Why Don't Federal And Texas Homicides By Illegal Immigrants Match?

I recently encountered a piece at American Thinker purporting to show that illegal immigrants murder at much higher rates than the general US population. Randall Hoven's piece starts with a March, 2011 GAO survey (PDF) estimating the number of murderers among the ranks of "criminal aliens" (a group that includes people legally in the United States) in state jails under the DOJ's State Criminal Alien Assistance Program, or SCAAP) and ones in the Federal Bureau of Prisons (BOP) committed over two periods:

To determine the number and nationalities of criminal aliens incarcerated, we analyzed BOP data on criminal aliens incarcerated in federal prisons from fiscal years 2005 through 2010 and SCAAP data on criminal aliens incarcerated in state prisons and loca l jails from fiscal years 2003 thorough 2009.

Hoven takes 2009 DHS estimates of 10.8 million illegal aliens in the United States, and a total alien population of 25.9 million, from the GAO study. Given there are two periods and two prison populations studied by the GAO, this leads to obvious problems with coming up with the right numerator and denominator. Over the study period, on page 21 (PDF page 27) they claim 25,064 criminal aliens were imprisoned for homicide. Assuming a static population over the longer of the two terms studied (which is clearly false but will serve to increase the rate otherwise), that means

Let’s take homicide as an example. The GAO estimates “criminal aliens” were arrested, convicted and incarcerated for 25,064 homicides. If non-citizens committed them over seven years, the annual rate would be 14.2 per 100,000 non-citizens. If illegal aliens committed them over four years, the annual rate would be 58.0 per 100,000 illegal aliens. Either way you compute, those are high rates.

So far, so bad (at least for what this means). We should expect to see similar figures from the states. Yet this hasn't materialized. Texas, according to the DHS, had 1.9 million illegal immigrants as of 2014 (PDF). The Texas Department of Public Safety records only 225 homicide convictions over a seven-year period from 2011 through 2018. This correlates with an annual homicide rate of around 1 per 100,000, which is dramatically lower than the general US population figure of around 5/100,000 — more than an order of magnitude of those calculated by Hoven, in fact.

So why the discrepancy? Digging deeper into the Bureau of Justice Assistance website, it turns out SCAAP eligibility (PDF) is contingent upon

Each applicant government is to provide detailed information about the individuals —

(1) whom the applicant government “incarcerated” for at least four consecutive days during the “reporting period,” and
(2) who the applicant government either knows were “undocumented criminal aliens,” or reasonably and in good faith believes were “undocumented criminal aliens.”

Which is to say, there's a considerable financial incentive to overreport such individuals. This isn't the end of this discussion, but it bears substantial further investigation, especially given the policy stakes.

Update: One interesting comment on this comes from the National Council of State Legislatures, which mentions that SCAAP eligibility changed starting in 2012: "SCAAP payments will only be made to states and localities only for those inmates that DHS can verify as illegal aliens. DHS will no longer reimburse states for “unknown” inmates (58 percent of the program in 2010)." Indeed, looking at the 2010 data (Excel spreadsheet, the closest year available before the 2011 GAO report, more years available here), 43% of the total presented to UCA for SCAAP reimbursement (21,831 of 50,402 total from Texas) fell into the "unknown" category. This seems ripe for exaggeration, as well.

Update 2: It would appear that SCAAP data would also admit illegal aliens, suspected or actual, who had been jailed in preparation for a trial in which they may have been exonerated. So the "crime" listed might entirely be speculative.