At the Atlantic, sociology prof Philip Cohen provides this graph:
Looking at it from the perspective of 1990, it was easy to assume a strong causal relationship between the rise in single motherhood and the murder epidemic. By my reading of the research, it is true that children of single mothers are more likely to commit crimes. But other factors are more important.
Contrary to Mitt Romney, it seems unlikely that increasing marriage is, in the current situation, the best way to reduce violent crime.
I wonder what relationships you would see if you broke the “single mother” graph down by adding additional factors. Say by income, or by divorced (and had the child/children prior to the divorce) vs. never married.
I would say that more data would be valuable (per RonF).
How about the SAME data broken out with families where the head of household is a single dad.
I agree with RonF and ASH, but a more important adjustment would be to change the right side/blue line part of the graph from ‘violent crime per 100,000 population’ to ‘violent crime per 100,000 people age 40 and under.’ People commit much less crime as they become older adults, and the period of declining overall crime rates is strongly correlated with the ‘aging out’ of the baby boomers.
There also needs to be an age adjustment to the left side/red line as well. The percentage of households comprised of adult female seniors living alone is obviously not germane to the question of whether the phenomenon of single mothers correlates to their children being more likely to commit crime. After 2000, a growing portion of ‘female only’ households will be seniors, and if those haven’t been removed from the dataset the diverging direction of the red and blue lines in this century in the OP graph is misleading.
Once you make these adjustments, I suspect the correlation (between single motherhood and crime) will be significantly restored.
I should add here that I don’t advocate the promotion of marriage as the top priority in trying to reduce crime; my ‘prescription’ would be more towards job rights, universal health care, stronger public schools, and other egalitarian measures. However, I do think it’s wrong to ignore how the peripheralization of fathers from their families may potentially be linked to more disaffected youth.
Post-hoc data massage is not generally a route to truth. Or a good argument.
I clicked through to the original data. They’re only counting “family households”. Depending on which piece of data they’re using (I can’t quite figure that out), it’s possible that it might end up counting several adult family members living together, in which the person who owns the house is a single woman, as a single-mother family, but those are probably rare enough to not affect the data too much. (So, say, an unmarried woman whose elderly parents move in with her might be counted as a single woman who is the head of a family, but a single woman living alone or with people who aren’t related to her doesn’t get counted as a family household at all.)
Post-hoc data correction is only necessary when the ‘pre-hoc’ data set is sloppily and potentially misleadingly selected, as appears to be the case here, Ben. The need for the corrections I suggest will be self-evident to anyone with a basic understanding of American demographics (and statistics). I suspect even you realize this; I notice you don’t actually make an argument against what I’m suggesting.
Ballgame, as Ruchama says, “adult female seniors living alone” are simply not included in the graph you’re criticizing.
I had the same thought initially, but I decided that it wouldn’t actually make a large difference. Here’s a graph from the Bureau of Justice Statistics:
As you see, even if we look only within young age groups, there’s been a huge decline in violent crime per capita. (For homicide, at least.)
ballgame, as I pointed out in my comment 5, your point about the single women data is flat-out wrong. As for the age of criminals, an aging population can’t possibly explain the huge drop in violent crime rates starting around 1990. That chart shows that the number of violent crimes per 100,000 people was nearly cut in half between 1990 and 2010. The percentage of the population that’s under 40 did not decrease by anything approaching that amount.
I’d appreciate it if both of you would dial it back a notch or two. Thank you.
even if we look only within young age groups
I feel like you’re understating the case here.
The entire drop in violent crime is young age groups, 14-24. Among the 25+, violent crime has been (negligibly) increasing since 2000.
Pfff… sociologists. Cheap video games are responsible for the crime rate drop. Everyone can see that.
If you wanted a paper, a grant, even just matching curves, you have to plot “amount of video game playtime household income can buy”.
If only people listened to Engineers…
OK, so I did a quick check of some of the figures, and it turns out I was right: a significant amount of the drop in crime rate in the first graph was due to aging of the population. The original chart shows a significant drop (~15%) in violent crime rate from 1980 to 2000; in fact, the violent crime rate per 100,000 18-34 year olds was almost exactly the same during those two years.
Amp, your chart is interesting, but the crime of homicide does have an issue that may be affecting these figures. It’s not clear to me how much the homicide rate is affected by advances in medical technology — i.e. if the victim doesn’t die, how much does that affect the likelihood the assailant will be charged with attempted homicide, or do the figures the chart is based on even include attempted homicide? I tried to dig up overall violent crime by age on the crime websites but it doesn’t appear to be available.
Describing that as “right” seems like a serious stretch. As in, no, you were wrong. Taking age into account has some effect, but in no way does the drop off after ’95 in crime rates go away once you look at crime rates for 14-24 year olds. But there is no equivalent drop off in single motherhood, so the correlation remains very bad.
Tell us, please, what is the correlation between single motherhood and the crime rate for 14-24 year olds, and why you consider that a significant correlation.
My sources were the same as everyone else here, Ruchama. I calculated the age by population using census data and crime rate using UCR data:
1980 18-34 yrs: 67,104,046
Violent crime: 1,344,520
Rate per 100k: 2,004
2000 18-34 yrs: 72,291,981
Violent crime: 1,425,486
Rate per 100k: 1,972
The original figs (doing my best to read the graph):
1980 crime rate per 100k: ~600
2000 crime rate per 100k: ~505
I don’t know what the correlation is for that age group because I can’t find the violent crime figures for that age group, Charles S. The only thing I see are the homicide rates for that group as indicated in Amp’s chart, which — as noted — is potentially misleading if it fails to include attempted homicides. The decline may partially be due to advances in medical technology saving more lives of victims who would otherwise die from their wounds (and therefore no homicide will have occurred, just an attempted homicide). You apparently know the violent crime rates for this age group somehow, so why don’t you tell me?
Ballgame, you are now saying that the crime rate when adjusted for demographics remained flat, at a time when single motherhood rose sharply. I’m not entirely convinced by your seat-of-the-pants equations, but even if you’re right, I don’t see how that supports your claim that “the correlation (between single motherhood and crime) will be significantly restored.” Whether you go by the figures you use, or the figures Phil Cohen used, the correlation simply isn’t there.
You could say that flat vs. increasing is closer to a correlation than falling vs. increasing, and in a way you’re right. But in neither case is there a correlation.
(It’s like saying that a baseball player who hits 5 home runs in a year is closer to being the new Babe Ruth than someone who hits 1 a year. Yes, that’s technically true, but it in no way contradicts Phil Cohen’s claim that neither player is a home run king.)
Also, I would appreciate it if you’d address your claim that the chart is “grossly misleading” because of “adult female seniors living alone.” Is that something that you still stand by, or have you been persuaded it was an error? (Nothing wrong with that, of course.)
I took two data points to verify that the chart you re-posted from The Atlantic omits an important factor — namely the graying of the population — and therefore misleadingly exaggerates the reduction in crime rate in recent years. I was correct in my suspicions; however I would not use two data points to measure the overall correlation between single female households and crime rate. I would want to take a number of data points to measure that correlation, and frankly I don’t have the time to do that at the moment. If and when I do have the time to do that, I’ll likely post more complete thoughts about this issue over at FC.
Not an error, because that wasn’t my claim. I never used the phrase “grossly misleading.” My exact words were:
I stand by what I said. First of all, the chart IS misleading because of the failure to compensate for the effect of an aging population on the violent crime rate, as noted earlier. But the chart would have been more misleading if — and please note that I used the word “if” here — senior adult woman living alone had not been removed from the dataset. I’m satisfied that senior adult women living alone aren’t part of the dataset because they don’t fall under the census definition of “family,” so that’s good news.
However, I’m not entirely convinced that senior adult women are completely a non-factor here. It has been noted in many places that adult children are finding it more difficult to afford to move away from home in recent years, with significant numbers of post-college progeny ‘returning to the nest’ until they can land a job that would allow them to afford to live independently. AFAICT, if those adult children are returning to a ‘mother only’ family home, they would be a part of the red line dataset and would be inflating the numbers of ‘single women families’ for recent years.
Now, I have no idea how big this particular demographic would be. I imagine it’s probably not huge, but it’s possible that it’s making a small but significant contribution to the divergence of the blue and red lines in the most recent time period.
Apologies for inserting the word “grossly” – it was an accidental misreading.
So, if homicide rates are effected by technology, wouldn’t that effect all age groups similarly? Why do we see a (modest) increase amongst 25-34 year olds and an enormous plunge among 14-24 year olds? Is it that 25-34 year olds are more likely to be trained assassins or something?
I claim no expertise in this subject, and I’m not going to look up crime data to debunk a baseless claim. It is your claim that there is a significant relationship once you correct for X. The onus is on you to find the data and show the results.
However, I don’t actually need to look at crime stats to prove your claim wrong. As it happens, your claim about population changes post-1980 actually fails entirely to support your larger claim. If you were only going to choose two points to look at, you really should have chosen 94 and 05, which could actually have proven your claim wrong or right, rather than 80 and 05, which are basically irrelevant to the discussion.
So lets compare 1995 and 2005. As it happens, 1995 was the year that the trough of the baby bust fell in the 14-24 age range. 2005 actually had a slightly higher percentage of the population between ages 14 and 24 than 1995 did. So the change in the size of the 14-24 cohort explains none of the decrease in crime from 1995 to 2005, and is in fact a (trivial) counter-balancing factor. (for another view of the data, here are 1990 and 2000 in 5 year cohorts in table form).
The decline in the percentage of population in the 14-24 cohort (that shows up in your 1980-2005 comparison) actually occurred simultaneously with the increase in violent crime, between 1980 and 1995. The decline in violent crime rates from 1995-2005 occurred along with no decrease in the percentage of the population in the 14-24 cohort. Since the percentage of the population between 14 and 24 did not decrease, the (non-existent) decrease in that population can not be a confounding factor to the plot in the original post, and we don’t need to hunt down age categorized over-all violent crime or age categorized attempted murders to remove the (non-existent) confounding effect.
The graph is looking at mother-only families as a percentage of all families. Unless you’ve got some evidence that children of single mothers are moving back home significantly more often than children of married couples, this wouldn’t do much of anything to the statistics. (And actually, since the rate of single motherhood has been rising, I’d suspect that families with adult children would be more likely to be two-parent families than families with young children.)
Interesting read. Thanks, folks. I would like to stress that I’m not making a strong argument about what caused the crime rate to go up and down. You all have already spent more time worrying about the (important) details than the “family values” crowd of the 1980s and 1990s, who blamed the crime wave on the “breakdown of the family” with abandon. And that was my target: the easy blame game of that time.
To clarify, the red line on the chart reflects this ratio:
(family households with female householder and no spouse present / family households)
where “family household” is one in which the householder (in whose name the home is owned or rented) lives with any relative by birth, marriage, or adoption in the home. So, a single woman (of any age) living alone is not in the numerator or the denominator. A single woman living with her elderly mother would be a “single mother” family as I’ve labeled it, I’m afraid. So the title of the figure is a little off, but to get consistent numbers back to 1960 without doing a lot of work you have to use what the Census has. The great majority of these households are women living with their own children.
FYI, the terms are defined here:
http://www.census.gov/cps/about/cpsdef.html; and the table is this one: http://www.census.gov/hhes/families/files/hh1.xls
This page of historical series is great for hashing all this family stuff out: http://www.census.gov/hhes/families/data/historical.html
I seem to recall that Freakonomics (the book) made the claim that the legalization of abortion is what caused the drop in crime starting in the 1990’s.
The availability of abortion (which BTW, is not the same as legalization) vastly reduced the number of children born to (poor) mothers who weren’t in a position to take care of them, which two decades later resulted in a reduction of the number of people most likely to commit violent and street crimes.
Granted, this is by the Freakonomics folks, who seem more interested in throwing out unorthodox ideas than in doing the sort of painstaking research it would take to actually prove them, but I thought that it is at least as well-founded as anything else in this thread or in the public discussion of crime and family values.
What I don’t see discussed here is the role of economic conditions. One big reason why poor families tend to be single-mother families is because the very poor often can’t afford to be married — poor women report that most men that they know can’t earn enough to carry their share of the financial cost of a two-parent household. I’ve heard this from a number of fairly reliable sources. But of course this doesn’t fit into the neo-conservative world-view, which is why nobody talks about improving the financial state of the poor as a way to increase the proportion of two-parent families.
Damn, someone else beat me to the Freakonomics article.
Simple version, from memory:
1) Crime has always been at least peripherally related to the # of people who are in “high criminal activity” categories.
2) both “unwanted children” generally, and especially “unwanted children of poor single mothers” are in a high activity category. (I think the actual categories were different)
3) Right after Roe, a lot of those parents started being able to abort babies rather than bear then unwillingly.
4) Rght around 1990, the # of high crime kids who would have been EXPECTED to turn 18 and start committing lots of crimes, didn’t… because they were never born.
5) Because the # of crime-committing folks went down, so did crime.
5) Ergo, the
I wonder what the graph would look like if, instead of looking at what proportion of families were headed by single mothers, we looked at what proportion of children were being raised by single mothers. I’m sure the data exists, and I’m honestly not certain what I’d predict it would look like — I know that the average number of children per family has been decreasing, but I don’t know how that decrease is spread over various income groups or family types. (My mom told me that, when she was a kid in the fifties, in a blue-collar heavily Catholic city, families have 8 or more kids was not unheard of. When I was a kid in the eighties, in an upper-middle class heavily Catholic town, I had several friends whose families had 4 or 5 kids, but I didn’t know anyone with more than 5.)
Interesting graph, analysis and comments.
What puzzles me but I do not think it has been addressed directly is: ignoring the red-line, what made the blue-line shift (okay, AMM and g&w touched on it in a freaky way).
I don’t know that the graph makes much sense (either from a correlation argument or a causation one), if you don’t understand why there was a drastic shift.
Abortion? Maybe. Video games? Maybe. The rise of the internet (with its seductive lure of free porn, music downloads, and blogs by comic book writers)? Probably.
Does Cohen explain that?
It’s not freaky. It’s just that there are a huge number of factors which affect (or at some point have been believed to affect) crime rates, ranging from # of police to gun control to stricter jail laws. It turns out that “number of people who have a statistically high likelihood of committing crimes” is a really big one.
Sorry, g&w, I was trying to be punny. “Freakonomical” would have sounded odd.
Kevin Drum made a pretty convincing case today that the drop is due to a drop in environmental/gasoline lead.
And he posted a graph of levels of lead in children’s blood at the preschool level, and matched it up to a graph of violent crime 19 years later:
I’m sure it’s not the only story, but it’s certainly the most convincing explanation I’ve heard.
If the ‘lead’ story is true, we can expect a crime wave in China. The sister of a friend of mine has become her company’s go-to-person for handling Chinese matters. The stories she has about industrial pollution in China were so unbelievable that she left a party of ours because she could not take the questioning (Did her brother have something to say about that when he found out!)
And two weekends ago, she showed up with two USB sticks worth of pictures and videos she shot and commented on herself. On a lot of them we see her escorts having to physically protect her camera. Some of the shit was unfuckingbelievable. As in, tri and hexa valent Chrome solution poured from a plating tank into the ditch running along the street, and the camera panning to the river where you can see people wading. Dark streets at noon. 2-3 mm of grease/metal mixture on leaves. Her mask’s filters after 30 minutes on the street.
She now has an official dispensation to show at parties empty handed. Not that she would, no one else drinks that strawberry smirnoff lemonade abomination.
interesting, but you did not have a graph for the U.S. If memory serves, the lead paint in houses warning was in 1978 and unleaded gasoline was prominent around the same time. The sharp decline in the graph Amp gave was in 1990-1995 and it did not flatten the way yours did; it is still dropping.
I don’t know what to say, except “maybe, but still not convinced that free internet porn did not play a role.”
The full study is here, and does have a lead-to-violence chart for the US.
Here’s what actually happened:
1960s: A fashion emerges of painting houses blue on the outside.
1990: A statistical correlation is noted between the growing number of houses painted blue on the outside and common funnybird mortality.
1991: Ornithologists explain how painting houses blue harms funnybirds: they mistake the blue paint for the sky and get killed as they fly into walls at full speed, evidenced by many funnybird corpses concentrated around blue-painted houses.
1993: “My house, my color choice” advocates successfully push the Synanthropic Funnybird Relocation Act. All funnybird populations living in areas inhabited by humans are forcibly relocated into wilderness.
2000: Apparent success! Funnybird population decline stopped. Blue houses more trendy than ever before.
2012: Funnybird population back at pre-1960 levels. Blue houses are all the rage. Ampersand happily presents the updated graph as a proof that there was never a relation between painting houses blue and funnybird mortality.
Interesting unsubstantiated nonsense, TheSameDog.
Tell me, what would the real world equivalent be for the Synanthropic Funnybird Relocation Act in your scenario?
Don’t look for an exact entity-for-entity analogy.
In rectrospect, I could just have said that sometimes causal links between phenomena get severed, and the fact that two variables don’t appear correlated over a larger time frame is not a proof that there was no real causal link in the time frame when the correlation was very strong.
What would be this thing that severed the link between fatherlessness and crime rate? Everything that worked toward removing the negatives of being a single mother and being raised by one. Like, decreased social stigma; child support and alimony becoming easier to obtain and more
ruthlesslyeffectively collected; legislation focused on giving women advantage without any concern for fairness; legislation not focused on giving women unfair advantage, but still having that effect as a result of being designed without explicit concern for fairness; and quite possibly decrease in effective fatherlessness when fathers learned how to navigate the hostile legal territory and avoid being completely severed from their children.
But actually my Occam bone likes the lead pollution theory even better.
I completely agree with this:
“sometimes causal links between phenomena get severed, and the fact that two variables don’t appear correlated over a larger time frame is not a proof that there was no real causal link in the time frame when the correlation was very strong.”
But it is an uphill climb for tha theory after the trends diverge. It takes something like a good new variable (lead, video games). I don’t buy the stigma or supposed favoritism toward women things.
Because it isn’t obvious from the layout of Kevin Drum’s post, the plot that Myca quoted comes from the paper, Understanding international crime trends: The legacy of preschool lead exposure (Nevin, 2007). This paper has a ton of excellent detail, including plots of distributions of violent crime (and murder, property crime, robbery, and assault) by age in 1980, 1994, and 2001.
Crime per capita is not the same as criminals per capita. Just take a look at what also happened over the past few decades: http://en.wikipedia.org/wiki/File:U.S._incarceration_rates_1925_onwards.png There are probably 100 different countries out there with smaller populations than the US prison population.
Pingback: Conservatives doubling down on single mothers and crime « Family Inequality