime in Baltimore dropped drastically from 2003 up until the riot of 2015).

Now keep in mind arrests are not good on their own. It's very important what the data do not reveal. How many times did cops change behavior without resorting to arrest? My guess is a lot. More good policing does often lead to more arrests, but it's really important to put the horse before the cart. Policing is the goal. Not arrests. "More arrests" is never a good strategy.

I'd like to know how many people were arrested in Camden in 2017 and 2018 when murder really dropped. In the ideal world, violence and arrests (and incarceration) all go down in sync. That's the win-win(-win). But residents will always choose more arrests and less violence rather than the standard police reform package of less policing and more violence.

One moral, and you see it time and again, is you don't have to fix society's problems to fix violence. Violence is not inevitable. But equally important is the corollary that you can't fix society when violence is out of control. Most residents want more police. They want visible police who maintain order and treat people with respect. It's not too much to ask for.

Maybe what is going on in Camden is just slapping lipstick on a pig. But hey, it's hard to argue with success. Don't underestimate good PR and a progressive-sounding chief who both controls the narrative and won't give in to anti-policing naysayers. And it's likely that what the arrest numbers do not show -- better hiring, training, culture, attitude, accountability, and leadership -- is what makes effective aggressive policing possible, or at least palatable.

Camden homicide numbers
2018: 22
2017: 23
2016: 44
2015: 32
2014: 33
2013: 57
2012: 67
2011: 52
2010: 39
2009: 35
2008: 53
2007: 45
2006: 33
2005: 35
2004: 49
2003: 41

Non-fatal shootings
2017: 95
2016: 92
2015: 109
2014: 90
2013: 143
2012: 172

2009: 11,280
2010: 9,414
2011: 5,348
2012: 6,903
2013: 6,613
2014: 10,582
2015: 12,049
2016: 9,052

Notes: In the data (for 2016, ISPSR 37056 and 37057) variable "offense" the value 18 is the total for drugs. Subtotals follow. Values 180 and 185 are again subtotals of what follow. This makes drug arrest numbers very easy to triple count, as I did at first.

For 2009-2012, I'm assuming Camden City is the agency with the listed population of 77,665. I ignore the other Camden, which presumably is the rest of the county. After 2013 (unless I'm wrong) Camden City is the agency "Camden County Police Dpt" with a listed population of about 75,000.

As always comments and corrections are welcome. Replications welcome; data available on request.

Some sources: https://www.nj.com/camden/index.ssf/2018/01/camdens_2017_murder_rate_was_the_lowest_in_decades.html

January 3, 2019

More on state differences in cops shooting people

Inspired by some twitter threads -- mostly this one with Gary Cordner and this one with Andrew Wheeler -- I thought I'd look more at the cops getting killed as a factor in cops killing people.

I like presenting this stage of research. In part because coming up with ideas and hypotheses and basic number crunching is what I like doing most. (I'll leave the journal article submitted and advanced stats to others.) I'll explain my steps partly to help others, but also to help me go through this on the old assumption that if you can't explain it to others clearly, you don't really understand it yourself. (I used Excel and PSPP.)

I'm always partial to fewer better-data over more bad-data. So, as I often do, I'd like to stick with good old murder: officers shot and killed on duty (from the officer down memorial page, which over the years I've found close to faultless, which is more than one can say for the UCR or anything else.)

The problem (from a statistical not a moral sense) is that there are many states in which very few officers are killed. So I went back to gather 20 years of data (for no particular reason, just a choice, it could have been 10 or 30) and got the number of officers killed between 1999 and 2018, by gunfire, for each state. 50 states. 990 total deaths. I dropped the states where n < 10. That leaves 33 states. Texas and California top the list, which isn't surprising because they're big states. But then come Georgia, Florida, and Louisiana. Interesting...

But what's the best denominator? I mean obviously one needs to look at population to get a rate. But which population? In order, I'm going to consider 1) number of cops, 2) population levels, and 3) violent crime levels, 4) population density, and 5) percent of population that is African American.

1) Perhaps we should look at cops killed in terms of how many cops there are in any given state, so as to consider the chance of any given cop being killed on duty. Makes sense to me, the problem is that the official data even on how many cops there are looks dodgy. It seems unlikely to me, for instance, that Mississippi went from 5,222 cops in 2007 to 2,524 in 2014 (the two years anybody attempted to count, but reporting is voluntary). If I don't trust the data, I don't want to use it. But I still did the numbers, based on the average between 2007 sworn officers and 2014 sworn officers.

For presentation purposes, let's use the USA average (using all 50 states) as a baseline, set that to 0, and compare all the states:

Cops are more likely to be killed in MS, LA, AR NM, SC, GA, and AZ. Keep in mind the small and safe states have been removed from the calculation. I don't like this. If nothing else because I don't trust the Mississippi numbers.

2) So let's just use overall population as the denominator. I'm using 2016 population because that's what I already have in my file. Some states have grown a lot in the past 20 years. Oh, well. I don't think it matters that much for these purposes. If it does, we can consider it later. Keep in mind these are ratios, the actual numbers by themselves are meaningless. But as a ratio, yes, a value of 1 means a cop is twice as likely to be killed per capita. It does appear that a cop in Louisiana is about 4 times as likely to be shot and killed as a cop in New Jersey.

This says that Louisiana, by far, is the most dangerous state to police in. Arizona is next. And given that its population has grown drastically in the past 20 years, it should really be higher. And that would make LA seem like less of an outlier.

New Jersey, Massachusetts, and New York are all comparably safe. I won't say the safest because the 17 safest (and smallest) states have all been dropped for the statistical reason of having fewer than 10 cops murdered over the past 20 years.

I think number 2 (population) is better than number 1 (number of cops). But they're not drastically different. You get the same states on top and the same states on bottom. But I'm going with state population as the denominator because I don't trust the count of cops.

3) Now let's consider violent crime as an independent variable (which is the variable that affects something else, on which something else is dependent). And back to using all 50 states.

I just got some crude numbers off wikipedia and then took an average of 5 years of data for each state. (Not the best methods, but probably accurate. Certainly fine for preliminary work.)

Let's run some correlations. I like correlations because they're easy to understand. They also tell you where you should look for deeper answers.

First question: at the state level, is violent crime rate correlated with cops getting killed? Absolutely (Pearson Correlation = .62, Sig = .000). This is a strong and unsurprising relationship.

Next, at the state level, is violent crime correlated with being killed by cops? Surprisingly, technically, statistically, no. (correlation = .23 sig = .104) Not at the state level; not with an N of 50. Now I know from other research that violent crime is correlated with being killed by cops, but you've have to delve down into the neighborhood level to see that effect. But still, if it that doesn't come out at the state level, it's a clue that something else is also at work! This is where things get interesting. Something else is also at play on a state level that is more significant than straight-up levels of violent crime.

4) What about geographic area? This is where wikipedia is great because you can get state size in seconds. And then if you already have population and you're handy with cut-and-paste and sorting on spreadsheets, you can get population density in minutes.

And it turns out the population density is indeed correlated with a lot.

Lack of density -- more space -- is correlated with being more likely to be killed by cops. Think of what this means. Common sense tells you it's not a view of "big sky country" that makes cops shoot someone. Whatever really matters, is correlated to density (or lack thereof). Maybe it's single person patrol. Or the time for backup to arrive. Or meth labs. Or gun culture. This is why they say "correlation doesn't equal causation" (which is also the most frustrating phrases in social science, because correlation can very much indicate causality, and the phrase is often used to dismiss meaningful correlations as meaningless.)

Population density (lack thereof) is also correlated with cops being killed. Density is not at all correlated with crime (like not even leaning in one direction). And yet both crime and density are heavily correlated with a lot of other factors. And both are correlated with cops being killed. More crime = more cops killed; more density = fewer cops being killed.

So now lets do a brief multivariate analysis, which is about as far as I go. This means that we look at more than variable at the same time. Which is more important (plays a greater role) in cops being shot and cops shooting people? Crime or density? (Or something else.)

Density seems to be more predictive than crime in terms of cops killing people and less important in terms of cops being killed (though for the latter both are correlated).

When I move "cops killed" to the independent variable side and keep a focus on people killed by cops, density becomes less important and violent crime becomes more important. This makes intuitive sense. Because the issue with a spread out area is that cop, alone, would face greater threats.

Keep in mind the above is about cops being killed. Much more talked about (by non cops) is people killed by cops. I wrote about that a few days ago.

If you're still with me, kudos. Causes here's where the whammo happens!

Were one to only look at individual variables, the key would seem to be density followed by crime and rate at which cops are killed. But it turns out that much of what is measured in those variables are simply correlated with and less important than the percentage of black population in a state. Crime matters. Police being killed matters (independently of violent crime), population density may matter a little, and of course other variables that I'm not even looking probably matter a lot. The question is always if they can be identified and accurately quantified.

Last year I observed that cops shoot more often in states that have fewer blacks. So I already had a strong hunch to look in this direction.

When one puts the state's percentage of African-American residents into the equation, things start to fall into place. This is also taking into account how often cops get shot, crime, and density (which finally starts to lessen in importance -- because as we know is only indicative of other factors -- but still probably important in terms of gun laws and culture and police-backup).

If one considers crime, density, and black percentage -- but only when one does so all together -- all three are significant (with an R-squared of .55). When one adds the rate at which cops are killed, r-squared goes up to .62.

[R-squared is technically the distance (squared to take account of negative numbers) that data points are from the trend line of a chart. At some level, r-squared is supposed to indicate how much of what is being looked at is explained by the independent variables in the statistical regression. But that's more in a statistical sense than a real-world sense. Still, generally, other things being equal, a high r-squared is better than a low r-squared. And an r-squared of 0.63 ain't shabby for this kind of game.]

So what does all this mean? Density matters, but not so much for what it is but for things correlated with it (same could be said for race). All these variables have "intervening variables," the way people act, the choices they make, the factors that make us do what we do. Things that may be harder to measure than crude indicators like "population density" and "race."

Still, looking a these variables, density seems mostly to correlate with the lack of African-American in a state. The black percentage of a state seems to be the most significant factor in determining how many people are shot and killed by police (with overall violence and cops being killed also being important). But, contrary to what many people believe -- and basically all of the "narrative" of the past few years -- the relationship is inverse. The greater the percentage of blacks in a state, the less likely cops are to shoot and kill people.

This is counter-intuitive to a lot of people, particularly if you think cops only shoot black people. But it makes perfect sense if one thinks about it in two parts:

1) Whites don't really care about who police shoot; period; end of story. And without the pressure over bad (or even good) police-involved shootings, cops never learn how to shoot less. Other things being equal, cops simply shoot more people if there isn't any push-back from (to over-generalize) blacks and liberals and media and anti-police protesters. Call it the Al Sharpton Effect, if you will. Basically, in many places, police organization and culture do need to be pressured into changing for the better.

2) Police can be recruited, trained, and taught to less often use legally justifiable but not-needed lethal force less. The state variations in police use of lethal force are huge. Some states (and particularly jurisdictions within states) do it better than others. Instead of saying "police are the problem" we could look at the states and cities and department that are doing it better and learn.

Ultimately what we need are well and better trained police officers who shoot less often, but still shoot when needed.

I'll leave you one final bit of data. I don't know if there's a there here or not. My guess is this does matter. But maybe it's just a clue that leads to the above. Or maybe it's something else. Maybe you can figure it out.

This is a table that shows a simple ratio: the number of citizens killed for each cop killed. Good people can debate what this ratio should be. I don't want to go there. The correct ratio is no cops getting killed and few criminals getting killed. But what's interesting to me is the that there is such a large difference between the states, and by a factor of 10! By and large the states on the high-end (more citizens getting killed) are very white and the states on the low-end (fewer citizens getting killed) are disproportionately black.

Take Oklahoma. Cops in Oklahoma are not getting killed a lot, per capita or per number (0.6 per year over the past 20 years). There's not a lot of violent crime, and yet in the past 4 years cops in Oklahoma have killed 118 people. Again, I don't want to get into what the correct ratio is, but seeing how the national average is 20 civilians-to-cops shot and killed, and seeing how some states are down under 10, why the hell is Oklahoma pushing 50?

Louisiana cops are getting shot at and killed three-times more often than cops in Oklahoma (and 8 times more often than cops in New Jersey). Both Oklahoma and Louisiana cops shoot a lot of people. But in Louisiana, dare I say, they have good reason to.

Pushing the Ideological Narrative

I updated the Brennan Center's crime report from 2016, to update it for 2018. I still have this urge to show how goofy their methods are. Why? Because, the authors are still cited by reputable journalists as experts, despite never acknowledging or correcting their past efforts to intentionally mislead journalists and the public. It's advocacy data-analysis. It's unethical, wrong, and harmful to the cause of truth.

Here's my parody of the Brennan Center style, adopted for 2019. The numbers I use are actually accurate, based on the best available city-data. The logic and conclusions and push, however, are just as absurd.
Crime in 2018: Final Year-End Data

Chicago accounted for more than 34 percent of the murder decrease last year, according to a new analysis of crime data based on faulty methods often used by the Brennan Center.

January 4, 2019

This analysis finds that Americans are less safe today than they have been at almost any time since 2014.

Based on new year-end data collected from the 30 largest cities, murder in 2018 remained higher than just 4 years ago. Although there are some substantial decreases in murder in specific cities, these trends do not signal the start of a new national crime drop. What’s more startling, this analysis finds that the decrease in murders is even more concentrated than initially expected. Just three cities — Baltimore, Chicago, and Columbus — accounted for more than half (59.9 percent) of the decrease in murders. Chicago alone now accounts for more than 34.3 percent of the total decrease in urban murders.

Final Year-End Findings:

• The murder rate fell in this group of cities last year by 7 percent.

• Amazingly, Chicago accounted for 34.4 percent of the total decrease in urban murders.

• Three cities — Baltimore, Chicago, and Columbus — accounted for more than half (59.9 percent) of the decrease in murders.

• Some cities are experiencing a decrease in murder while other forms of crime remain relatively high. Celebration about a national crime drop are premature, but these trends suggest a need to understand how and why murder is decreasing in these cities.

Highlights of this style (faulty logic obscured by dressed-to-impress layout, footnotes, and statistical concepts).

1) The murder rate fell in this group of cities last year by 7 percent.
* "In this group of cities" added only when called out. http://www.copinthehood.com/2017/07/two-year-increase-in-homicide.html

2) Amazingly, Chicago accounted for 34.4 percent of the total decrease in urban murders.
*Note: this simply is not true. But is a reflection of only looking at a number cities.

3) Three cities — Baltimore, Chicago, and Columbus — accounted for more than half (59.9 percent) of the decrease in murders.
*This is true when one includes the caveat "of the sample used." And if one includes this caveat, the statement is statistically worthless.

4) Celebration about a national crime drop are premature — America remains much more violent than just 4 years ago — but these trends suggest a need to understand how and why murder is decreasing in these cities.
*If you cherry pick the baseline year, you can say anything!

One lesson is always be suspicious of data presentation. Is somebody pissing on your leg and saying it's raining? Trust your gut or your "lying eyes." When crime is up and people say it's not, be wary. But use the same vigilance when crime is down and people say "be afraid!"

Know your source, if possible. Assuming people aren't just making numbers up, see when people use one form of logic when data go one way, but sing another tune when the same data go in the opposite direction. (Could be crime, the stock market, gas prices, etc.)

Luckily, murder really was down in 2018. I wouldn't want to waste your time pretending otherwise.

January 1, 2019

State variation in police-involved shootings

Welcome to 2019!

I've compiled the past four years of Washington Post data on those shot and killed by police. Four years gives us a reasonable amount of data. The first thing that jumps out is that the number of people killed by police has remained strikingly constant each and every year for which we have data (from the Washington Post).

The other thing that continues to jump out (I've written about this before) is the state-by-state variation.

The national annual average (2015-2018) is 0.31 (rate per 100,000). And yet New Mexico is 0.98 and New York is 0.09. This is a large difference.

Or take Utah (because of this story in the paper). Utah has a murder and violence rate below the national average, a low poverty rate, and is 90 percent white. And yet people in Utah are almost 5 times as likely an in New York to be killed by a cop. Utah has murder rate lower than NYC, 1/5 the poverty rate, far fewer cops, and Utah is 90% white. In 2018, the rate of people shot and killed by police in Utah is multiple times higher than NYC.

I'd speculate significant variables are (in no particular order) training, fewer cops per capita, fewer cops per mile (no backup), one-person patrol, more guns, gun culture, more meth, more booze, and race (with more white states having more police-involved shootings).

The ten leading states -- as in cops most shootingest states -- in rank order, are New Mexico, Alaska, Oklahoma, Arizona, Colorado, Nevada, Wyoming, West Virginia, Montana, and Idaho. It certainly seems like if we were to focus on the states that have the highest rates of police-involved shootings (and by far), we could find some low-hanging fruit to reduce the number of said shootings. But to do this we'd have stop thinking of police-involved shootings as primarily related to race.

Collectively the top-10 states are 4.9 percent African American (compared to 13 percent nationally). These are the cowboy states out west. The 10 states with the highest percentage of black population (collectively 25%) have a rate of police-involved homicide (0.24) that is below the national average.