About . . . . . . Classes . . . . . . Books . . . . . . Vita . . . . . . . Links. . . . . . Blog

by Peter Moskos

February 5, 2017

The Curious Case of Poverty and Crime

When I'm charming people at a cocktail parties with talk of rising crime and the role of police, the good people I talk to, rather than even considering the possibility that police matter and post-Ferguson protests might matter (in a negative way), inevitably try and shift the discussion to greater social issues: poverty, racism, and inequality, the so called "root cause" of crime.

The "root causes" position has long annoyed me. I care about those poverty, racism, and inequality, but in terms of effective crime-preventing policing today, the "root causes" are nothing but a distraction. It's basically a defeatist way to say we can't lower crime until we fix society. I'm all for fixing society, but I'm not willing to hold my breath till it happens. Also, the idea that the only way to impact crime is to address structural issues is consistently and demonstrably false.

Last year poverty went down and murder went up. In 2008, the economy tanked, and criminals barely noticed. Between 1965 and 1975, poverty is the US was way down; violent crime way up. In the 1990s, during New York's great crime decline, the number of New Yorkers living in poverty increased 21 percent. Inflation adjusted household and family income declined. Unemployment approached 10 percent.

[A few academics buck the poverty-causes-crime trend (Orlando Patterson and Marcus Felson are two that jump to mind), but despite all the evidence to the contrary, "poverty causes crime" is still pretty much accepted as scripture.]

Anyway, cause I'm not much of a football fan (or Satan fan), I thought I'd graph poverty and homicide over time. And here's what you get:

[click to embiggen]

Those lines are almost flip visions of each other (Especially if one ignores the 1990s.) Turns out, at least since 1959, there an inverse correlation between poverty and homicide in the US. Homicide goes up when poverty decreases. Statically significant and everything. Well, that's awkward.

[Update: A commenter makes a very good point that I'm overstating any statistical significance because of the high poverty years up to 1995.]



Does this mean we shouldn't reduce poverty because homicide will go up? Of course not. (I often make fun of the "correlation doesn't equal causation" mantra -- because sometimes correlation does indicate causation, and correlation certainly doesn't eliminate possible causation -- but the "correlation doesn't equal causation" mantra is well worth repeating here.) I think homicide is far less linked to macro economics than, well, macro economists would have you believe.

But this fact remains: there is an inverse correlation between poverty and homicide. [Correction: Eh... maybe, maybe not. Probably not. See comments section.] The question then is to figure out how and why and through what intervening variables. I'll leave that for better statisticians than me to figure out. But let's assume, just for a moment, that this correlation isn't random. Why might this be so?

Hell, I don't know. But if I had to hazard a guess... I'd think that perhaps some of same good policies that help reduce poverty and suffering in our country might go along with a certain ideology that occasionally has its head up its ass when it comes to policing and crime. Conveniently this might also explain the 1990s, when both poverty and crime decreased. President Clinton and Vice President Gore managed to reduce poverty while still being firmly on the police side of the ideological divide. Broken Windows was working in NYC. Welfare was being ended as we knew it. And the Feds even coughed up a chunk of change for a few more cops, to boot.

15 comments:

Unknown said...

What about other kinds of crime? Crimes like theft or assault- is there any connection with poverty there?

Otis Blue said...

@Unknown. Take a look a few posts below, specifically "No, it's not just Chicago" for some insight on your question.

Jim from BC said...

You know it's a sidenote but until I saw this chart I didn't realize how successful Johnson was at reducing poverty.

aNanyMouse said...

Typo alert: I'll bet you wanted to say "I only kinda sorta of Knew it."

Maybe this correlation owes something to criminals seeing more (easily) steal-able stuff dangling in apartment windows in (once) poor Hoods. Or to resentment that most of the new gravy is going to certain best-connected people.

aNanyMouse said...

More on typos:
1) You may as well say "kinda sorta knew it."
2) "...theft might go UP when money is flush."

On Baltimore, it could also be that bad guys know that witnesses will sing less on them, esp. now that cops are going fetal, incl. on helping witnesses.


Peter Moskos said...

Man... Problem is comments can't be corrected. I'll re-post it. Thanks.

Peter Moskos said...

I only kinda sorta knew it. But I was a bit surprised, too. I knew there was a basic truth to the much disparaged "the 'Great Society actually did reduce poverty." But I didn't know the numbers.

Peter Moskos said...

Thieves need something to steal. So it's not crazy to think that theft might go up when money is flush. And addicts might buy more drugs, leading to more drug market violence. The problem is you could make the counter assertions just as easily (money tight, more drug violence). We don't know.

What I think the chart should make one realize is that if one wishes to reduce murder, a federal focus on reducing poverty overall won't do it. (Not to say reducing poverty isn't good without further justification.)

The economy at the block level certainly matters. But even that isn't the be all and end all. At least no short term. The violence increase in Baltimore after the riot had nothing to do with changes in the local or national economy. The only variables that changed had to do with policing: less low-level enforcement and fewer interactions with violent criminals.

[But still, I can't get over the seeming (even if indirect) *inverse* relationship between national poverty and crime (and the 1990s exception). Something else is going on here, too...]

Peter Moskos said...

As to people talking less. That certainly matters. But I think the bigger (related) problem is failed prosecution. Worst thing you can do is snitch and then *not* have the guy put away!

It's probably less that people *think* they can get away with murder (most criminals always *think* they can get away with it) than the fact that more shooters are walking around shooting. Unless they get deterred by police, incarcerated, find God, or age out: shooters gonna shoot.

Donna Richardson said...

If you look at the graph that you include but barely mention in the blog, for poverty rates below about 15% (the left side of that graph), the dots appear to be nothing but noise. The entire correlation (and an r squared of 0.2 indicates a moderate relationship at best) seems to be caused by the 7 dots in the lower right quarter of the graph. These points must correspond to 1959-1965, as those were the only years when poverty levels were that high. I challenge you to try this same exercise using only 1965 until the present. If your thesis is real, then the correlation should still be there when you remove those 7 years from the analysis. Actually, you can still argue strongly that crime and poverty are not related; you just can't claim the inverse relationship anymore if the correlation goes away.

Peter Moskos said...

Thanks very good point! And removing those data does eliminate the statistical significance. (Though the non-significant correlation / trendline still goes in the same direction)

And indeed, as observe, my main point is that these two variables are not related. I don't think increasing poverty decreases murder or vice versa. And I'm quite happy to go with no correlation at all. But...

[But statistically, I'm not convinced these data should be removed; they're not exactly "noise" since they correlate with low homicide years. But given your on-the-mark observation, I do assume you're a better statistician than me. So please explain if this makes sense: Since those early years are grouped together chronologically, wouldn't it be better to consider them an extreme, but not as outliers to be excluded? Would it not make sense to include them, but reign them in a bit? For instance, if you consider the poverty rate from 1959 to 1966 to be constant, at the level of 1966: 17 percent, then we still do (just) get a statistical significance (F = 2.4, sig = .044).]

Peter Moskos said...

Here's the data if you want to play with it. Not certain the best way to share in the comment section of a blog. Columns 1-4 are year. 5-8 = poverty. 9-11 = homicide. 1980 homicide rate is 10.2. I replaced it with *** because 10.2 is 4 digits and would screw up the column format.

195922.44.9196022.25.1196121.94.8196221.04.6196319.54.6196419.04.9196517.35.1196614.75.6196714.26.2196812.86.9196912.17.3197012.67.9197112.58.6197211.99.0197311.19.4197411.29.8197512.39.6197611.88.8197711.68.8197811.49.0197911.79.7198013.0***198114.09.8198215.09.1198315.28.3198414.47.9198514.07.9198613.68.6198713.48.3198813.08.4198912.88.7199013.59.4199114.29.8199214.89.3199315.19.5199414.59.0199513.88.2199613.77.4199713.36.8199812.76.3199911.95.7200011.35.5200111.75.6200212.15.6200312.55.7200412.75.5200512.65.6200612.35.8200712.55.7200813.25.4200914.35.0201015.14.8201115.04.7201215.04.7201314.84.5201414.84.4201513.54.9

Joseph Dundee said...

But the idea that the poverty and homicide are unrelated is ridiculous on its face. It is abundantly clear the high murder rates are invariably in poor areas, and very wealthy areas invariably have low murder rates. Poverty may not be a sole determining factor but it is a key piece of the puzzle. Of course causation is hard to prove here, and I agree with you that entrenched structural problems like poverty don't mean we shouldn't look for more near-term solutions like improved policing, but to suggest that poverty doesn't matter is a real stretch.

Peter Moskos said...

Related. A bit. But I think poverty is a red herring in terms of violence reduction. We reduced crime in NYC in the 1990s without reducing poverty; we reduced poverty in America in 2015 without reducing crime. Other factors (eg: culture, having a working parent or two, nice neighbors, a lead-free home) matter in a way that poverty does not. And though poverty *can* serve as a proxy measure of these harder to quantifiable variables, but it's still only a proxy measure. Poverty *by itself* is not a good predictor of violence. Not on the individual level. Not on the national level. And certainly not on the international level.

High murder rates are seen in poor areas, but not all poor areas have high murder rates. Poor immigrants, no matter how, collectively refrain from homicide. Haitians in Florida are an interesting case study. So is East Liverpool, Ohio (which I wrote about here.)

Let's focus on reducing poverty, by all means. Just don't think of it a violence prevention program. Violence is not an inevitable result of poverty. (Unlike say, public illegal drug dealing and guns, which do pretty cause violence.) Talk of poverty and other "root causes" is too often a distraction technique to avoid nitty-gritty talk about policing. When it comes to violent crime, we know what has worked. And we need to stay focused on criminals and policing (and ending drug prohibition). That really is what I'm getting at.

Also in terms of direct cause and effect, the poverty-to-homicide link is very inefficient. There were 42-some million Americans in poverty last year. And (just?) some 15,000 murderers. Even if one were to assume that every murderer is in poverty, we're still only talking perhaps a 3,000:1 ratio. And then we know from 2015 that even reducing poverty by 3 million didn't even correlate with (much less cause) a reduction in homicides. Would 4 million have done it? I doubt it. What about 40 million? I don’t know. I suspect most violent offenders would be holding out till the bitter end.

DWPittelli said...

Joseph Dundee: "the idea that the poverty and homicide are unrelated is ridiculous on its face. It is abundantly clear the high murder rates are invariably in poor areas, and very wealthy areas invariably have low murder rates."

Violent criminals tend to be poor and live in poor areas, both because in the long run such crime doesn't pay, and because residents of middle class and rich areas would not tolerate their presence. But that doesn't logically require that growing the economy or lowering the poverty rate will make them no longer be violent criminals. At any rate, the changing poverty and murder rates over time do not support the notion that poverty causes murder in this way.