Distributed Mind

[Update, March 8: This analysis is actually pretty bad; what was I thinking? Most notably, the apparent aggregation of "white" and "hispanic" into "white" screws everything up.]

Lately, masochist that I am, I have been keeping an eye on some of the rhetoric coming out of the racialist camps on the Internet. The first step to defeating an opponent is always understanding the opponent. You have to always make sure not to get sucked in, though. When I saw several times the frequently repeated claim that blacks commit some ridiculous proportion of crime in the United States, a claim that I have seen before but hadn't thought about in a while, it gave me pause. Now, of course, due to certain of my ideological leanings (ones which most conservatives, even arch-conservative backwards ones, should agree with), it doesn't really matter if blacks really did commit crime at a hugely disproportionate rate without any other explanatory factors, since everyone deserves to be treated as an individual. But you know me, I'm not the kind to let that sort of thing fly without ripping it to shreds, so, let me do so.

This is one of those issues where 30 seconds of work had a huge payoff, because I remembered from my college sociology class (where I learned just about everything I know, it seems) the FBI's national crime statistics (tracked in the the Uniform Crime Reporting Program). And, when I looked for them, it took me less than a minute to find the FBI's helpful summary of persons arrested in 2004.

Now, some sanity chacking. Note first of all that these are arrests, not convictions. If the systematic errors affecting the arrest rate are not much different than those affecting convictions, though, we should expect it to be similar. More problematic, conviction is not the same as guilt - we have no way to measure that, and why that is problematic is because we know empirically that a persons "race" affects people's response to them, and we know that specifically in the context of law enforcement and punishment, such as the fact that blacks convicted of murder are much more likely to receive the death penalty than whites convicted of murder (I say convicted not guilty since the rate of false convictions for murder is I would guess an obscene number, for whites and blacks). So, a priori we would expect blacks to be arrested more often than whites, and convicted more often than whites, even if they were only guilty at the same rate as whites. Furthermore, these sorts of numbers will almost certainly be badly affected by hidden variables. And some also are not so hidden: We have the issue of economics (poor persons are more likely to commit crime than relatively wealthy persons), geography, and prior victimization (I don't know the degree to which prior vicitimization makes a person more likely to commit violent crime or even if prior vicitimization makes a person more likely to commit a property crime, but I do know that blacks are more likely to be vicitmized) just to name the ones that come to me off the top of my head. If all of this makes it sound like I don't think that a persons "race" (outside of the specific circumstances that being a member of such a group places one in in this nation) contributes much to whether they are likely to commit a crime, you are right, I don't. Having said that, some arguments don't care that correlation does not imply causality, but rather their proponents just want license to treat blacks differently based on that correlation, so we can't get by merely pointing out that there are many contributing factors besides race (unless perhaps we could demonstrate a better one, such as saying that poverty is much more likely to contribute to criminality so we should, um, get rid of all poor people or some such).

Now, on to the numbers. First of all, I note that regardless of race, 82% of arrestees for violent crimes were male, so even allowing that they might be arrested more often than women when assuming the same probability of guilt, I suggest that racialists, kinists, and others offering these arguments would be better off getting sex-change operations and developing cloning so that we can found an entire society of only women, and cut our violent crime rates by 60% overnight, well probably less since correlation does not imply causality, and women might find a way to pick up the slack. Except that when it comes to gender (unlike "race"), there is an unquestioned component of sex when it comes to violent crime, so we actually could drastically improve our crime rate by getting rid of men. That's kind of scary. Let me get back to the topic at hand.

Next, I see that 26.8% of arrestees were black while 70.8% were white. The white category seems to be including hispanics which will probably not make the racialists happy, and is certainly going to make analysis a little more complex. For violent crimes, the proportion was worse, 36.9% black, 60.9% white. So, based on the 2000 census numbers, and some quick calculations, 1.5% (60.9% x 586,558/ 82% x 281 mil) of whites were arrested for violent crimes and 3.6% of blacks were arrested for violent crime. So, blacks were twice as likely to be arrested for violent crime. Not great, but not nearly as bad as it's made out to be. I also notice that while some people are claiming that blacks commit half the crime in this country, it's not even close, if you think arrests are indicative,which, admittedly they may not be. But anyway, even for violent crime, you're looking at 36.9% of violent crime, and only at 26.8% of crime overall. I suspect one source of the claim that blacks commit so much crime comes from the much higher incarceration rate (something like 60% it sounds like). Which itself is definitely indicative of institutional racism - blacks commit well less than half the crimes in this country and yet still they represent the majority of prisoners. How does that work? Just thinking about a number like that wants to make me shout vulgarities, especially when it is used to show the exact opposite, that we aren't being tough enough on blacks. Give me a break. Anyway, back to the topic at hand, I think that the proportions are worse in certain age ranges, so that may be one source of some of the more outlandish claims about crime rates.

One interesting thing to look at is what crimes all these people were being arrested for, anyway. I quote: "Whites were most commonly arrested for driving under the influence (893,212 arrests) and drug abuse violations (821,047 arrests). Blacks were most frequently arrested for drug abuse violations (406,890 arrests) and simple assaults (288,286 arrests)." Although, looking at the numbers themselves (see Table 43) here I notice that blacks do commit more burglaries than whites - not percentage-wise, but absolutely. Ouch. But for all other crimes the numbers are basically as you would expect based on what we have seen already. So, this doesn't tell us too much new, other than that looking at all of the numbers is more informative and interesting than looking at some (incorrect!) isolated "statistic" like "blacks commit half the crimes in the U.S."

Slightly more relevant, one could also look at the Bureau of Justice Statistics crime and victim statistics. Guess what? "Violent crime rates declined since 1994, reaching the lowest level ever recorded in 2004." That sentence doesn't do the chart justice - there's a precipitous drop between 1994 and 2004 where the violent crime rate is literally halved (although the number they give from the UCR is less than half, so I am confused). All this despite a declining proportion of whites, I might add. But, really, I point out the decreasing crime rate to remind everyone that crime, like terrorism, is often used because it invokes fear, regardless of the facts.

In closing, I want to say that I know the real picture is complex - much more complex than I have made it sound here. And I know that not all numbers are good for blacks. But that's not to say we should use those numbers to make blacks second class citizens or advocate for their removal or any of the idiocy that the racialist slobs promote. What it is to say is that certain communities like large urban communities need our help to fix the problems that they do have. We're all God's children; let's start acting like it.

(Also, standard disclaimer: My analysis was dreadful, and it has more holes in it than my car. Although I more or less know how to do it properly, it would take me days or weeks to do it. None of which would necessarily be able to eliminate the underlying problems with the data. And at the end a good statistician or demographer could probably still find problems with my analysis. This is a "back of the envelope" calculation to assure me - and you - that we are not crazy, and not a substitute for a proper analysis which I am sure is out there somewhere, since demographers do have to eat, after all.)

[Update, 11:34: One huge flaw in this analysis is that it doesn't take into account the effect overt time. So, if you arrested the same whites every year but an entirely new set of blacks, the proportionality of criminals among blacks could be higher. Now, it's not likely the affect is drastic, but it could certainly have an effect, and should be looked at - but you would need different data than we have here. You might find a larger imbalance that way (though you might also find less of one). Dang, statistics is complicated. Which is part of the reason I love them.]