My Concerns with John Lott’s Arizona Study

Conservative media outlets and their friends on Twitter have been ablaze with a John Lott study of Arizona criminal convictions that purports to show that undocumented immigrants in Arizona commit crimes are much greater rates than citizens in Arizona.

If true, the Lott study challenges the academic consensus that undocumented immigrants commit fewer crimes than the general population (in AZ at least).  Too many on the left are simply dismissing the study because of the author’s past work.

I’ll admit up front, absorbing data that is critical of undocumented immigrants is a big struggle for me. I’m an empiricist but I also have close relationships with “dreamers,” so it’s personal. I try hard to resist the urge to discount research that doesn’t fit my narrative.  For immigrant rights activists, there is stuff here to reckon with.

Here’s a takeaway stat:

Undocumented immigrants are 163% more likely to be convicted of 1st degree murder than are U.S. citizens, 168% more likely to be convicted of 2nd degree murder, and 189.6% more likely to be convicted of manslaughter. Those three categories and negligent homicide added up to 987 convictions. Undocumented immigrants are also much more likely to commit sexual offenses against minors, sexual assault, DUI, and armed robbery.

I have four core issues with this work. One flaw I find with it is that Lott creates his ratio of “undocumented crimes” by using estimates of the undocumented immigrant population in Arizona with actual convictions of undocumented immigrants in that state. If there are more undocumented in AZ than estimated, it throws off his ratios.  The main challenge of studying undocumented immigrants is the difficulty inherent in counting populations that don’t want to be counted.  In fact, Lott cites this as a key problem with earlier research.  Nonetheless, he’s using estimates as his denominator.  This might be the best he can do in terms of getting numbers, but it does render the rest of the data questionable.

Second, Arizona seems like the “best case scenario” if you want to paint a group as criminally minded.  As the most porous part of the southern border, it has become a flashpoint for drugs, trafficking, etc., so it may be the best case study to pick if you want to paint undocumented people as prone to criminality. If you replicated this study in California or Texas, you’d probably get wildly different results.

Third, I’m puzzled that Lott is presenting descriptive statistics with no effort at more rigorous causal statistical analysis.  Why no inferential study? Lott presents ratios with little control for third-party effects and little context (so if 1 out of 100 citizens are convicted of armed robbery but 2 out of 100 undocumented immigrants are convicted of armed robbery, then undocumented immigrants are 100 percent more likely to commit crimes than citizens. He does control for age (sort of), but why not do a multivariate analysis so you can control for more factors like income or educational attainment or any other typical predictors of crime rates.  This might be because of the limitations of his data, but without some kind of hypothesis testing, we really don’t know what we have here.

The final thing that stick with me is why “documented immigrants” are below average in committing crimes but documented immigrants are way more prone to engage in illegality.  One would presume that these populations aren’t significantly different (other than the minority of actual criminals crossing the border that no one would deny is true). What’s the theory that explains this? Are undocumented immigrants drawn from a different pool than documented immigrants? Doesn’t this challenge the notion that legal immigration needs to be reduced?

Without addressing these concerns, the study gives the veneer of “science” as possible cover for discriminatory views (whether that is the study’s intent or not).  It worries me that I haven’t seen anyone take on this work, not because it isn’t flawed, but because people just want to dismiss the author as “tainted.” To me, that is a cop-out. This study is all over conservative media and in a democratic society, we owe each other the benefit of the doubt (even if that isn’t always reciprocated). I get where restrictionists are coming from. Social cohesion is an important element of a democratic society and immigration has the potential to undermine that cohesion (although I have much more faith in the American project’s ability to incorporate people under American ideals). I have my doubts about the veracity of this data, but I’m open to having my mind changed on the subject, even if I’m squarely on the side that immigrants (undocumented or documented) are an important driver of growth and progress in this society.

Correction: Lott does include a logistic regression analysis in one of the appendices of the paper, but he doesn’t appear to discuss it in the text of the paper.

Jose Marichal is a professor of political science at California Lutheran University.