John R. Lott Jr. responded to my criticism of his working paper where he claims to have found that illegal immigrants are more likely to be admitted to Arizona state adult correctional facilities than other Arizona residents. Lott did not respond to my main criticism directly, which is that the Arizona Department of Corrections (ADC) data do not allow him to identify illegal immigrants with nearly as much precision as he claimed in his paper. 

Praising the supposedly precise ADC data, Lott claimed that the “huge advantage of using the data that will be presented here from the Arizona Department of Corrections is that over our 32.5‑year period we know each prisoner who entered the prison system, their criminal convictions history, and whether he is a documented or undocumented immigrant [emphasis added].” 

Lott was so confident in his data’s ability to identify illegal immigrants that he wrote: “It is the entire universe of cases, not a sample, and thus there are no issues of statistical significance.” 

The rest of his paper depends upon the immigrants in the “non-US citizen and deportable” variable in his dataset being illegal immigrants. We showed that that variable does not exclusively contain illegal immigrants. Thus, Lott’s characterization of the utility of his data is false. This is the only point in my criticism of his paper that counts and Lott did not challenge me on it in his rebuttal, except to point out that the description of the variable in question has a comma in the codebook rather than a conjunction. I dispute that, but it is irrelevant. I’ll take his non-response to my main point as an admission that he misinterpreted the variable and that his paper does not accurately describe illegal immigrant admissions to ADC facilities. 

Lott then spends a lot of time attempting to rebut my back-of-the-envelope (BOE) calculation for 2017, which is what I called it in my blog. It’s hard to take any BOE calculation seriously and I even included a note at the end of the blog about how my BOE calculation is probably wrong. 

Lott’s focus on my BOE calculation is a sort of tacit admission of my main criticism that he misidentified the “non-US citizen and deportable” variable. As part of his criticism of my BOE, he then tried to show that most of the people in the “non-US citizen and deportable” category are actually illegal immigrants in order to argue that his data allows a pretty good estimate of illegal immigrant crime rates. That is an admission that the “non-US citizen and deportable” variable does not exclusively record illegal immigrants, that it includes some legal immigrant, and that he cannot reliably separate the two. I will take this section of his rebuttal as an additional admission that the data, on which he rested his entire paper, does not identify illegal immigrants in a separate variable. 

Furthermore, it’s telling that this section of Lott’s response sounds awfully like he’s making the assumptions that he so criticized in the “primitive” studies that he claimed were inferior to his own because he had the real data:

Before we proceed, there are a couple of numbers to put together. The percentage of Arizonans that are lawful permanent residents and non-immigrants on other visas is lower than the national average — about 12% lower. If lawful permanent residents in Arizona are deported in proportion to their share of the state population, they would account for less than 8.8% of deportations. Also note that temporary foreign workers are not very numerous, accounting for only about 12.6% of the estimated number of undocumented immigrants in the United States (=1.42 million/11.3 million). Assuming that they are deported at the same rate as illegal aliens, lawful permanent residents and temporary foreign workers would together account for 10.5% of all deportations.

Lott’s criticism of my BOE calculation (which I even criticized in my Cato blog in a note acknowledging comments by Dara Lind) notwithstanding, the 219 times that he uses the phrase “undocumented” in his working paper to describe certain admitted prisoners, needs to be amended to read “a mix of undocumented and legal immigrants in unknowable proportions” to be accurate. That he did not challenge me on this point is a tacit admission that my criticism regarding his misidentification of the “non-US and deportable” variable stands and that the findings in his paper should be significantly discounted.