My thanks to Ron Unz (Unz) for publishing a detailed reply to our recent critique of his article on Hispanic crime rates. However, since, speaking as the numbers guy for our critique, I believe his analysis is largely mistaken, I'm providing a response.
In his original article, Unz argued that total federal, state and local imprisonment rates were misleading, and that federal rates ought to be excluded, because so many Hispanics were in federal prison on immigration related charges - which, in his opinion, shouldn't count. In our critique, I demonstrated that simply removing immigration related offenses from the equation did not dramatically reduce the white/Hispanic contrast. Unz does not challenge this result. Instead, he tries to justify his exclusion of federal imprisonment statistics by coming up with yet another category of Hispanic crime which, in his opinion, shouldn't count: drug smuggling.
He writes: "illegal drug activity is certainly a crime, but including these numbers may also distort our estimates of ethnic crime rates...there exists a thriving drug-smuggling operation all along our southern border, with low-level Mexican 'drug mules' being hired to evade border guards and bring their illicit cargo into the United States on a regular basis. If they succeed, they go back home for another load, but if they're caught, they end up in federal prison. Now these individuals are obviously Hispanic, but since they actually live in Mexico rather than in the U.S., it hardly makes sense to regard them as 'American' criminals."
Well, I guess that's all right, then - nothing to see here!
Anyway, Unz continues: "The main focus of my analysis is to attempt to estimate the relative criminality of America's whites and Hispanics with regard to ordinary day-to-day 'street crimes' such as robbery, rape, murder, theft, burglary, fraud, arson, and assault. Such crimes are almost always prosecuted in state courts and almost no federal inmates are currently held on such charges. Therefore, it seems reasonable to exclude federal incarceration rates lest the large numbers of immigration violators and border drug-smugglers generate a statistical artifact distorting the overall ethnic data."
As usual, Unz is playing fast and loose with the facts. According to the Federal Bureau of Prisons, at last count there were 37,293 Mexican citizens in federal prison. Even if they were all in there for either immigration related offenses or drug smuggling, that would still be only 17.9% of the total. Unz calls these "large numbers." Meanwhile, how many federal inmates were in there for crimes "such as robbery, rape, murder, theft, burglary, fraud, arson, and assault?" More than a third! Unz calls this "almost no[ne]."
By what alchemy can Unz multiply a "large number" by two and come up with "almost none?"
But enough of that. As we wrote in our first critique: Unz is entitled to ask whatever question he wants to, provided that he answers it honestly. So if it's state and local imprisonment that he prefers to discuss, it's state and local imprisonment that we shall discuss.
* * * * *
Which brings us to those bizarre numbers he came up with for the comparative state and local imprisonment rates of whites vs. Hispanics in the 18-29, 15-34 and 15-44 age ranges. (See Fig. 1.) By way of explanation, Unz writes: "Since the BJS state+local incarceration data is unfortunately stratified neither by age nor gender, I was forced to develop a methodology to adjust these figures for the age and gender differences of the relevant ethnic populations...my approach was simply to divide the total number of inmates by the total high-crime-age male population for each ethnic group to produce an age-adjusted incarceration rate..."
At this point I must confess that it's only with the greatest difficulty that I've been able to persuade myself that Unz actually means what he seems to be saying here - i.e., that in an attempt to produce, e.g., an "age-adjusted incarceration rate" for white males in his various chosen age ranges, he simply took the total number of white inmates regardless of age and sex and divided by the total white male population in those cohorts - and likewise for blacks and for Hispanics. But apparently that really is exactly what he did. If you take a few hours to work through this absurd procedure, you really do end up with the sort of numbers shown in his Fig. 1 - numbers that are not even remotely in the ballpark of the truth.
And here I was, naively hoping that he had just mis-labelled the scale on his graph!
I mean, what can one say, in the face of such open and apparent statistical malpractice? - except that somebody needs to take away this man's calculator before he sins again.
Do I really need to point out that the result of this procedure is not an "age-adjusted incarceration rate?" - that, on the contrary, it is a completely meaningless number? Obviously - or so, at any rate, I would have hoped - if one wants to know the incarceration rate for, e.g., 18-29-year-old white males, one needs to start with the number of 18-29-year-old white male inmates alone - and not the total number of white inmates of all ages and both sexes?
Ye Gods.
Unz concludes this part of his reply with a challenge: "Since I don't claim that my methodological approach to age-normalization is perfect, I'd welcome MR+SB to suggest and implement a superior approach."
Well, I could be snarky and suggest throwing darts at a dart-board and recording the results. But I guess that would be...ummm...snarky. So here goes:
In the first place, one should begin with better and more recent data. As Unz admits, there is a "slight discrepancy" between the figures reported by BJS for state+local as opposed to federal+state+local imprisonment in 2005. What he calls a "slight discrepancy" I would call a "gross contradiction." Besides, these are old numbers, and there's a perfectly straightforward way of extracting reasonable estimates from the latest data - so why not give it a try?
According to the BJS' "Prison Inmates At Midyear 2008 - Statistical Tables" the total number of Hispanics in federal + state prisons + local jails as of June 30, 2008 was about 460,400 (See Table 17). According to Prisoners in Federal prison at year-end, published by the Federal Justice Statistics Resource Center, there were 55,345 Hispanics in federal prison at the end of 2007, and 56,565 at the end of 2008. So the number at midyear 2008 was presumably about 56,000 - i.e., about 12% of the total.
To the extent that the profile of federal inmates by age and sex resembles that of state and local inmates, we can now estimate the state + local imprisonment rates for Hispanic males in various age ranges simply by multiplying the figures in Table 19 of "Prison Inmates At Midyear 2008" by 0.88. Since federal inmates skew somewhat older than average, this will give us figures that are a little too low for younger prisoners and a little too high for older prisoners - but not by much. Here are the results for Hispanic males aged 18-29:
18-19: 2091 inmates per 100,000 population
20-24: 3767
25-29: 3337
Now let's follow the same procedure for non-Hispanic whites. There's a complication, since the Federal Justice Statistics Resource Center does not break down their racial data by ethnicity. But, as usual, let's do our best for Unz, and assume that absolutely all of the Hispanics in federal prison were counted as white. Even so, there were still at least 47,000 non-Hispanic white inmates at midyear 2008 - i.e., about 6% of the federal + state + local total.
Reducing the figures for non-Hispanic white male imprisonment rates in Table 19 by 6%, we get these results:
18-19: 917
20-24: 1470
25-29: 1457
Here are the ratios of the Hispanic male to the non-Hispanic white male rates:
18-19: 2.28
20-24: 2.56
25-29: 2.29
Or, as Unz prefers to say, the rates for Hispanic males exceed those for non-Hispanic white males in this age range by between 128 and 156%. So much for his absurd claim, based on his absurd "methodology," of 13-31%.
* * * * *
Let us now speak of urban crime comparisons.
Unz continues to insist on his contrast between Seattle and San Jose. He writes: "Seattle and San Jose are very similar cities in most respects, and...although the former is among the whitest cities in America and the latter is one-third Hispanic, San Jose's crime rates are actually much lower, with roughly half the rate of robberies or violent crime in general. MR+SM argue that the crucial underlying difference is that Seattle is 8.4% black while San Jose is only 3.5% black. But this doesn't make any sense. If Hispanics actually had a much higher crime rate than whites, then a 4.9 point difference in the black population couldn't possibly swamp the impact of a 30 point difference in the Hispanic population..."
Once again, let's look at the latest data. As it turns out, Wikipedia, relying on the American Community Survey conducted by the Census Bureau, has interesting, and sometimes surprising, numbers for the very different demographics of Seattle and San Jose as of 2007.
non-Hispanic white: 64.9%
non-Hispanic black: 11.5%
Asian: 15.4%
Hispanic: 6.2%
non-Hispanic white: 31.7%
non-Hispanic black: 3.1%
non-Hispanic Asian: 30.5%
Hispanic: 31.3%
Suppose, in line with most estimates (other than Unz's) a Hispanic crime rate about twice the white rate, and a black rate about seven times the white rate, and an Asian rate about one quarter the white rate. In that case, given the above demographics alone, with no further information, I would expect the overall crime rate in Seattle to exceed that in San Jose by about 30%. (If anybody would like to see my math on this, just ask, and I'll provide it in the combox.) Now add in the fact that "the median income for a household in [San Jose] was the highest in the US for any city with more than a quarter million residents," and exceeded Seattle's median income by more than 40%. Finally, add in the fact that an astounding 36.9% of San Jose residents are foreign born (compared to 16.9% for Seattle) - and that first-generation immigrants are, by all accounts, far less criminally inclined than their progeny.
So, remind me: what was Unz's point supposed to be, here?
This example nicely illustrates the complete uselessness of Unz's lengthy exercise, in his original article, of "calculating the weighted-average correlation coefficient between the Hispanic percentage of a city and its various crime rates and performing the same calculation for the white-plus-Asian percentage as well." If one lumps in Asians with whites and ignores blacks entirely, one might well find the contrast between the crime rates of Seattle and San Jose puzzling. But once one disaggregates Asians from whites and includes blacks in one's analysis, the mist dispels and the light dawns.
The same weaknesses fatally undermine Unz's discussion of the decline in crime in Los Angeles over the past two or three decades. The percentage of the black population has decreased dramatically, while the percentage of the Asian and first-generation immigrant Hispanic population has increased. So of course the crime rate has gone down. What else would one expect? But what does this prove about the relative average criminality of Hispanics compared to non-Hispanic whites? Nothing whatsoever. All it proves, as Edwin S. Rubinstein puts it, so succinctly, is that, when it comes to relative contribution to crime rates, "Hispanics are not blacks."
No more than Asians are whites.
* * * * *
And on and on it goes. Unz wants to talk about Santa Ana (53.3% first generation immigrants) while ignoring cities like Tucson and Phoenix and Albuquerque with their large and long-established Hispanic populations - and appallingly high crime rates. And he wants to talk about El Paso (26.1% first generation immigrants) as if it existed in a void, and not as a sort of relatively upscale suburb of its big cross-border brother, Ciudad Juárez, teeming cess-pit of crime. He seems blissfully unaware of the possibility that, for those who are into the cherry-picking game, there are much bigger and juicier cherries to be picked on the other side of the argumentative fence.
But I am no cherry-picker. I prefer to leave that sort of thing as an exercise for the reader.
Comments (5)
Good response, Steve.
Posted by M.A. Roberts | March 2, 2010 10:12 PM
I could tell he was throwing the cherries in the basket. You did a good job of showing which ones!
Posted by Terry | March 3, 2010 1:13 AM
O tempores, o mores! How I wish this were the case. This is a prime example of why there ought to be a law against political scientists: they run around telling people things about a subject, law, about which they actually know nothing. This statement would have been true forty years ago. Today, federal criminalization has proceeded at such a pace that vast numbers of seemingly ordinary state violent crimes wind up in federal courts through various jurisdictional hooks. Almost any crime involving a firearm can potentially result in a federal prosecution, for instance.
Here's another methodological weakness, although I doubt it's unique to this Unz fellow: local jails ought not really to be included in this statistic. Persons incarcerated in jails are principally defendants held pending or during trial, or petty misdemeanants. The first category shouldn't fairly be characterized as inmates and the second population turns over too quickly to be accurately measured by yearly statistics.
Posted by Paul | March 3, 2010 8:31 AM
Here is an overview of the Unzism debate now taking place on many websites:
http://conservativetimes.org/?p=4723
Posted by M.A. Roberts | March 3, 2010 1:51 PM
Very nice, Steve. Unz's chicanery is pretty astounding, in the worst tradition of agenda-driven statistics.
Posted by Cyrus | March 3, 2010 8:51 PM