The March issue of The American Conservative features Ron Unz's long essay, "His-Panic: The Myth of Hispanic Criminality." His surprising conclusion, contrary to most of the technical literature, is that Hispanics in the U.S. do not commit crimes at higher rates than white Americans.
Unz's article is laudable for its straightforward, dispassionate discussion of a sensitive issue. His methodology, however, is problematic, and his conclusion is wrong. A proper analysis of the data indicates that Hispanics have a substantially higher crime rate than whites.
The key statistic here is the Hispanic incarceration rate divided by the white incarceration rate. Let's call this number HDW for short. So hypothetically, if four out of every 100 Hispanics are in prison, and two out of every 100 whites are in prison, then the HDW is two, meaning Hispanics are incarcerated at twice the white rate.
At first glance, the national HDW seems to be rather damaging to Unz's case. Table 13 in the 2005 Bureau of Justice Statistics (BJS) report, featured prominently in Unz's article, makes the HDW calculation straightforward. When we include prisoners of both genders and all ages in federal, state, and local confinement, HDW is 2.62.
But the main point of Unz's article is that 2.62 is a major overestimation. Unz prefers to use only the state and local incarceration data found in table 14, which immediately reduces HDW to 1.80. He then crudely controls for age and gender, which gets him down to the 1.1 to 1.3 range. With a national HDW this close to one, Hispanics are actually seen to be less crime prone than whites in several states, including Florida where HDW is below 0.5. Only in the Northeast are Hispanics still substantially more criminal than whites.
Unz goes on to cite surprisingly low crime rates in certain majority Hispanic areas, and he concludes by saying: "Taken as a whole, the mass of statistical evidence constitutes strong support for the 'null hypothesis,' namely that Hispanics have approximately the same crime rates as whites of the same age."
The problems start with Unz's exclusion of the federal prison data. His reason is superficially plausible -- the sole offense of some Hispanics in federal custody may be an illegal border crossing. It hardly seems fair, as Unz writes, to group "illegal nannies convicted of illegal nannying" with robbers and rapists who are typically put in state prisons. After throwing out the federal data and reducing HDW from 2.62 to 1.80, Unz says the reduction "presumably reflect[s] the exclusion of immigration-related federal offenses."
That's not true at all. The reduction mainly reflects the vagaries of state-level data collection. The reality is that relatively few people are held in federal prison for immigration violations unless they have also committed a more serious crime. It would be difficult to find any illegal nannies doing hard time.
In fact, the BJS data make it clear that mere immigration-related offenses cannot explain the reduction in HDW when the federal data are thrown out. As Matthew Roberts and Steve Burton recently pointed out in Chronicles, the number of people in federal prison for immigration violations is very small compared to the combined number of Hispanics in federal, state, and local custody. Even under the strong assumption that they should all be discounted, immigration-related offenses can account for only a tiny part of the HDW reduction.
Confronted with that observation, Unz developed a new reason to exclude the federal data -- too many cases of drug smuggling, which he says are perpetrated by Mexican "drug mules," not real Hispanic Americans. This is a weak argument, since otherwise law-abiding illegal immigrants also go back and forth across the border. Should they not count as American residents either?
But the larger point here is that neither immigration nor drugs nor anything else can even come close to explaining the HDW reduction. The BJS has actually published incarceration data that are grossly inconsistent.
Consider that table 13 of the BJS report lists incarceration rates at the combined federal, state, and local level. The resulting HDW, as discussed above, is 2.62. We know exactly how many federal prisoners are white versus Hispanic, so we can subtract them out and get just state and local rates. (Note that we have eliminated not just immigration or drug violators, but all federal prisoners.) When we do that, HDW moves from 2.62 to 2.44.
Now, table 14 reports state and local incarceration data only. We would expect the HDW derived from table 14 to be right around 2.44 as well, but it is actually the 1.80 that Unz uses. What causes this large discrepancy? It's a mystery. I've spoken with two different statisticians at the BJS about this problem, but neither had a good answer, other than to say that the two tables are calculated from different source material.
I can speculate that some states in table 14 have counted many Hispanics as white, which would explain why the HDW is unusually low compared to other estimates. In any case, when one table implies a state and local HDW that is over 35 percent higher than what is reported by another table, we must conclude that the report is not reliable enough for our purposes. Unfortunately, all of Unz's national and state-level analyses depend on it.
An even more serious problem with Unz's analysis is his glossing over of the generational issue. U.S.-born Hispanics are much more likely to be incarcerated than foreign-born Hispanics. Though it may seem counter-intuitive, this is a finding that is well known to researchers in the field, and it implies that Hispanic crime will become more of a problem as time goes on, not less.
While neighborhoods with mostly first-generation Hispanics do not have a large crime problem, entrenched Hispanic communities are usually more dangerous. In many of Unz's most prominent examples of low-crime Hispanic communities -- Silicon Valley, the Jackson Heights section of Queens, even Los Angeles proper -- the adult Hispanic population is primarily immigrant.
This immigrant-native dichotomy is also one of the reasons that some states, particularly those in the South, appear to have such law-abiding Hispanics. Unz writes that Hispanics have crime rates "somewhat higher than white New Englanders, but somewhat lower than white Southerners." However, most Hispanic adults in the South, especially in states like Georgia and Florida, are immigrants. The reality is that U.S.-born Hispanics have a significantly higher crime rate than whites in every state in the union. Hispanic crime is not just a regional problem.
Unz adds:
Wrong. Consider again what happens between the first and second generations. The children of Hispanic immigrants are much wealthier and better educated than their parents, but they also commit crimes at higher rates. The poverty-crime correlation here is negative, not positive.[1]
Unz downplays the second generation increase in crime by stating, correctly, that the magnitude of the increase has been exaggerated by unreliable Census data on the immigrant status of prisoners. However, the crime increase among U.S.-born Hispanics is still quite real and substantial. According to the Center for Immigration Studies report that Unz cites, the Census Bureau has vastly improved the accuracy of its immigration status variable since 2000.[2]
An annual mini-census called the American Community Survey (ACS) still shows a large crime increase in U.S.-born generations, and a similar increase is evident in independent survey data. Furthermore, elevated crime in the Hispanic U.S.-born generation is consistent with an increase in several other underclass behaviors. Labor force dropout, illegitimacy, and welfare usage are all much higher among Hispanic natives than among Hispanic immigrants. Crime is no different.
So how do we get a good estimate of HDW? Probably the best method is to use the recent ACS data on institutionalization.[3] The pooled 2006 through 2008 dataset provides a large sample that can be cleanly broken down by age, gender, race, immigrant status, and state of residence.
The dataset is certainly not perfect. Minor lingering problems with the immigration status variable cannot be ruled out, and federal prisoners are included. However, we have seen that federal crimes, especially immigration crimes, account for a rather small portion of incarcerations. The underlying conclusions described below remain unaffected by these minor objections.
The following table summarizes HDW for a variety of Hispanic groups using ACS data. The analysis is limited to men ages 16 through 64, and the numbers are statistically controlled for age and state of residence using regression analysis. To illustrate, U.S.-born men of Mexican ancestry are 2.4 times more likely to be institutionalized compared to white men who are the same age and who live in the same state.
Overall, immigrant Hispanics are institutionalized at a slightly higher rate than whites, while the rate for U.S.-born Hispanics is roughly two and a half times higher than whites. Combine the generations together and we get an HDW of about 1.8, which can be expected to go up as more Hispanics are born in the U.S.
To put 1.8 in perspective, the black to white rate is roughly 6.2. Fears that Hispanic areas of the country will become as dangerous as black ghettoes are clearly unfounded. Ron Unz is right to say that there is no massive Hispanic crime wave coming to America.
This being said, Unz is wrong when he says that Hispanics are no more criminal than whites. Hispanics are, in fact, substantially more likely than whites to commit serious crimes, and U.S.-born Hispanics in particular are about two and a half times more likely. These findings are not due to age differences or immigration violations or other statistical artifacts. The reality of Hispanic crime should be one of the many factors we consider when setting immigration policy.
Notes:
1. One could argue that this crime spike is purely a first to second generation phenomenon, and that crime would begin to go down as income goes up from the third generation onward. The problem is that Hispanic income and education do not go up relative to whites after the second generation.
2. The Bureau "allocated" -- i.e., guessed -- the immigration status of over half the prisoners interviewed in 2000. According to CIS, guesses were down to 6 percent by 2007.
3. Though recognized as a good proxy for incarceration, "institutionalization" also applies to people in mental hospitals and nursing homes. Given the pre-retirement age range I use, nursing homes should not be skewing the data. And if Hispanics disproportionately end up in mental hospitals, then frankly that's something we would want to know about anyway.
Unz's article is laudable for its straightforward, dispassionate discussion of a sensitive issue. His methodology, however, is problematic, and his conclusion is wrong. A proper analysis of the data indicates that Hispanics have a substantially higher crime rate than whites.
The key statistic here is the Hispanic incarceration rate divided by the white incarceration rate. Let's call this number HDW for short. So hypothetically, if four out of every 100 Hispanics are in prison, and two out of every 100 whites are in prison, then the HDW is two, meaning Hispanics are incarcerated at twice the white rate.
At first glance, the national HDW seems to be rather damaging to Unz's case. Table 13 in the 2005 Bureau of Justice Statistics (BJS) report, featured prominently in Unz's article, makes the HDW calculation straightforward. When we include prisoners of both genders and all ages in federal, state, and local confinement, HDW is 2.62.
But the main point of Unz's article is that 2.62 is a major overestimation. Unz prefers to use only the state and local incarceration data found in table 14, which immediately reduces HDW to 1.80. He then crudely controls for age and gender, which gets him down to the 1.1 to 1.3 range. With a national HDW this close to one, Hispanics are actually seen to be less crime prone than whites in several states, including Florida where HDW is below 0.5. Only in the Northeast are Hispanics still substantially more criminal than whites.
Unz goes on to cite surprisingly low crime rates in certain majority Hispanic areas, and he concludes by saying: "Taken as a whole, the mass of statistical evidence constitutes strong support for the 'null hypothesis,' namely that Hispanics have approximately the same crime rates as whites of the same age."
The Mystery of the BJS Data
The problems start with Unz's exclusion of the federal prison data. His reason is superficially plausible -- the sole offense of some Hispanics in federal custody may be an illegal border crossing. It hardly seems fair, as Unz writes, to group "illegal nannies convicted of illegal nannying" with robbers and rapists who are typically put in state prisons. After throwing out the federal data and reducing HDW from 2.62 to 1.80, Unz says the reduction "presumably reflect[s] the exclusion of immigration-related federal offenses."
That's not true at all. The reduction mainly reflects the vagaries of state-level data collection. The reality is that relatively few people are held in federal prison for immigration violations unless they have also committed a more serious crime. It would be difficult to find any illegal nannies doing hard time.
In fact, the BJS data make it clear that mere immigration-related offenses cannot explain the reduction in HDW when the federal data are thrown out. As Matthew Roberts and Steve Burton recently pointed out in Chronicles, the number of people in federal prison for immigration violations is very small compared to the combined number of Hispanics in federal, state, and local custody. Even under the strong assumption that they should all be discounted, immigration-related offenses can account for only a tiny part of the HDW reduction.
Confronted with that observation, Unz developed a new reason to exclude the federal data -- too many cases of drug smuggling, which he says are perpetrated by Mexican "drug mules," not real Hispanic Americans. This is a weak argument, since otherwise law-abiding illegal immigrants also go back and forth across the border. Should they not count as American residents either?
But the larger point here is that neither immigration nor drugs nor anything else can even come close to explaining the HDW reduction. The BJS has actually published incarceration data that are grossly inconsistent.
Consider that table 13 of the BJS report lists incarceration rates at the combined federal, state, and local level. The resulting HDW, as discussed above, is 2.62. We know exactly how many federal prisoners are white versus Hispanic, so we can subtract them out and get just state and local rates. (Note that we have eliminated not just immigration or drug violators, but all federal prisoners.) When we do that, HDW moves from 2.62 to 2.44.
Now, table 14 reports state and local incarceration data only. We would expect the HDW derived from table 14 to be right around 2.44 as well, but it is actually the 1.80 that Unz uses. What causes this large discrepancy? It's a mystery. I've spoken with two different statisticians at the BJS about this problem, but neither had a good answer, other than to say that the two tables are calculated from different source material.
I can speculate that some states in table 14 have counted many Hispanics as white, which would explain why the HDW is unusually low compared to other estimates. In any case, when one table implies a state and local HDW that is over 35 percent higher than what is reported by another table, we must conclude that the report is not reliable enough for our purposes. Unfortunately, all of Unz's national and state-level analyses depend on it.
The Generational Issue
An even more serious problem with Unz's analysis is his glossing over of the generational issue. U.S.-born Hispanics are much more likely to be incarcerated than foreign-born Hispanics. Though it may seem counter-intuitive, this is a finding that is well known to researchers in the field, and it implies that Hispanic crime will become more of a problem as time goes on, not less.
While neighborhoods with mostly first-generation Hispanics do not have a large crime problem, entrenched Hispanic communities are usually more dangerous. In many of Unz's most prominent examples of low-crime Hispanic communities -- Silicon Valley, the Jackson Heights section of Queens, even Los Angeles proper -- the adult Hispanic population is primarily immigrant.
This immigrant-native dichotomy is also one of the reasons that some states, particularly those in the South, appear to have such law-abiding Hispanics. Unz writes that Hispanics have crime rates "somewhat higher than white New Englanders, but somewhat lower than white Southerners." However, most Hispanic adults in the South, especially in states like Georgia and Florida, are immigrants. The reality is that U.S.-born Hispanics have a significantly higher crime rate than whites in every state in the union. Hispanic crime is not just a regional problem.
Unz adds:
"There actually does exist a connection between poverty and crime ... and ... there is every reason to expect this crime rate will drop further as Hispanics continue to move up the economic ladder."
Wrong. Consider again what happens between the first and second generations. The children of Hispanic immigrants are much wealthier and better educated than their parents, but they also commit crimes at higher rates. The poverty-crime correlation here is negative, not positive.[1]
Unz downplays the second generation increase in crime by stating, correctly, that the magnitude of the increase has been exaggerated by unreliable Census data on the immigrant status of prisoners. However, the crime increase among U.S.-born Hispanics is still quite real and substantial. According to the Center for Immigration Studies report that Unz cites, the Census Bureau has vastly improved the accuracy of its immigration status variable since 2000.[2]
An annual mini-census called the American Community Survey (ACS) still shows a large crime increase in U.S.-born generations, and a similar increase is evident in independent survey data. Furthermore, elevated crime in the Hispanic U.S.-born generation is consistent with an increase in several other underclass behaviors. Labor force dropout, illegitimacy, and welfare usage are all much higher among Hispanic natives than among Hispanic immigrants. Crime is no different.
Better Estimates of Hispanic Crime
The dataset is certainly not perfect. Minor lingering problems with the immigration status variable cannot be ruled out, and federal prisoners are included. However, we have seen that federal crimes, especially immigration crimes, account for a rather small portion of incarcerations. The underlying conclusions described below remain unaffected by these minor objections.
The following table summarizes HDW for a variety of Hispanic groups using ACS data. The analysis is limited to men ages 16 through 64, and the numbers are statistically controlled for age and state of residence using regression analysis. To illustrate, U.S.-born men of Mexican ancestry are 2.4 times more likely to be institutionalized compared to white men who are the same age and who live in the same state.
Hispanic Institutionalization Rate Divided by the White Rate
 |
| Note: The analysis is limited to men ages 16 to 64, and statistics are controlled for age and state of residence. Source is the American Community Survey, pooled 2006-2008 |
To put 1.8 in perspective, the black to white rate is roughly 6.2. Fears that Hispanic areas of the country will become as dangerous as black ghettoes are clearly unfounded. Ron Unz is right to say that there is no massive Hispanic crime wave coming to America.
This being said, Unz is wrong when he says that Hispanics are no more criminal than whites. Hispanics are, in fact, substantially more likely than whites to commit serious crimes, and U.S.-born Hispanics in particular are about two and a half times more likely. These findings are not due to age differences or immigration violations or other statistical artifacts. The reality of Hispanic crime should be one of the many factors we consider when setting immigration policy.
Notes:
1. One could argue that this crime spike is purely a first to second generation phenomenon, and that crime would begin to go down as income goes up from the third generation onward. The problem is that Hispanic income and education do not go up relative to whites after the second generation.
2. The Bureau "allocated" -- i.e., guessed -- the immigration status of over half the prisoners interviewed in 2000. According to CIS, guesses were down to 6 percent by 2007.
3. Though recognized as a good proxy for incarceration, "institutionalization" also applies to people in mental hospitals and nursing homes. Given the pre-retirement age range I use, nursing homes should not be skewing the data. And if Hispanics disproportionately end up in mental hospitals, then frankly that's something we would want to know about anyway.
