In Baltimore Arrests are Down and Crime is Way Up

by on June 12, 2015 at 7:28 am in Current Affairs, Economics, Law | Permalink

A debate forum at the New York Times begins:

A rise in gun violence in New York, Baltimore and other cities after months of angry protests over police killings of unarmed black men, have led some to see a ”Ferguson effect,” in which police, spooked by criticism of aggressive tactics, have pulled back, making fewer arrests and fewer searches for weapons.

But has a wave of murders and shootings brought an end to the long drop in crime?

I don’t think that we will see a sustained increase in crime at the national level. But there is no question that we are seeing a Ferguson effect in Baltimore–more precisely a Freddie Gray effect. Arrests in Baltimore have fallen by nearly 40% since Freddie Gray’s funeral and the start of the riots on April 27. In the approximately 3 months before the Gray funeral police made an average of 87.7 arrests per day, since that time they have made only 54.6 arrests a day on average (up to May 30, most recent data).

Arrests

As Peter Moskos argues:

In Baltimore today, several police officers need to respond to situations where formerly one could do the job. This stretches resources and prevents proactive policing.

Not all arrests are good arrests, of course, but the strain is cutting policing across the board and the criminals are responding to incentives. Fewer police mean more crime. As arrests have fallen, homicides, shootings, robberies and auto thefts have all spiked upwards. Homicides, for example, have more than doubled from .53 a day on average before the unrest to 1.35 a day after (up to June 6, most recent data)–this is an unprecedented increase–and the highest homicide rate Baltimore has ever seen.

Put differently, the unrest in Baltimore and subsequent reduction in policing is responsible for roughly twenty “excess” deaths. (so far)

Homicides

It’s not just homicides, the number of shootings in Baltimore has more than tripled. Shootings increased from .82 a day before the unrest to just over 3 a day. Since the onset of the riots there has hardly been a day without a shooting.

Shootings

Robberies are up from 8.1 per day to 13.25 per day on average.

Robberies

Even auto thefts are up from 9.6 per day to 13.6 per day on average.

AutoTheft

The increase in homicides and other crime is terrible and it is also putting a strain on police resources.

Police Commissioner Anthony W. Batts said the rise in killings is “backlogging” investigators, just as the community has become less engaged with police, providing fewer tips.

With luck the crime wave will subside quickly but the longer-term fear is that the increase in crime could push arrest and clearance rates down so far that the increase in crime becomes self-fulfilling. The higher crime rate itself generates the lower punishment that supports the higher crime rate (see my theory paper). In the presence of multiple equilibria it’s possible that a temporary shift could push Baltimore into a permanently higher high-crime equilibrium. Once the high-crime equilibrium is entered it may be very difficult to exit without a lot of resources that Baltimore doesn’t have. I have long argued that high-crime areas need more police but the tragedy is that they also need high-quality policing and that too is made more difficult to achieve by strained budgets and strained police.

1 Rich Berger June 12, 2015 at 7:48 am

I am shocked that the police would react this way to being categorized as an occupation force. They have feelings, too.

2 PIzza Man June 12, 2015 at 11:30 am

I don’t think it is primarily an emotional reaction. Large riots have big social costs in terms of lives and property. So it makes sense for them to adjust their strategy now that the risk profile to their actions has changed.

3 colleteral June 12, 2015 at 1:54 pm

If they are on strike then they ought not to be paid.

4 Big Bill June 12, 2015 at 5:04 pm

They are not “on strike”. They are giving the people of Baltimore the level of policing they demanded.

5 collateral June 12, 2015 at 7:02 pm

If they are superfluous they ought to be laid off. Any way you cut it, police officers shouldn’t be getting a paycheck for not doing anything.

6 Panamared June 13, 2015 at 3:20 pm

This is exactly what the people are paying for with their tax dollars. They demanded more accountability from the police, so the police are giving them that accountability for their actions. But only when they decide something is warranted enough to show up for it and the consequences are something they can prove validate their interference. Otherwise, buy a gun and good luck.
Welcome to the hell you created Baltimore

7 Popeye June 14, 2015 at 11:30 pm

So some government workers were abusing their power, and when they were called out on it they decided to stop doing their jobs. Obviously these leeches sucking on the taxpayers’ teats would be very popular in this community.

8 Mike June 14, 2015 at 8:45 pm

It’s says ‘several officers have to respond to situations where one used to be the norm.’ They aren’t on strike, they are doing the job the way they are instructed. Sad fact is that all the attention turned to one shooting by one cop, while the rest of the city is shooting 3 people a day.

9 Larry Siegel June 13, 2015 at 3:41 am

Actually, an occupation force is exactly what is needed. If any population other than lower class African Americans were suffering from the level of violence that occurs in those communities, they would consider it an act of war and would demand (and probably get) an overwhelming military response, which would take care of the problem in about a half hour. Jesse Jackson and Al Sharpton ought to be demanding safe streets and eradication of the gangs instead of complaining about the occasional abuse of police power.

10 Ellie Kesselman June 14, 2015 at 7:57 pm

Jesse Jackson and Al Sharpton don’t live in Baltimore, so it is easy for them to complain about things then leave a mess behind them. They have the choice of living somewhere else.

Occupation forces do work well, but the problem is that they need to be maintained. The National Guard could make Baltimore safe again, but then no one would ever want them to leave.

11 mavery June 12, 2015 at 7:49 am

And we’re sure this isn’t something due to the cyclic nature of the calendar? I’m honestly asking. I have no idea if violent crimes tend to increase as we move from winter to spring to summer. It’s just an obvious possible explanation.

12 mavery June 12, 2015 at 7:52 am

It’s also worth asking how much we should trust numbers that aren’t things like homicide. Crime rates for these have historically bounced around depending on the political whims of the mayor/police commissioners. Homicides and shootings are tougher to sweep under the rug, so I tend to believe that number more than the others.

13 John Thacker June 12, 2015 at 9:08 am

Then you should in general be persuaded by the post. The homicide rate has increased the most, and it’s an all time record.

Baltimore does need policing, and more of it, to move to a better equilibrium; it also needs high quality policing, as Alex said. What is was previously getting was a lot of low quality policing, partially because Mayor O’Malley felt that he wanted to make the tradeoff of lowering the quality of policing in order to get more of it. (He saved spending money for things like the city owning a money-losing Hilton next to the Convention Center.)

14 KB June 12, 2015 at 9:04 am

No, we’re not. Which is what makes this a quite irresponsible post.

15 Cliff June 12, 2015 at 9:31 am

Yes, we are.

16 Thiago Ribeiro June 12, 2015 at 10:47 am

Yes, we we isn’t not.

17 Dan June 12, 2015 at 1:17 pm

May was Baltimore’s most murderous month in more than 40 years. That span included more than 100 Junes, Julys and Augusts which would have been hotter. If you want to rationize away the obvious, you’ll have to try harder.

18 Ellie Kesselman June 14, 2015 at 8:05 pm

It isn’t an irresponsible post. The Baltimore Sun newspaper has been reporting on this for the past week.

19 XVO June 12, 2015 at 9:52 am

So you’re saying you would discard the obvious conclusion in favor of some unexplained never before existed seasonal factor that reduces arrests at the same time that crime is increasing?

Your cognitive dissonance is showing.

20 K. June 12, 2015 at 10:30 am

Occam’s razor cuts both ways. There are a couple plausible simple explanations: either Alex’s assertion is correct and crime has gone up due to reduced policing; or there is a seasonal trend that would be revealed simply by extending the graphs by a couple of years to the left.

Another plausible explanation is that the F Gray event caused a social disruption and a temporary increase in violence, independent of what the police are or aren’t doing, and the violence will reduce itself. Similar to unemployment responding to an economic shock.

Or it’s a combination of all three factors, and maybe some others that we haven’t thought of. Whatever the truth is, there isn’t enough data in this post to show clear causation; at best we get correlation.

21 adam June 12, 2015 at 10:46 am

There is actually well-observed correlation between summer months and increasing crime.

http://www.bjs.gov/content/pub/pdf/spcvt.pdf

22 Jay June 12, 2015 at 12:36 pm

There is a pretty marked difference in the timing and as others have said, the degree of both measurements going up and down seems to imply something more than seasonal shifts.

23 adam June 12, 2015 at 12:54 pm

And the timing also coincides with the weather warming. The post doesn’t even show year-to-year changes. It’s worthless for figuring out causation.

24 Mike June 12, 2015 at 12:55 pm

“seems to imply” is pretty weak tea. There’s obviously going to be seasonal/weather-based trends, and if it were me I would at least *try* to account for some of them before accusing a city of going on a crime spree.

25 Urso June 12, 2015 at 12:41 pm

I thought everybody knew this. “Increased ice cream sales lead to increased crime” is a chestnut in Stats 101 textbooks to show the correlation/causation distinction. Summer leads to increases in both

26 Rich Berger June 12, 2015 at 7:52 am

I have read reports that crowds are gathering around police when they are making an arrest, hoping for an incriminating video. Under these conditions, combined with the pressure coming from the Obama administration, what would a rational cop do? Reduce the opportunities for prosecution. Hey citizens! You’re on your own, best of luck!

27 mavery June 12, 2015 at 7:54 am

You appear to be arguing that the earlier arrest rate was the optimal one. I don’t think that’s clear.

28 Cliff June 12, 2015 at 8:49 am

Yes, it is so tough to determine the optimal murder rate. We will have to do some more studies.

29 mavery June 12, 2015 at 11:24 am

Who said anything about murder rates? I was talking about arrest rates. Rich is saying that cops are facing more scrutiny when they arrest people, and as a result, fewer people are getting arrested. I read this as him implying this was a bad thing, which would presuppose that the prior arrest rate was better (not necessarily “optimal” as was pointed out below) than the new arrest rate. I drew this conclusion because he framed this in terms of cops becoming derelict in their duties as a result of the increased scrutiny. (“Hey citizens! You’re on your own, best of luck!” is what Rich posits the cops are thinking.) But perhaps this wasn’t what he meant, as he states below.

30 Matthew June 12, 2015 at 11:30 am

If what you want is zero murders, that’s somewhat easy. We could put tracking chips in everyone. We could even kill everyone (no people -> no murders). But if you want perform an optimization that considers cost and morality of tactics to reduce crime, it’s significantly more difficult

31 Malthusian Delights June 12, 2015 at 1:20 pm

To be fair, I think killing everyone would probably leave you with the highest possible murder rate.

32 Thiago Ribeiro June 12, 2015 at 1:33 pm

In the short run, but in the short run, no more deaths. In the long we will have been all dead.

33 jon June 12, 2015 at 7:25 pm

It’s even easier than that:
1. Invent a drug that causes birth defects that can lead to precognitive abilities.
2. Set up an entire Precrime police division based around three such “precog” children.
3. If you are the creator of the program and understand its capabilities and limitations, DO NOT kill the mother of those three precogs within their telepathic range. Seriously, just meet the mother one town over, it’s worth the extra time and expense.

34 wm13 June 12, 2015 at 12:20 pm

Alex said the increase in homicides was “terrible.” Obviously, if the new arrest rate is optimal, then the increase in homicides is not “terrible,” is at most a regrettable byproduct of a desirable change. So your real quarrel is with the sob sisters like Alex who wail about an overall beneficial event..

35 Thomas June 13, 2015 at 1:41 pm

“if the new arrest rate is optimal, then the increase in homicides is not “terrible,” is at most a regrettable byproduct of a desirable change.”

Sorry, that is not logical.

36 Rich Berger June 12, 2015 at 8:52 am

No, I wasn’t.

37 Pshrnk June 12, 2015 at 9:11 am

Optimal is a silly word to use. Better or closer to optimal, sure.

38 mavery June 12, 2015 at 11:19 am

True.

39 ZZZ June 12, 2015 at 9:49 am

No, we haven’t determined the optimal policing rate but we have determined what side of the curve we are on. It may or may not have been too low before Freddie Gray but it’s definitely too low now.

40 Slocum June 12, 2015 at 9:17 am

First of all, video had nothing to do with the Freddie Gray case — the injuries that caused his death occurred off-camera in the police van. As for the other prominent cases (Walter Scott, Tamir Rice) where video was critical — these were not borderline cases where police could claim that ‘minor’ mistakes or errors in judgement were being criminalized. In Ferguson, video would likely have supported the police version of events and prevented a great deal of harm and the cop in question would probably still have a job. Police should welcome rather than fear video.

41 dearieme June 12, 2015 at 9:27 am

“Police should welcome rather than fear video.” I suspect that that should read “Most police …..”. If I were a policeman, I’d also like to know who gets to edit the video.

42 Sam Haysom June 12, 2015 at 10:53 am

Exactly perceptions of the Rodney King case varied around two factors: race and how much of the video people had seen.

43 JonFraz June 12, 2015 at 2:49 pm

Re: the injuries that caused his death occurred off-camera in the police van.

Of immediately before he was placed in the van– one line of speculation here is that he roughed up pretty bad during his apprehension. The whole business has a huge question mark hanging over it.

44 Hasdrubal June 12, 2015 at 11:55 am

This strikes me as a false dichotomy: If we cannot use our previous unfettered policing techniques, we cannot police at all.

It seems more like a negotiating tactic intended to create a problem, more crime in this case, that can be used as an example of what happens if the police give up _any_ power. Of course, they’re also communicating their intentions loud and clear, so that the people they’re policing will take advantage of the lull, and proving their point for them.

I expect officials in multiple jurisdictions to cite this Baltimore crime spree as a strong defense against any police reforms they find unpalatable.

45 JonFraz June 12, 2015 at 2:48 pm

To be fair to the police I don’t think they are getting very much guidance The bureaucracy in this city is remarkably dense in its intelligence and glacial in its reactions. I’m willing to accord the police a bit of understanding given that the city’s leadership has shown itself so remarkably incompetent over the whole business.

46 Thiago Ribeiro June 12, 2015 at 5:49 pm

“I have read reports that crowds are gathering around police when they are making an arrest, hoping for an incriminating video.”
“So what? If you did nothing wrong, you have nothing to hide.”

47 Eirik June 12, 2015 at 7:53 am

Sorry Alex, but your analysis seems to be a bit quick. We know that crime is seasonal (see http://www.bjs.gov/content/pub/pdf/spcvt.pdf for example), and you show us a difference estimate? At least provide the diff-in-diff if you are going to draw conclusions.

48 Moreno Klaus June 12, 2015 at 7:59 am

I was also thinking about that… but i think he states the homicide rate is the highest Baltimore has seen, or something like that

49 Jamie_NYC June 12, 2015 at 12:26 pm

Global warming!

50 JonFraz June 12, 2015 at 2:45 pm

The homicide rate is higher than it has been for a good long while, but it is not a record high (the drug war period in the 80s and 90s had a higher rate)

51 Alex Tabarrok June 12, 2015 at 8:26 am

Seasonality doesn’t explain the decline in arrests! Moreover, it’s the sharpness of the decline in arrests and the increase in crime that is telling. Seasons change slowly but the change here is quick–within days or a week or two. I have done the year to year comparison, however, and the results are the similar. There is an uptick in the summer months but there are big increases in crime this year compared to the same time in previous years. I didn’t include for reasons of space and time.

52 Bill June 12, 2015 at 8:50 am

Yeah, I just wonder why they took a 4 month period and didn’t show the same period in the previous year. How many property crimes are in February versus May?

Also, are there significance bars? What is the variance?

Also, look at the base for some of these numbers: shootings go from two to 4 and down to zero, and up to 4, etc.

Come on. And, the statement: ” Seasonality doesn’t explain the decline in arrests”, has many problems: First, seasonality could explain it, particularly when you don’t show the previous year; Moreover, though, could it have been
that there had been a decline in frivolous or unlawful arrests??? Why do you assume that the arrests in both periods were lawful? Notice that the arrest table doesn’t say arrested for what. It just says: arrests.

53 John Thacker June 12, 2015 at 9:15 am

“Yeah, I just wonder why they took a 4 month period and didn’t show the same period in the previous year.”

For one reason, as you yourself point out Bill, there was a big change in marijuana arrest policy in December which would be confounding. Looking at year over year would be more misleading than this data.

54 Bill June 12, 2015 at 9:30 am

John, Look again. The table at the top is ARRESTS, it is not ARRESTS for murder. Total arrests include arrests for marijuana.

And, I am not saying that fewer MJ and false arrests lead to higher murder rates. I am not your puppet that you can put your words in my mouth and then claim I said them. It just shows how poor you argue when you do not confront facts, such as the change in arrest for MJ (which by the way is not instantaneous, but over time) and not correcting for the composition of arrests over time but just looking at total arrest, and also not looking at statistical significance in changes in murders, for example, over a small base.

Second, the same article and data show a decline in arrests over time re marijuana (see above). Yet, Alex says he looked at arrest data for his eyeball time series and reported that there was no seasonality and that the decline in arrests just happened recently. Hmmm. I claim Alex did not do this, unless MJ was a small proportion of arrests.

Third, again, you do not know the composition of arrests: MJ, resisting, etc. in each period. You also overlook the WSJ article which says police are being more careful in their arrests.

Fourth, there is no seasonality comparison, no comparison of arrests by crime; nor is there any seasonality adjustment for property crimes, murders, etc.

55 John Thacker June 12, 2015 at 10:05 am

Bill, do you disagree that the answer to your question “I just wonder why they took a 4 month period and didn’t show the same period in the previous year” is because there was a big change in marijuana arrest policy starting October, definitely with massive effect by December, and thus the same period in the previous year would definitely show a lot of extra pot arrests?

The change in overall drug arrest numbers by December, year over year, is so massive that, with a light assumption that not every single drug arrest included pot (and that Baltimore continues to make *some* pot arrests), it was likely nearly complete by February 1.

In any case, since drug arrests of all types had already fallen by over one half– which led to a 29% decrease in overall arrests– even if the change in pot arrest policy had led to absolutely zero drug arrests of any kind you still wouldn’t expect a 40% fall from the February-April period to May.

56 Bill June 12, 2015 at 11:18 am

John, You have no data on the composition of arrests in both periods. The total arrest number is meaningless if the composition changed…that is, if there are less MJ arrests, less spitting on the sidewalk and loitering arrests, less resisting arrests (because that is more carefully reviewed), etc. Alex’s piece is really flawed and superficial.

57 Eirik June 12, 2015 at 8:53 am

Good to know that you have done the year to year comparison, and I agree and/or understand all points.

However, “Seasons change slowly but the change here is quick–within days or a week or two” rubs me the wrong way. There are plenty examples of seasons that change dramatically and suddenly: e.g. consumption of eggs and yellow soda during Easter in Norway or (probably) alcohol related crimes in Mardi Gras in New Orleans to take two silly examples.

By the way, thank you (and Tyler) for the excellent blog. It is often though provoking and forces me to evaluate my own political views often. Much appreciated!

58 Jimmy Bae June 12, 2015 at 9:51 am

> I didn’t include for reasons of space.

That is non-sensical. One of the very best things about web publishing is that there are no limits on space.

If you want to make the post easier to read, then do a TL;DR version and a fully documented version. If you don’t have the time to make the fully-documented one pleasant to read, then just footnote the TL;DR and link to a data-dump.

59 mavery June 12, 2015 at 11:29 am

Not to mention a chart showing a longer time series and accounting for seasonal changes would more informative (if less dramatic) than the ones included.

Regardless, a time series fit of some sort rather than a simple mean would (1) be trivial to do and (2) be a lot more useful.

60 JonFraz June 12, 2015 at 2:45 pm

You seem like a very bright guy. I assume you have heard of the “post hoc, propter hoc” fallacy?

61 cb June 12, 2015 at 4:53 pm

This year homicides in May were about 80% higher than April. Last year they were 90% higher for the same month to month comparison.

It also looks like every month this year has been higher than last for homicides.

62 brad the enlightened lib June 12, 2015 at 7:54 am

“I have long argued that high-crime areas need more police but the tragedy is that they also need high-quality policing”

High quality policing? What’s that? Let me guess, use “root cause education” rather than police? The root cause of violence is a propensity towards violence and poor impulse control. Good luck confronting and dealing with violent people without using strong arm tactics. They don’t care about your education or your blue values for that matter. Education and policing aren’t the same thing. Liberals need to stop confusing themselves that they are the same. Blue lib polices created what Baltimore has become. Telling the cops they can’t do their jobs won mid-term elections but cost residents of criminally infested areas what little peace they had left.

Send in the shock troops and grin and bear it.

63 Art Deco June 12, 2015 at 7:56 am

For quality policing, see James Q Wilson and William Bratton. “Strong arm tactics’ is a stupid way of describing it.

64 brad the enlightened lib June 12, 2015 at 8:00 am

Send in the shock troops and grin and bear it.

65 Just Another MR Commentor June 12, 2015 at 8:11 am

Yeah Kill those animals! That’s just noticing

66 Thomas June 13, 2015 at 2:40 pm

This troll is so man, lol. Talk about failing as a troll.

67 Floccina June 12, 2015 at 8:37 am

Schooling like most cure all’s does nothing, you need to address problems as directly as possible.In fact if your problem is violence and theft switch money and effort from schooling to more and better police, courts and punishment. Also at least lets try end the war on drugs and other non-violent or theft crime.

68 Art Deco June 12, 2015 at 9:25 am

Also at least lets try end the war on drugs and other non-violent or theft crime.

Rubbish. Theft is properly punished and both are components of the disorder that breeds violent crime. And it’s not as if drug dealers and thieves are incapable of violence.

69 Floccina June 12, 2015 at 11:37 am

Sorry I meant to write:

Also at least let’s try end the war on drugs and other non-violent crime.

70 Floccina June 12, 2015 at 11:42 am

Also at least let’s try ending the war on drugs and other non-violent crime.

71 Floccina June 12, 2015 at 12:01 pm

I have been really messing up. I meant to say:

“Also at least let’s try ending the war on drugs and legalize other non-violent, non-theft crimes.”

I was on a device hard to type on and it was before coffee.

72 Albigensian June 12, 2015 at 10:22 am

“Quality policing” surely is in need of a good (or at least usable) definition.

I’ll start by saying that at a minimum “quality policing” means maintaining a reasonably high clearance rate for serious crimes of violence. For without that, few will be willing to risk themselves and their families in order to inform on these violent criminals, let alone be willing to testify against them in court.

It may seem from the outside (especially by those who romanticize violent criminality) that everyone in these neighborhoods is plugged into a culture of violent crime, but that’s not even close to being true. There are plenty of people just trying to get by in these places who can’t afford to move to someplace better. These people deserve and desperately need quality policing.

And the core of that is, solving serious crimes of violence. Witness intimidation is all too real, and few will risk informing or testifying unless they can expect that they and their families can and will be adequately protected, and that their efforts are likely to result in conviction(s) and lengthy sentences.

73 Agra Brum June 12, 2015 at 2:00 pm

The police were not doing their jobs when they arrested Gray (for the crime of running from the same police). This isn’t the police feel they can’t “do their job” – they just are just purposefully engaging in a slowdown. There is a third way between a lawless city and lawless police.
How is random police brutality helpful to a city?

74 Aaron June 14, 2015 at 2:27 pm

Yes they were. Stopping suspicious people, such as known and convicted repeat drug dealers, who are acting suspiciously in a high crime area while carrying a knife is what i expect cops to do. Thats what makes cities a safer place. When Freddie ran, the cops chased, again like they are supposed to do. The problem is that if the citizens don’t want they type of law enforcement, then that is their right. The citizens in Baltimore have made it clear that they don’t want that, so the cops reduce the level of enforcement. End result is higher crime; but the Baltimore citizens have the right to want that scenario. The cops serve communities, and the communities should get what they want. It is sad what the result is for places like Baltimore, but fortunately I choose to live away from that poor city…

75 Art Deco June 12, 2015 at 7:55 am

Baltimore already had elevated crime rates from understaffing of the police (consequent to a truncated tax base) and from a refusal by the ticket-punchers and dolts in its political class to go to school on New York’s successes. Congratulations to Mayor Rawlings and State’s Attorney Mosby, this is your legacy.

76 brad the enlightened lib June 12, 2015 at 8:03 am

Baltimore already had an out of control criminal element. Freddie Gray was nothing more than an excuse to do what they enjoy doing.

77 John Thacker June 12, 2015 at 9:10 am

It’s Martin O’Malley’s legacy as well. He decided to flood the city with lots of low quality policing (which is cheaper than doing a good job) for a short term drop at the expense of the long term, which lead to the inevitable backlash.

78 Art Deco June 12, 2015 at 9:27 am

The city isn’t ‘flooded’ with police and you can verify that yourself by examining the numbers on ‘city-data’ (and what’s your evidence that the police force as individuals are ‘low quality’?).

79 John Thacker June 12, 2015 at 9:59 am

what’s your evidence that the police force as individuals are ‘low quality’

Didn’t claim that. High quality individuals can be pressured to implement low quality policing, if that’s the metrics by which they get commendations and promoted.

80 John Thacker June 12, 2015 at 10:00 am

You seem to think that any criticism of police tactics is the same thing as criticism of the police as individuals themselves. I happen to think that Baltimore police officers as individuals are perfectly capable of engaging in high quality policing, if the leadership asks for it, and equally capable of engaging in low quality policy, if leadership and incentives ask for that too.

81 Art Deco June 12, 2015 at 11:59 am

You’re going to have to explain to us all why ‘low quality’ policing is cheaper. It does not reduce your payroll and legacy costs, so what’s the savings?

82 Floccina June 12, 2015 at 1:03 pm

@Art Deco the problem as I see, is what I call the Phil Crosby quality problem. The criminal justice system works so poorly that some police have taken it upon themselves to become the courts and punishment system and that pushing the problem down to low a level and so the problem never gets fixed.

https://www.google.com/webhp?sourceid=chrome-instant&rlz=1C1CHMO_enUS527US527&ion=1&espv=2&ie=UTF-8#q=phil+crosby+quality+is+free

83 Moreno Klaus June 12, 2015 at 7:57 am

“Fewer police mean more crime.” Thanks Cpt Obvious. The main point is that if the communities see the police as the enemy, their job will be always much harder, and my question is whether it is the police that “shot themselves on the foot” on this subject? I mean, it seems a lot like “Shoot first, ask later”, is the standard procedure if the person is black, (and it this is the image black people have of the police welll….) but even if this totally just “msnbc”-gibberish, this is still the country (“the land of the free” loool) that has higher black incarceration rates than South Africa during apartheid looool….

84 Moreno Klaus June 12, 2015 at 7:59 am

Well, criminals are not stupid… if they know the police is on the retreat…

85 Pshrnk June 12, 2015 at 9:32 am

Actually most criminals are relatively stupid. The folks who cannot figure out what the stupid criminals have understood ????

86 Cliff June 12, 2015 at 8:55 am

White police are less likely to shoot a black suspect than a white one. Black police are more likely to shoot a black suspect. Differences are small, though.

87 Moreno Klaus June 12, 2015 at 9:22 am

p-value fishing?

88 Cliff June 12, 2015 at 9:33 am

??

89 Cliff June 12, 2015 at 9:35 am

Anything that does not line up with your priors is probably a poorly done study? Is that the argument?

http://slatestarcodex.com/2014/11/25/race-and-justice-much-more-than-you-wanted-to-know/

90 Tarrou June 12, 2015 at 9:54 am

Thanks Cliff, I was gonna link that! It has become my go-to for all race-related CJ arguments. I certainly don’t have the patience to look all that up 😛

91 Tarrou June 12, 2015 at 9:20 am

Blacks are less likely than whites to be shot in a situation in which a suspect is firing a gun at police (something I think we can all agree on where deadly force is justified). Police do discriminate in use of deadly force…..against whites.

92 Red June 12, 2015 at 12:58 pm

this is still the country (“the land of the free” loool) that has higher black incarceration rates than South Africa during apartheid looool

What’s your point? “Democratic” South Africa has a higher black incarceration rate than apartheid South Africa did.

93 Larry Siegel June 13, 2015 at 3:54 am

Well, it is democratic, and the incarceration rate *just might* have something to do with the number of people who have done something that justifies incarcerating them.

94 R Richard Schweitzer June 12, 2015 at 8:20 am

What are the demographics behind the graphs presented?

WHO is getting murdered; by WHOM?

WHO is getting robbed (or burglarized) by WHOM?

Who is paying ( or expected to pay) for the prevention of those incidents? How do they benefit?

Are we beginning to see (and understand) cases where certain populations, with particular motivations (and lacking others) self-destruct?

95 what? June 12, 2015 at 8:32 am
96 Floccina June 12, 2015 at 8:26 am

1, Suppression of crime is by far the most important job Government does. The Baltimore Government should slash spending in other areas (schooling) and spend it on more and better police.
2. IMHO We really at least need to try ending the war on drugs and see if that helps.
3, How long until offsetting behavior starts to make black society very careful, protected and polite?
$. Would allowing more segregation by race hurt or help? That is with say black police in segregated black areas?

97 JonFraz June 12, 2015 at 2:32 pm

Re: 1, Suppression of crime is by far the most important job Government does.

Not even. Historically vastly more people died from tainted food and water than from criminal activity. Maintaining basic sanitation and clean water is therefore more important than policing (not that policing and justice are trivial– they aren’t.)

98 8 June 12, 2015 at 8:37 am

The solution is to reduce arrests way, way more. Legalize drugs.

99 Floccina June 12, 2015 at 8:43 am

+1

100 John Thacker June 12, 2015 at 9:17 am

Maryland largely decriminalized pot last year, and Baltimore changed its arrest policy at the end of last year. That didn’t seem to have a huge effect one way or the other on violent crimes (yeah, it’s possible people moved into other drugs, and legalization should reduce crime more than decriminalization) unlike the recent actions.

101 Tarrou June 12, 2015 at 9:17 am

necessary but not sufficient, methinks.

102 Gunther June 12, 2015 at 9:57 am

“Legalize drugs.”

Anesthetize yourself and pretend nothing’s wrong.

103 Art Deco June 12, 2015 at 12:32 pm

About 20% of prison and jail inmates are there on drug charges, and the allocation of national product to local police forces is modest (< 2% thereof). The notion that vast resources are being devoted to enforcing drug laws is a dumb meme that will never die.

104 fwiw June 12, 2015 at 10:13 pm

I’d argue that 20% is vast resources, and that number is probably understated by people in jail for, for example, not paying court fees or any of that other bs we jail poor people for.

Anyway, do you have citations for those stats? I’d like to read the original source; sounds interesting.

105 Floccina June 13, 2015 at 6:34 pm

+1

106 rayward June 12, 2015 at 8:45 am

The Freddie Gray effect isn’t just a reaction to (perceived) police brutality but to the dismal opportunities for black men, especially young black men; the incidents between blacks and police exposed much more than (perceived) police brutality, and the spike in crime is likely attributable to released frustration with the dismal opportunities as reaction to (perceived) police brutality and the reduction in (effective) policing. And while I would expect that the tension between blacks and police will subside, the tension caused by dismal opportunities for black men won’t. In this respect, Tabarrok’s solution, more (effective) policing, won’t solve the problem.

107 Cliff June 12, 2015 at 8:57 am

Dismal opportunities like being admitted to any school you apply to if you have mediocre test scores? Like being preferentially hired for high status jobs like big firm attorney positions?

108 Craig June 12, 2015 at 9:56 am

Preferential hiring for life-long government jobs…

109 Tarrou June 12, 2015 at 9:13 am

There are always opportunities. I was a poor minority HS student, and wilder than most. I still got a scholarship to college, then doubled up and took the other incredibly simple (though definitely not easy) route of joining the services. There you go. Two things any teen, no matter their poverty, race, or sex can do to get to college or a job that provides college later on.

Allow me a short story, I live in a majority-minority city hollowed out by the fall of the auto industry. Massive unemployment, huge crime problems. My company hired me to start a branch in that town, we set out to hire a relatively small starter staff of five people. We advertised all over. Got hundreds of applications. Did weeks of work setting up interviews. Two people showed up to their interviews. Two. We’ve been running the adverts for a year and a half now, and we’ve found two more people who can spell their name and will show up. It’s an easy job, requirements are minimal, company pays for training, pay is far better than minimum wage, and people will not take it.

So I have one thing to say about the “lack of opportunity”. Horseshit. No such thing.

110 Pshrnk June 12, 2015 at 9:37 am

How many drug related convictions did you have before joining the service?

111 Tarrou June 12, 2015 at 9:49 am

None unless we’re counting alcohol as a drug. I did serve with many who did, however, including one guy who was basically a 15-year career gang member with thirty or forty arrests and convictions to his record including four for attempted murder. I don’t recall which ones he was convicted of, he might have beat the most serious charges, but I handled his personnel jacket, and his police record was enormous, and dated back to when he was ten.

Of course, what they will let you in with depends a lot on how badly they need the bodies. Back in ’05, you could have attacked the First Lady with a giant dildo and gotten in. Now, probably not so much.

112 Pshrnk June 12, 2015 at 10:48 am

Great foreign policy to send thugs overseas to win friends and build nations.

113 ivvenalis June 12, 2015 at 11:03 am

When I read your first paragraph I was going to make a comment about The Surge but you beat me to it. Those people don’t get let in anymore. They tended to be bad soldiers, but we needed cannon fodder.

114 Tarrou June 12, 2015 at 11:19 am

and Pshrnk squirts squid ink to avoid the issue! Silly me for treating the commentariat here like adults.

Oh, and ivv, that guy was actually a great soldier. One of the most promising I ever saw come up. The basic attributes that make many criminals are also those that make good soldiers, if they have the discipline to channel it.

The pissing matches over foreign policy are irrelevant. The military is still a job option available to just about anyone, and you don’t have to be “cannon fodder” like me. Lots of jobs in the Navy and Air force if you don’t want to be a real soldier! And I note no one will deal with the second part, which is that I basically walk about a town with sixty-five thousand unemployed in it, mostly poor blacks, and offer a good job with minimal requirements and no one wants it. I go to churches, talk to schools, do presentations, and all that in addition to the actual job! Don’t tell me there aren’t jobs available, I’m sitting on a dozen of them I can’t fill.

115 Tarrou June 12, 2015 at 9:55 am

And nice attempt at deflection.

116 Pshrnk June 13, 2015 at 7:38 pm

Regarding your 11:19 comment:

If you are going to go ad hominem on me at least call me calamari.

117 Larry Siegel June 13, 2015 at 4:00 am

Tarrou, you are probably in the top 10% (of any group) in intelligence. I can tell by the way you write. There isn’t much opportunity for low-ability people. I don’t know how to fix that problem although I’d be interested in your views. I am glad our system provided opportunity for you because there was a time it wouldn’t have.

118 Tarrou June 13, 2015 at 9:28 am

It is true that the first path I noted is easier for those with high cognitive ability, but it is by no means exclusive to them. Public school in the US is so incomprehensibly easy that with a bit of work anyone not completely cognitively retarded can pull a 4.0, though the standardized tests are harder to deal with. The military required (at the time I joined) a 29 on the ASVAB, but you could get a waiver down to a 22…..on a 100-point scale of a multiple choice test. If you can fill in “C” and not eat the pencil for two hours, cognitive limitation is no bar to the services (and it is a prerequisite for officers!).

Low cognitive ability makes many things harder, but it is dealt with by inculcating a few basic attributes. Punctuality, work ethic, honesty, these things will get just about anyone a job. I have a brother with Downs Syndrome. He is severely limited in what he can do, but he holds down several part time jobs because he shows up, every day, on time, and works. The future, as the saying goes, belongs to those who turn up. So do the opportunities.

I was transporting a client the other day, an elderly black lady who was alternately bemoaning and boasting of her grandkids. She was saying her granddaughter was going to beat her older brother (three years older) to a high school diploma. According to her, her granddaughter worked at a department store, was at the top of her class, was class president, captained the track team, and her grandson was a world champion mooch and pot smoker*. If true, one of these kids will have all the opportunity in the world if she can keep focus, and the other will at best drift about the margins of society.

I get infuriated when people insult me and people like me by claiming that we had no “opportunity” and hence should be excused for acting like barbarians. We have all the opportunity in the world here in America. Just try pulling this crap in Russia, or Lebanon, or hell, even France! No matter your cognitive ability (within normal range), if you can not get pregnant, stay out of serious prison time, do your homework, actually apply for stuff, and turn up every day to whatever it is you do and bust your hump, you can advance in this country! And if you can do this and live in the Sagnasty, give me a call, because there are jobs to be had.

*Not to be taken as a slam on pot, I know some very heavy users who are great workers and very successful.

119 Art Deco June 12, 2015 at 9:39 am

“dismal opportunities”? The vast majority of blacks in this country are working class, just like the vast majority of whites. The black population has an unusually large lumpenproletariat, but that’s maybe 15% of the total. A great many working blacks have bad jobs, but have a gander and the occupational enumerations of the Bureau of Labor Statistics: about a quarter of the working population at any one time is ensconced in a crummy job. Having a crummy job is a variant of normal in this country.

You could be of help to slum blacks in general by establishing a greater measure of public order on the streets and in the schools, as well as doing some serious tracking in the schools so that each student not excised for bad behavior is in a program which makes optimal use of his time. More human capital and more congenial public spaces is good. You do have to deal diligently and unflinchingly with threats, though.

120 Tarrou June 13, 2015 at 9:41 am

Quite so. The idiotic focus on college for everyone is doing untold damage to the left side of the IQ bell curve. And the ceding of densely populated ghetto communities to criminals is doing damage to all of them. End the drug war, sure, but violence will always be illegal, and it is rampant in these communities. Even within my town, one of the top ten most violent mid-sized cities twenty years running, the vast majority of the crime only happens in a few neighborhoods on the east and southeast sides. Much of the town isn’t bad at all, but small slices are terrible.

Everyone needs justice and protection, the poor and disadvantaged moreso than the rich and connected. This does not mean police brutality, and we need serious reform of the systems of immunity for both cops and prosecutors. But it does mean a lot of heads are gonna get knocked, and some people will be killed. And there will inevitably be mistakes and violations. But turning these into a race issue when the data clearly shows it is not one is extremely counterproductive. We have a police militarization problem, we have an overcriminalization problem. It has yet to be shown that we have much in the way of a police racism problem.

121 E. Harding June 12, 2015 at 12:12 pm

Dismal opportunities or laziness&criminal records? Blacks have lower average SAT scores at every income level.

122 Art Deco June 12, 2015 at 12:36 pm

So what? The intake pipe for baccalaureate granting institutions encompasses maybe a third of each age cohort. That’s not salient in the lives of most people starting out.

The ratio of employment levels to population is lower for blacks than it is for non-blacks (about 10%), but not near enough different to justify calling blacks ‘lazy’, certainly never without qualification. The black population has a larger lumpenproletariat and a smaller bourgeois element, but the overlap in experience with the remainder of the population is enormous.

123 Floccina June 12, 2015 at 3:01 pm

+1

124 Cooper June 12, 2015 at 7:04 pm

Violent criminals are not a large fraction of any population but they have a hugely disproportionate impact.

It’s possible that 0.1% of white men in Baltimore are violent criminals compared to 0.2% of black men. In the statistics, we would see twice as many black criminals as a share of the population than white criminals. The total criminal share of either population is still quite low.

It might only take a small change in the number of criminals to create a community death spiral.

People will continually flee the more dangerous community in favor of the less dangerous one, even if the initial disparity is small. On the margins, ghettos become self reinforcing.

125 Pshrnk June 13, 2015 at 7:44 pm

Prisonpolicy.org lists 2010 incarceration rates of 2,207 for African American and 380 for white.

126 Floccina June 12, 2015 at 2:53 pm

dismal opportunities for black men, especially young black men

As compared to what? And opportunity to do what? Make enough money to feed, clothe and shelter oneself.

In any city the size of Baltimore in the developed world their is tremendous opportunity.

127 Richard Besserer June 12, 2015 at 8:55 am

Anybody with a working knowledge of time series analysis want to test for a structural break here?

More than four months of data would be nice too. Let’s disentangle the Freddie Gray effect from spring.

128 ibaien June 12, 2015 at 9:00 am

having just watched ‘the wire’, my cynical guess would be that the department has made the decision to juke the stats to prove a point to the community – something to the effect of ‘if you start calling us on our shit, we’ll just stop doing our jobs and you can see what it’s like when we’re not around’.

129 AF June 12, 2015 at 9:13 am

+1

We all know Martin O’Malley did just that to provide the “Baltimore Miracle”

130 Mark Thorson June 12, 2015 at 9:23 am

Exactly what I’m thinking. It’s not that individual cops have become more sensitive, it’s that the union has put the word out to throttle back on policing to send a message to the city, both the people and its government.

131 Art Deco June 12, 2015 at 9:40 am

having just watched ‘the wire’,

Uh. Fiction is fiction.

132 ibaien June 12, 2015 at 9:45 am

you’re totally right, art. a show about the failures of baltimore bureaucracies – especially the police force – written by a former baltimore sun police beat writer, who spent years reporting from inside the department, has nothing to tell us here.

133 Art Deco June 12, 2015 at 11:56 am

It’s fiction and reporters tend to trade in stories. Stories illustrate social reality. They cannot capture it, and they can mislead.

134 fwiw June 12, 2015 at 10:16 pm

but… statistics capture social reality?

135 Agra Brum June 12, 2015 at 2:03 pm

Now they are juking the other way – when a felony dropped to a misdemeanor, or attempted murder was dropped to a simple assault, they are now flipping the stats the other way to help paint a picture of a “City out of control”

136 Jan June 12, 2015 at 9:02 am

Show year over year changes please.

137 Bill June 12, 2015 at 9:06 am

Gee, maybe the decline in arrests has something to do with a decline in FALSE ARRESTS, or not arresting people for non-crimes and then later arresting them for resisting.

Here is a WSJ article on how false arrest and imprisonment charges on three officers is leading to fewer arrests, presumably because officers are being more careful to do LAWFUL arrests:

:University of Baltimore President Kurt Schmoke, who was the state’s attorney in the 1980s before he was elected the city’s mayor, said the charge could help cut down on improper or dubious arrests. “That line, for some officers, seems to have gotten a bit blurred over the last few years,” he said. “That feeds into the young person’s view of being harassed.”

Here is the link:
http://www.wsj.com/articles/illegal-arrest-charge-ripples-beyond-baltimore-1430683705

Also, there has been a decline in marijuana arrests due to a change in public policy beginning in December:

Arrest data from Baltimore City Central Booking and Intake Center (CBIC) shows a precipitous decline in drug arrests in Baltimore City, continuing an ongoing slide that has become more pronounced since Maryland’s marijuana-decriminalization law went into effect on Oct. 1. In December there were 507 drug arrests processed at CBIC, down from 1,018 in December 2013, a 50.2-percent decline.
On one day, Dec. 14, only six drug-arrests were made, compared to 22 on Dec. 14, 2013.
City Paper’s analysis shows that overall arrests are down, too, though more modestly. In 2014, 40,468 arrests were processed at CBIC, down from 44,109 in 2013, a decrease of 8.25 percent. The overall arrest numbers for December, though, show a large, 29-percent drop, from 3,106 in December 2013 to 2,204 last month.
– See more at: http://www.citypaper.com/blogs/the-news-hole/bcp-baltimore-sees-50-percent-decline-in-drug-arrests-in-december-20150109,0,472021.story#sthash.JXubtSru.dpuf

138 John Thacker June 12, 2015 at 9:13 am

You’re arguing that fewer false arrests and marijuana arrests lead to the largest homicide rate Baltimore has ever seen? I certainly hope not. Are you attempting to make the pro-false arrest, street harassment, and marijuana arrest argument?

The marijuana arrest data is irrelevant, since as you point out the change in policy started in December. That’s a very good reason why Alex only looked at the last four months– going back further would confuse the matter. It’s *not* the marijuana arrest policy change that led to the increase in real crimes, it’s the effective police strike. (If they can’t do their jobs badly, they won’t do them at all.)

139 Bill June 12, 2015 at 9:34 am

Reply to your comment was posted above for some reason, so am posting below:
John, Look again. The table at the top is ARRESTS, it is not ARRESTS for murder. Total arrests include arrests for marijuana.

And, I am not saying that fewer MJ and false arrests lead to higher murder rates. I am not your puppet that you can put your words in my mouth and then claim I said them. It just shows how poor you argue when you do not confront facts, such as the change in arrest for MJ (which by the way is not instantaneous, but over time) and not correcting for the composition of arrests over time but just looking at total arrest, and also not looking at statistical significance in changes in murders, for example, over a small base.

Second, the same article and data show a decline in arrests over time re marijuana (see above). Yet, Alex says he looked at arrest data for his eyeball time series and reported that there was no seasonality and that the decline in arrests just happened recently. Hmmm. I claim Alex did not do this, unless MJ was a small proportion of arrests.

Third, again, you do not know the composition of arrests: MJ, resisting, etc. in each period. You also overlook the WSJ article which says police are being more careful in their arrests.

Fourth, there is no seasonality comparison, no comparison of arrests by crime; nor is there any seasonality adjustment for property crimes, murders, etc.

140 John Thacker June 12, 2015 at 9:47 am

Total arrests include arrests for marijuana…It just shows how poor you argue when you do not confront facts, such as the change in arrest for MJ (which by the way is not instantaneous, but over time)

Sorry Bill, you’re the one who is not confronting facts. Your article is from January. Yes, the change in MJ arrest was not perfectly instantaneous, but it had massively declined by December 2014, fairly soon after the law took effect in October. That was enough to account from a 29 percentage point drop in arrests for December year over year– that was not accompanied by an increase in violent crime. The magnitude of the decrease by December actually makes it rather unlikely that the post April 27 drop in arrests was caused by further decrease in MJ.

OTOH, we saw a similar, nearly instantaneous drop in arrests after April 27 that was accompanied by a step change in violent crime.

Again, the very fact that the decrease in the arrest rate accelerated throughout 2014 makes year over year comparisons worse than what Alex presented. The pot decriminalization is a very good point to raise, but it’s a point about why year-over-year wouldn’t make sense.

141 John Thacker June 12, 2015 at 10:14 am

Your article provides excellent evidence that the greatest part of the MJ arrest shift occurred by December. That is good reason to discount a dramatic change in MJ arrests between Febuary-April and May. It is also good reason to avoid using year-over-year comparisons, which would definitely compare pre and post pot decriminalization and arrest changes.

I am confused, Bill, because your article provides some excellent evidence that appears to contradict what you’re actually saying.

You also overlook the WSJ article which says police are being more careful in their arrests.

Perhaps, but if so doesn’t that have negative implications? It’s undisputed that the violent crime rate is way up. Yes, seasonality and the possibility of faking statistics, but homicides are much harder to fake than other crime states and the homicide rate is definitely at an all time high. (Something not true of alarmist articles about non-Baltimore cities.) That’s more than enough to overcome the seasonality objections.

Therefore, if you’re going to sit there and tell me that the crime rate is up, and that the main reason that arrests are down since the violent crime rate went up is because false arrests went up (since it seems very unlikely that the pot arrest rate further declined much from Feb-Apr to May, considering how low it already was in December), well, it sounds like you’re saying that false arrests prevent violent crime. I disagree, but that’s what you sound like.

142 Bill June 12, 2015 at 10:55 am

John,

What you are saying doesn’t make sense, or you are failing to include the logic of it.

Look at what you are saying: A decline in arrests leads to an increase in violent crime.

“”We saw a similar, nearly instantaneous drop in arrests after April 27 that was accompanied by a step change in violent crime.”

Where is the logic to this, particularly when there was a drop in arrests also in the previous period where there was not an increase in violent crime.

Again, go back and think about the composition of the term “arrest” in the first graph, also go back and think in term of causation: what you are saying is let’s go out and arrest people and that will lead to less violent crime. ???? No, we arrest people after they committed a violent crime. One follows the other; it does not precede it. The other funny part of your argument is that arrests declined in the previous period if you include the MJ change in arrests, and there was no change in violent crime, so you dont have evidence that a decline in arrests leads to an increase in violent crime.

143 John Thacker June 12, 2015 at 9:56 am

I can’t put words in your mouth, but I can read the words and facts that you quote.

Bill: “I claim Alex did not do this, unless MJ was a small proportion of arrests.

Bill’s article: “In December there were 507 drug arrests processed at CBIC, down from 1,018 in December 2013… The overall arrest numbers for December, though, show a large, 29-percent drop, from 3,106 in December 2013 to 2,204 last month.”

Pre pot decriminalization policy drug arrests of all type were roughly one-third of all arrests. By December of this past year they had declined by half and to fewer than one-fourth of all arrests. The drug arrest are for all types of drugs, so MJ arrests alone in December must be even fewer as a percentage of all arrests. (Overall arrests declined by more than the drug numbers would imply, which can be argument in favor of the decriminalization policy since the violent crime numbers did not go up.) With such a rapid change in pot arrests, it is highly likely that the drug arrest policy had come close to its new equilibrium before April (and even fairly close before February.)

If we looked at comparisons that included the time when the decriminalization started, such as year-over-year data, it would be worse than looking at the last four months, where it is likely that the pot arrest policy had largely settled down.

144 Bill June 12, 2015 at 11:04 am

John, Total arrests declined in the previous period but violent crime did not increase, which is contrary to your statement that a decline in total arrests leads to an increase in violent crime.

Second, arrests can decline for many reasons, including decriminalization, changes in composition of arrests (no arrests now for loitering, spitting on sidewalk, resisting arrest because cameras are on, etc., or a change in policing (more community policing).

145 Pshrnk June 12, 2015 at 9:40 am

Why is no one discussing that even when police do the best they can they will still have some false positive erroneous arrest rate? There is also going to be a false negative arrest rate. Police may change their criteria to honestly try to improve their performance and get it wrong.

146 Art Deco June 12, 2015 at 9:41 am

a decline in FALSE ARRESTS, or not arresting people for non-crimes and then later arresting them for resisting

Because police in an environment like slum Baltimore have nothing better to do than arrest people at random for sh!ts and giggles.

147 ICantTrustMyFans June 12, 2015 at 9:28 am

Let’s go ahead and call this Chris Hayes/Rachel Maddow effect.

You wanted Occupy Wall St., unicorn-in-the-sky Anarchy? Well, enjoy your new SJW-approved murder rate. The Wire was such a great show. Fucking cops are all scumbags. The heroic drug dealers and honorable, street-wise citizenry are going to make B’more a bastion of PeoplePower and equality.

Mencken and Poe in their graves: Are they chuckling or weeping?

148 indybones June 12, 2015 at 9:35 am

finally real estate prices might go down

149 Jimmy Bae June 12, 2015 at 10:04 am

The NYC reference is misleading. There has been an increase of 5 murders from 11 to 16 over the same time period last year. The data is too noisy to draw any conclusions from. But the over-all NYC crime rate is down 2,500+ from ~39,000 incidents. Those numbers are much less noisy.

http://www.washingtonpost.com/blogs/the-fix/wp/2015/06/05/the-overwrought-very-political-hand-wringing-over-crime-in-new-york-city/

Similarly, the choice to use the date of Gray’s funeral as an end-point in the averages is misleading, leading to have an exaggerated importance. It looks like someone trying to fit the data to a per-determined result. Why not plot a running average over the entire time period and simply mark dates of interest?

150 John Thacker June 12, 2015 at 10:15 am

Yes, I don’t think that the numbers justify claims about non-Baltimore cities. But something is going on in Baltimore.

151 eric June 12, 2015 at 10:11 am

It’s like the trade-off between type 1 and 2 errors. Lots of progressives think there are none.

152 collin June 12, 2015 at 10:19 am

Alex,

If you truly want to cut down the ‘police state’ then don’t go overboard on four months of crime stats in a few cities! (Especially since there is a significant seasonality that effecting the data here.) Crime stats are extremely noisy, have seasonality and we analyzing one city rocked with riots.

Realize the Police State didn’t just grow because of evil politicians, it grew because people feared crime. Thanx for gasoline on the fire LINO!

153 TallDave June 12, 2015 at 10:20 am

Police generally can’t fix institutional problems, so their effect on crime is marginal.

The people of Baltimore, including their police and DA, have to decide to reject savagery and embrace civilized values.

They chose to riot.

154 Moreno Klaus June 12, 2015 at 11:40 am

Yes, black people dont have any reasons to riot… at all… 😉

155 TallDave June 12, 2015 at 3:21 pm

No civilized person ever has a reason to riot, no savage ever lacks for one.

156 TallDave June 12, 2015 at 3:24 pm

BTW, most rioting used to be directed against blacks. Race has nothing to do whether one has civilized values. That’s a choice.

157 TheAJ June 12, 2015 at 11:53 pm
158 TallDave June 13, 2015 at 11:05 pm

Whether or not Samuel Adams helped plan the Boston Tea Party is disputed, but he immediately worked to publicize and defend it.[65] He argued that the Tea Party was not the act of a lawless mob, but was instead a principled protest and the only remaining option the people had to defend their constitutional rights.[66]

Hard to find anything principled in the violence in Baltimore and Ferguson.

159 TallDave June 13, 2015 at 11:10 pm

also — According to Young, American writers were for many years apparently reluctant to celebrate the destruction of property, and so the event was usually ignored in histories of the American Revolution.

People today often assume the BTP was universally celebrated but in fact that wasn’t really the case, the Founders were ambivalent, especially the more conservative elements.

160 Art Deco June 12, 2015 at 11:55 am

Police generally can’t fix institutional problems, so their effect on crime is marginal.

No, they are crucially important, so long as they are properly staffed, properly trained, and given a franchise to do their jobs.

161 TallDave June 12, 2015 at 3:22 pm

Policing can only explain about 10% of the difference in crime rates. High crime areas are generally more, not less, policed.

162 Floccina June 12, 2015 at 4:02 pm

Policing can only explain about 10% of the difference in crime rates. High crime areas are generally more, not less, policed.

But that is not to say that yet more police will not help. For example almost no one holds up a convenience store while an armed policeman is checking out.

163 TallDave June 12, 2015 at 5:32 pm

Yes, but… policing is like giving some aspirin for a serious infection or cancer pain. It may help the symptoms, but the roots of a decision to commit violent crime are still going to dominate the incidence of crime.

164 Art Deco June 12, 2015 at 6:01 pm

Policing can only explain about 10% of the difference in crime rates.

He sayeth, pulling a datum out of his authoritative rectum.

165 TallDave June 12, 2015 at 10:24 pm

Statistics aren’t that hard.

166 TallDave June 12, 2015 at 10:49 pm

For instance, Tabarrok’s study finds crime rates were 15% higher on days without the massively increased police presence of an alert. As he notes, most studies find no effect.

But you know what? Fuck it. Let’s round that 15% up all the way to 100%. Because it makes no fucking difference. Even if losing that additional police presence doubled the crime rate, the institutional or cultural differences between high-crime and low-crime areas cause overall crime rates to vary by ten times between the highest and lowest deciles.

Police are there to avenge the wrong and they exact retribution by filling out a report that goes to a bureaucrat who is angling for a political career but will happily engage in the usual prosecutorial abuse of massive overcharging that so he can plea your violator out without the bother of a trial, or maybe just throw it out like they did Trayvon Martin’s if that’s more politically expedient. The main defense against crime is self-defense.

167 Mark June 12, 2015 at 10:46 am

Alex, why didn’t you include Jan 2015 in your analysis? From what I could tell January had the a higher crime rate in Baltimore than Feb – April. Including January would provide a better picture of the variance in crime from month to month. Also, in the summer of 2013 there was a similar spike in crime to what we are currently seeing. To me the theory makes sense, but the data isn’t as clear as you make it out. Analysis like this is why economists get a bad rap.

168 Ray Lopez June 12, 2015 at 10:46 am

I find it ironic that AlexT advocated “kid’s gloves” approach to policing minorities, then complains when that happens, that the crime rate goes up. Speaking out of both sides of his mouth, like a true economist. Of course he might counter that his kid’s gloves were with respect to Ferguson, MI not Baltimore, MD, but that’s just weasel words to avoid being pinned down.

169 Sam Haysom June 12, 2015 at 11:10 am

You are being unfair Ray. Baltimore is a disaster now but its legacy as a onetime major city means that soi distant elites like Tarroback look at it through the same indulgent standards that they view New York, LA, Boston etc.

Bill Cosby had it totally wrong-low standards aren’t a form of bigotry they are a privilege in the same way that high standards are a weapon. The kind of excruciatingly double-guessing standards which Tarroback would subject the Ferguson PD to would never work for a city that AT might have to visit sometime, or in which Tarrobacks kids might one day live.

It’s like Sailer always points out it’s not an accident that William Braton gets to honest about crime in a way Sheriff Joe or the police chief of let’s say Memphis never could be. If William Bratton had tried to same tactics in New Orleans or Jacksonville hed be remember as some malignant Bull Connor clone.

170 JonFraz June 12, 2015 at 2:22 pm

Surely there is plenty of daylight between the police acting like the alpha urban gang and beating up random black men, and the police sitting on their hands doing nothing.

171 dearieme June 12, 2015 at 2:36 pm

In the abstract, yes. But for that particular police force, in that particular city, maybe not.

172 Ricardo June 12, 2015 at 10:53 am

What happens when you adjust for temperature? There is a well known positive correlation between crime and temperature, and the funeral was at the beginning of May…

173 adam June 12, 2015 at 11:00 am

Precisely. This post is a bunch of meaningless jibberish dressed up as a statistical analysis.

174 scott cunningham June 12, 2015 at 11:05 am

Leave it to Alex Tabarrok to show us what a blog can still do in this day and age. Fantastic analysis.

175 E. Harding June 12, 2015 at 11:13 am
176 Krigl June 12, 2015 at 1:29 pm

I’m not sure Far Left fits here, I’ve met hardcore lefties, who were otherwise composed and agreeable people, PZ Myers, outside his specialization, is just SJW asshole crusader.

177 E. Harding June 12, 2015 at 3:24 pm

I don’t care about tone; I care about content.

178 Yancey Ward June 12, 2015 at 11:38 am

How long before Global Warming is blamed for it?

179 d June 12, 2015 at 2:34 pm

yesterday

180 Cooper June 12, 2015 at 12:44 pm

There is seasonal variability to crime.

Show me the results for last year as well and then we can confirm a trend.

181 Rich Berger June 13, 2015 at 7:16 am

https://data.baltimorecity.gov/view/n29z-hvc9

MJ was decriminalized 10/1/2014.

182 Krigl June 12, 2015 at 1:21 pm
183 JonFraz June 12, 2015 at 2:19 pm

To what extent is this due to the weather? I live in Baltimore: our winter was extremely cold, and our early spring was very wet. That tends to keep the low-life types at home. It’s only been since very late April that’s it’s been really warm (and now finally hot) here. I think we need to wait until a year has passed and see if this crime wave lasts, or if it just turns out to be a case of deferred criminality.

184 Robert June 12, 2015 at 3:20 pm

It’s wrong to call this the “Freddie Gray” effect. This is the “Barack Obama” effect.

185 João June 12, 2015 at 3:52 pm

We’ve had several natural experiments concerning the relationship police x crime in Brazil.

Several states already had police strikes, which would not have been possible in the US, I guess.

In every single one of them there was a huge surge in the crime rate. Case closed.

186 Brenton June 12, 2015 at 6:41 pm

Joao, thank you. If only this didn’t get lost in the noise of speculation and bickering.

187 Robert (evidence based) June 12, 2015 at 4:50 pm

I agree / sympathize with almost everything Alex wrote in his post.

There is though, a glaring omission. The police cannot be the police if they are also criminals, and I mean that literally. I have seen this in multiple states with my own eyes and no-one seems to care. People seem to think that ff a police officer does it, then it’s not a crime. This is inline with Nixon’s ideas on the law ‘If the President Does It, That Means It’s Not Illegal’’.

The police frequently (not ocassionally) testilie. http://www.politico.com/magazine/story/2015/04/north-charleston-shooting-116871.html
They commit assault and then cover it up.
They taser people who are not violent.
And they murder people from time to time.
The list goes on and on.

This makes a mockery of the law and it undermines the legitimacy of the law in the eyes of the people who are victimized. You cannot expect them to obey the law if the representatives of the law do not obey they law. I dispute that this is a matter of ‘strained budgets’. Real spending on police has increased in the 80s, 90s and 00s until the recession.

We need to first have an agency whose job it is to police the police, and ONLY the police. Second, we need to vigorously prosecute violations of the law by the police. We need to hold them up to higher standards than civilians. If they beat a civilian with a baton without cause the penalty needs to be harsher. We need to train them so that they understand that force is something that is a last resort only. They need to understand that perjury will cost them their jobs, their freedom and their entire pensions.

Very few people not effected by police criminality care and they often have a fatalistic acceptance that goes along the lines of ‘that’s the way the police just are’. Of course they don’t have the same opinion about civilian criminals. And all of these protests refuse to accept it. So may I suggest that we first put the police house in order before we moralize about the criminality going on in various places like Baltimore.

And once done, I am all in favor of rigorous law enforcement in Baltimore, in accordance with the constitution and human rights.

188 TallDave June 12, 2015 at 5:35 pm

It’s hard to argue we don’t need better police. Sadly this is one of those cultural problems that’s really, really difficult to fix, often harder than getting people to engage political corruption. And perversely the areas with more criminals tend to have worse police.

189 Art Deco June 12, 2015 at 6:02 pm

.And they murder people from time to time.

There are hundreds of thousands of police officers in this country who have interactions with difficult people every day. The number of homicides by police is in the low four digits, and you have to scrounge to find examples of dubious conduct.

190 Art Deco June 12, 2015 at 7:13 pm

low three digits.

191 chuck martel June 12, 2015 at 8:20 pm

” you have to scrounge to find examples of dubious conduct.”

Not really. http://nailheadtom.blogspot.com/2013/07/bad-cop-david-clifford-will-serve-2.html

192 ibaien June 12, 2015 at 8:36 pm

https://pbs.twimg.com/media/B3QlOM8CMAE54Y9.png:large this discrepancy doesn’t bother you?

193 Robert (evidence based) June 12, 2015 at 9:26 pm

If you google around you will see there is plenty of evidence that many so called ‘justified killings’ are far from that. The sheer number of people killed by police, even taking into account the higher murder rate and great availability of guns provides strong evidence to a huge numbers of unjustified killings. Civilized people call this murder.

In the first 24 days of 2015, police in the US fatally shot more people than police did in England and Wales, combined, over the past 24 years.
Police in the US have shot and killed more people – in every week this year – than are reportedly shot and killed by German police in an entire year.
Police in the US fatally shot more people in one month this year than police in Australia officially reported during a span of 19 years.
Police in Canada average 25 fatal shooting a year. In California, a state just 10% more populous than Canada, police in 2015 have fatally shot nearly three times as many people in just five months.
Police fired 17 bullets at Antonio Zambrano-Montes, who was “armed” with a rock. That’s nearly three times what police in Finland are reported to have fired during all of 2013.

Yes, we have more people but American police kill 10 to 100 times more people per capita than the police in the EU, Australia, New Zealand, Japan, Taiwain, and South Korea.

So instead of just being outraged when a civilian commits a murder, we should reserve some of our outrage for the police.

http://www.theguardian.com/us-news/2015/jun/09/the-counted-police-killings-us-vs-other-countries

194 Robert (evidence based) June 12, 2015 at 9:54 pm

US police are 70 times more likely to kill a civilian than the police in England and Wales. The murder rate in the US is around 4.7 per 100,000. It is 1.0 in England and Wales. So you would expect the number to be 4.7 not 70. Even if the police in the US would be generally more fearful that someone has a gun, it does not justify a killing rate more than 70 times that of England. I would say anything higher than 10 times are you are probably looking at murder beyond that. However I would say we can be certain that anything beyond 20 times means, we can be certain the police are committing murder.

This means that we can be certain that 5/7 of all police killings in the US are unjustified, unnecessary and are murder. We need to change the law in the US so as to be able to prosecute the police when they commit murder.

195 TallDave June 12, 2015 at 10:23 pm

You don’t have to work very hard to convince me that US police are grossly overmilitarized in both training and equipment. On the other hand, there are certainly portions of the US comparable to England and Wales in both murder rate and rated of police-caused death.

The institutional deficits in other areas produce both the higher murder rates and the higher police-induced fatality rates. They also produce higher rates of theft, rape, dishonesty, rudeness, and abandoned single mothers. Even if we could fix just the police (we can’t) it wouldn’t really help much overall.

196 chuck martel June 12, 2015 at 6:35 pm

For starters, no law enforcement personnel should be allowed to consume alcohol or drugs at any time and they should be subject to 24/7 random testing by an independent agency. Individuals with 24 hour access to high-powered automobiles and firearms cannot be under the influence. If airline pilots and railroad engineers have to be sober on the job, why not cops with guns?

197 ohwilleke June 12, 2015 at 6:30 pm

Probably seasonal. The relationship between crime and daily high temperatures is one of the best documented relationships in social psychology.

198 jorod June 12, 2015 at 11:17 pm

The welfare state continues to rev up. Wait till they raise taxes and businesses start to shut down.

199 Rich Berger June 13, 2015 at 7:25 am

For all the bleating about “seasonality” and “show me more than 4 months”, I could see no one who looked up the source of AT’s graphs. I did, and here is arrests going back to the beginning of 2013 -https://data.baltimorecity.gov/view/n29z-hvc9

Pot was decriminalized in October 2014: you can see that. You can also see the sharp drop after FG’s death.

In other news -http://www.baltimoresun.com/news/maryland/crime/blog/bs-md-ci-gray-protective-order-20150603-story.html#page=1

200 Demosthenes June 13, 2015 at 12:06 pm

And likewise, regarding potential seasonality in crime rates, at the linked source (https://data.baltimorecity.gov/Public-Safety/BPD-Part-1-Victim-Based-Crime-Data/wsfq-mvij), once you manage to set all the filters correctly, it’s pretty clear that the seasonality between March and May is nowhere near big enough to account for this change.

Indeed over 2011-2014, the average monthly changes (increases) between March and April, and April and May, were 2.75 and 1.25 respectively. Whereas the respective month-on-month changes in 2015 are 6 and 21.

201 Bill June 13, 2015 at 12:25 pm

Demo, Hmm. Looks like greater use of firearms in the criminal activity. Second, what I also see from the data is that you are not using what Alex used, which was total arrests; you instead are using a subset of data which now excludes MJ arrests. Your database is victim based crime, not MJ, not resisting arrests, etc, which is included in Alex’s graph. Third, in addition to seasonality, I hope that you are adjusting for another change since their riots: first, damaged property from a riot becomes a place where there is additional looting, and second, people report theft from property often sometimes later, after the riot or property damage occurred. So, just taking an event–an riot–and using that as a cutoff can be misleading, if theft is reported later, or is attributable to damaged property which is later looted. What you could do is overlay zip codes of riot locations and see what this shows. Third, perhaps MJ decriminalization leads to greater gang warfare as MJ demand increases and gangs vie for territory. Worth checking the link. Finally, there is no evidence that fewer arrests leads to more crime.

202 Bill June 13, 2015 at 12:38 pm

Demo and Rich, I hope you also understand that data over a two month period is probably not statistically significant for this time series. Try flipping a coin and seeing the frequency of two heads occurring in a row, which is the equivalent of two months data here. In looking at Alex’s graph, note the dispersion and variance as well.

203 Rich Berger June 13, 2015 at 2:42 pm

You can wiggle all you want, but AT’s point is a little too obvious for you to evade. The truth will set you free, if you just open your mind to it!

204 Bill June 13, 2015 at 3:04 pm

Rich, I’m not wiggling at all. You can’t take two observations–May and June–and make a time series conclusion about a change.

Take out a coin and flip it. If you get two successive heads, you are telling me the coin is a two headed coin.

205 Demosthenes June 13, 2015 at 9:39 pm

Sorry I wasn’t clear above (tired, was late at night where I was, and failed to proof read properly).

The numbers I quoted are for homicides. Since Rich had addressed potential seasonality in arrests (Alex’s first graph), I looked into seasonality in murders. So nothing immediately to do with arrest rates, although I apologise for not being clearer. As it turns out, there’s a little bit of seasonality. But nowhere near enough to explain this year’s spike.

Regarding significance, we’re talking about a monthly count series with the vast majority of instances being between 10 and 20. And then we get a jump to a record high,over twice the historical average, overnight. The monthly average over 2011-2014 was 14, with a standard deviation of 4.5. Looking at monthly changes, the average over the same period is negligable (no overall trend to speak of), with a standard deviation of 5.6. The jump in the course of two months from 12 to 39 is highly significant by either metric.

And it’s not really only 1 or 2 datapoints – most of the events that contribute to the count are relatively independent (i.e. we’re mostly talking about events that contribute 1-2 homicides at a time, rather than a mass shooting where the deaths are highly interdependent). So we’re talking about a series that looks something like a Poisson, where a spike of this magnitude is definitely significant. Alternatively, like Alex does, the daily homicide rate can be used to increase the sample size, and it’s pretty clear the recent change is significant.

206 Bill June 13, 2015 at 11:42 pm

Demo, It is still two months. It is two data points! Two data points! Here is a link on hypothesis testing using time series
https://www.cs.berkeley.edu/~jordan/courses/294-fall09/lectures/time/slides.pdf Poisson is irrelevant if you have only two data points on which to test your hypothesis.

I just flipped a coin two times in a row. Both times it came up heads. Must have a two headed coin. It’s a normal distribution, and I just got two heads in a row.

207 Demosthenes June 14, 2015 at 5:51 am

Yes, when constructed as monthly data, it is just comparing two data points. From a distribution about which we know quite a lot, and in particularly have a pretty good idea about the mean (~0) and standard deviation of monthly movements (~5.6). We see a monthly increase of 21. When testing the null hypothesis that the May outcome is produced under the same DGP as the prior series (this is essentially a difference in means test with equal variances i.e. the joint test of whether either the mean or variance has changed), the t-stat is 3.75. Sure, this sort of thing happens by random chance every so often, well under 1% of the time, which is rather consistent with the fact that the number of homocides in Baltimore in May is apparently higher than any monthly outcome in several decades.

I don’t think you understand the Poisson claim. There are good a-priori reasons to believe the series should be roughly Poisson (count data of fairly independent events). Not perfectly Poisson distributed, because homicides aren’t completely independent, and indeed the overall variance is a little higher than the mean, whereas in a true Poisson they should be the same. But it’s a good rough guide. Based on the historic data (2011-2014), we can make a guess at the parameters of this distribution – the monthly averages for March through July are 12.25, 15, 16.25, 16.5 and 16. The variance over the entire period is 20, so to be extremely charitable, to maximise the chance of large positive outliers, let’s consider a Poisson distribution with lambda = 20. Then, taking that distribution (i.e. assuming there has been no structural break), what is the probability of observing a month with 39 or more homicides? Again, a tiny fraction of a percent.

208 Bill June 14, 2015 at 7:55 pm

Demo, Your may data includes the riot period. See below.

209 bill June 13, 2015 at 12:42 pm

I’d like to see data that starts at least 12 months ago. Doesn’t crime rise in the summer every year?

210 Demosthenes June 13, 2015 at 10:24 pm

Lots of people here are hand-waving about seasonality. Of course this is a valid concern, and is good to check – the difference in differences is a much better measure. Fortunately Alex provided us with a link to the data, so we can check this. Take the homicide rate for example (https://data.baltimorecity.gov/Public-Safety/BPD-Part-1-Victim-Based-Crime-Data/wsfq-mvij, select the filters for homicide, then collapse by day or month). It’s pretty clear from the data that the seasonality between March and May is nowhere near big enough to account for this change.

Indeed over 2011-2014, the average monthly changes (increases) between March and April, and April and May, were 2.75 and 1.25 respectively. Whereas the respective month-on-month changes in homicides in 2015 are 6 and 21. So there’s a little bit of seasonality in the data, but nowhere near enough to explain this year’s spike.

Indeed, the monthly average over 2011-2014 was 14, with a standard deviation of 4.5. Looking at monthly changes, the average over the same period is negligable (no overall trend to speak of), with a standard deviation of 5.6. The jump in the course of two months from 12 to 39 is highly significant by either metric.

But can this really be significant, given they are just 1 or 2 data points, right? Well, most of the events that contribute to the count are relatively independent (i.e. we’re mostly talking about events that contribute 1-2 homicides at a time, rather than a mass shooting where the deaths are highly interdependent), so in terms of independent observations the data is a lot richer. Like Alex does, the daily homicide rate can be used to increase the sample size, and I suspect the effect would be highly significant here as well.

211 Bill June 13, 2015 at 11:47 pm

Still two data points. See response above re hypothesis testing and time series. And, you can’t say “I suspect etc.” to be the source of a conclusion. it doesn’t work that way. You test, you don’t suspect, and you dont frame a suspicion as a conclusion.

212 Demosthenes June 14, 2015 at 6:21 am

Yes, when constructed as monthly data, it is just comparing two data points. From a distribution about which we know quite a lot, and in particularly have a pretty good idea about the mean (~0) and standard deviation of monthly movements (~5.6). We see a monthly increase of 21. When testing the null hypothesis that the May outcome is produced under the same DGP as the prior series (this is essentially a difference in means test with equal variances i.e. the joint test of whether either the mean or variance has changed), the t-stat is 3.75. Sure, this sort of thing happens by random chance every so often, well under 1% of the time, which is rather consistent with the fact that the number of homocides in Baltimore in May is apparently higher than any monthly outcome in several decades.

You’re right – I shouldn’t use “I suspect” to be the source of a conclusion. Which is why I didn’t. It was a suspicion, a.k.a. I hadn’t had a chance to run regressions on the monthly data yet, so was merely noting my suspicions beforehand. Nothing wrong with that.

But at least I’m actually checking the data, rather unlike quite some number of people above who have some suspicion about seasonality, and essentially dismiss Alex’s findings without even stopping to see if there contary suspicion actually invalidates his argument.

So what does the daily data reveal? Controlling for month (for seasonality) and year, and allowing for a few different serial correlation structures in the error term, the coefficient for the period after the riots ranges between 0.75 – 0.79, which is highly significant in each specification I ran (t-stats of 5.94 using HAC standard errors, and 5.84 when controlling directly for autocorrelation). So the increase in daily homicides since the riots is very real, and controls change Alex’s story (he quotes 0.82 above) very little.

213 Bill June 14, 2015 at 8:07 am

Demothenes, Sorry, but you don’t fool me. 1. What is the period you use for monthly data. How many sample points 2. Was it controlled for seasonality; if so, you have a smaller set of comparable (you would have May through whatever period you chose for seasonal comparables, but no matter what period you choose (May to Sept), you would still have a small set. Or did you make the adjustment over all periods and again what was the period you chose. 3. What was the seasonal adjustment factor you used, or did you. If you didn’t adjust for seasonality, your data simply is showing seasonality, not a step change. 4. For your monthly increase of 21 (by the way, what is the monthly increase of..property crimes, assaults, arrests (all crimes), etc. apparently you did you exclude the riot period from the sample as an outlier. 5. What is the 21 number you refer to. What is it 21 of??? 6. How did you control for an assault which also included a property crime, so that you would have arrest number for assault increase when there are more property crimes. 7. What was your cutoff (or did you cutoff) the riot period. Did you exclude the riot period as an outlier, and if not, which data set did you include it in. 8. What are the variables to which you attach the coefficients (total arrests (and if so, did you adjust for MJ enforcement).

214 Bill June 14, 2015 at 8:18 am

Just eyeballing it, it looks to me like you included the riot period in your data set to reach the conclusion that there has been a step change in criminal activity for the period may/june. Gee whiz. Same thing Alex did in the way he drew his graphs. What d’ya get if you exclude the riot period. By the way you have to exclude the riot period if you are testing for a change; otherwise, all you are saying is that during a month in which there is a riot, there is more property crime. You cannot claim that following a riot period, using a month which included the riot, that there has been a step change in criminal activity, no matter how you define it.

Did you exclude the riot period.

215 Demosthenes June 14, 2015 at 10:40 am

Funny, you won’t believe a word I say when I run the stats on the monthly data, you don’t believe a word I say when I run the stats on the daily data, yet you expect me to believe some claim you make from eyeballing. Sheesh.

The regressions use daily homicide data from 2011 onwards. In case you missed that (since you missed it each time above, despite all these characteristics being mentioned above), the regression uses the number of HOMICIDES observed DAILY in the period 2011 ONWARDS.

Dummy variables for MONTH capture seasonality. Dummy variables for YEAR capture any trends across years. Data on neither arrest rates nor other crime categories enter my model, because as stated above, I looked at HOMICIDES.

It’s far from clear why the riot period should be excluded because I’m testing for a change in the HOMICIDE rate, rather than looking at property crimes and the like. But if I change the definitions to exclude the riot period, the point estimates for the post-riots period actually get a little (negligibly) larger.

216 Bill June 14, 2015 at 11:39 am

Demo, I dont take anything on faith, I dont believe a word anyone says, but I do appreciate you identifying the homicide variable, and I do believe you should take out the riot period as an outlier. So, if you take out the riot period, reducing sample size at the end tale, you claim that the model becomes stronger.

So I take it you excluded the period April 18 to May 4, which is the period in question based on the Wiki description of the riot period: See, Wiki for dates of actions and lifting of curfew:

https://en.wikipedia.org/wiki/2015_Baltimore_protests

What is the date range you excluded. You are saying that you have sufficient adjusted data from May 4 (lifting of the curfew and ostensibly the end of the riot period) to conclude that from May 4 to early June that Baltimore has now entered a new period of higher homicide rates based on the observation of May 4 to early June.

217 Bill June 14, 2015 at 11:43 am

Oh, and Demo, what is the population of Baltimore, and what is the homicide rate as a percentage of total population. How significant is the percentage difference in the number of homicides per 100 thousand from the period May 4 to early June?

218 Bill June 14, 2015 at 12:00 pm

Demo, You might also want to look at the murder rate charts based on DOJ crime data here:
http://www.thedailybeast.com/articles/2015/01/03/america-s-2014-murder-capital.html You will see the rates per 100,000 had been declining. Am waiting for the period you excluded from the analysis as the riot period, and whether you can claim there is sufficient data from May 4 to early June that there is a new phase of Baltimore murder rates.

Too small a sample. Go ahead, keep trying.

219 Demosthenes June 14, 2015 at 12:05 pm

Excluding the riot period, the point estimates became negligibly stronger, yes. The t-stats declined a little, which is not surprising given the sample for the post-riot period became smaller due to the redefinition (I had been using Apr 27 onwards as the sample for the structural break, given that is when the funeral was / seems to have been the key day / was the date Alex used, and changed it to May 4 onwards since that is the first day the curfews had finished, national guard had left etc). But it all remained significant etc, there’s still about 30 days in the post-window and that’s enough to get fairly small standard errors.

Google tells me that the Baltimore population is 622k (either way, it won’t have changed much between April and June), and the homicide rate in the post riots period is about 1.35 per day in this period (compared to about 0.5 per day before). So the homicide rate goes up by approx 170%. In terms of homicides per 100k per year (annualising those rates), that’s a jump from 29 to 79, if my math is correct.

220 Demosthenes June 14, 2015 at 12:09 pm

The regressions I ran control for year, so any long term trends like the one you speak of are captured.

And yes, the magnitude of the increase in homicides is well and truly sufficient for the effect to be significant based off the available daily data.

221 Bill June 14, 2015 at 4:54 pm

Demo, Wait a second, you are doing a day by day estimate of a murder rate?? There has to be a lot of variation within a month, and there is, if you look at the data on either side of the period before the riots. So that is the point estimate…not a month, but days from May 4 to early June and you are getting statistical significance for each day against other days, controlling for seasonality and evidently a trend you detected that apparently was going lower, rather than reverting to a mean. Earlier, you talked of monthly 0 standard deviations, not day variations in your model and you represented you were using a monthly comparison. (Go back to your statement above which begins: “Yes, when constructed as monthly data….” )

I don’t think you can construct a daily murder rate, particularly when there is so much variation within the month.

222 Demosthenes June 14, 2015 at 11:32 pm

“I don’t think you can construct a daily murder rate, particularly when there is so much variation within the month.”

TIL: And you’re wrong. Congratulations.

223 Bill June 14, 2015 at 8:07 pm

Demo, You might want to think about statistical significance and validity of your day by day model when the model says that there is a statistically significant 170% increase in the murder rate. Just sayin’

If you look at homicide data collected by the Baltimore Sun, I don’t think you see that (and, even that data is limited by a small date range for homicides after the riots). Here is the link: http://data.baltimoresun.com/bing-maps/homicides/?

When you look into the specifics of the homicides, you see how many involve 1, 2 and 7, 16 and 17 year olds, how many involve people who live at the same address, how many involve firearms, how many victims are black, etc. You also get a sense that homicide rates had been declining over last year and began increasing before the riot, returning to the mean of previous years.

224 Bill June 14, 2015 at 8:37 pm

Demo, regarding your use of Poisson distribution and small samples (daily crime rates from May 4) consider whether the assumptions for applying this methodology are met; specifically, absence of dependence, and offenders being influenced by others behavior, in determining whether there has been a change in the rate:

“Residual variance will also be greater than λi if the assumption of inde- pendence among individual crime events is inaccurate. Dependence will arise if the occurrence of one offense generates a short-term increase in the probability of another occurring. For aggregate crime data, there are many potential sources of dependence, such as an individual offending at a high rate over a brief period until being incarcerated, multiple offenders being arrested for the same incident, and offenders being influenced by one another’s behavior. These types of dependence would increase the year-to- year variability in crime rates for a community beyond λ i , even if the under- lying crime rate were constant.

For these two reasons, ‘‘overdispersion’’ in which residual variance exceeds λi is ubiquitous in analyses of crime data. Applying the basic Pois- son regression model to such data can produce a substantial underestim- ation of standard errors of the β’s, which in turn leads to highly misleading significance tests. ”

Source: http://www.udel.edu/soc/faculty/parker/SOCI836_S08_files/Osgood_JQC00.pdf

Also consider the small set of samples and your models projection of a 170% increase in the homicide rate. It is probably telling you something when the model says it is significant and the change is 170% over the base rate.

225 Bill June 14, 2015 at 9:39 pm

Forgot to tell the title of the article. The quotes above are from an article entitled “Poisson-Based Regression Analysis of Aggregate Crime Rates” by D. Wayne Osgood

226 Bill June 14, 2015 at 9:49 pm

Also, regarding the dependence issue and the relationship of the riot to murders in a subsequent period after the riot, here is one comment from the Baltimore Sun:

Some Baltimore officials have speculated that the spike in violence can be partly attributed to the looting of drugs from pharmacies during riots that followed the death of Freddie Gray.

During April’s riots, thousands of dollars in narcotics and other drugs were stolen from more than 20 Baltimore pharmacies. Twenty-seven pharmacies reported damage during the unrest.

The U.S. Drug Enforcement Administration and police are investigating the thefts, but neither knows how many opiates were taken. Investigators have yet to interview all the business owners.” Here is the link: http://www.baltimoresun.com/news/maryland/crime/blog/bs-md-ci-shootings-20150531-story.html

227 Demosthenes June 14, 2015 at 11:28 pm

Maybe, just maybe, this says something essentially identical to the caveats I provided when talking about testing using the Poisson distribution, and maybe, just maybe, this is why I used numbers that were extremely conservative.

And the regressions don’t use a Poisson distribution, or anything like that. They are completely different tests.

“It is probably telling you something when the model says it is significant and the change is 170% over the base rate.”

Yeah, it is. Generally that there has been a large, significant increase.

228 Bill June 15, 2015 at 6:05 am

Choosing “conservative numbers” doesn’t correct for the problem if there is dependence and small sample sizes.

229 Arthur B. June 15, 2015 at 8:12 am

Regarding your theory paper, what about the elasticity of police resources? Why doesn’t the police allocate more resources to high crime areas – or deallocate resources from low crime area?

230 TheBride July 1, 2015 at 10:47 pm

The people of Baltimore got what they wanted: Little law enforcement and a free hand to steal, loot, kill, and run wild. And–on top of that–they get the bonus of property values plummeting, good people moving away, and a community left to nincompoops. Let Jesse Jackson and Al Sharpton lead you to another decade of disaster. You all deserve it.

Comments on this entry are closed.

Previous post:

Next post: