NOTE: Due to a misunderstanding on my part, there’s problems with my initial conclusion, see updates below body of post.
I have an op-ed I’m hoping will get published in some of the local papers soon dealing with immigration and one of the issues I touch on is the movement to make E-Verify mandatory for all employers in America. Since it is still being considered, I won’t publish any of it here, but I want to expand on the E-Verify point.
Even as anti-immigration activists seek to make E-Verify mandatory at the national level, they are pushing hard to get states to mandate it in the meantime. Simply put, this is a horrible idea.
I could point to a lot of different reasons why:
But I want to focus on one in particular. False positives.
For those unfamiliar, a false positive is when a test finds a person positive for something when they should be negative. You hear it a lot with regards to medical tests, like AIDS for example.
Since the E-Verify system is essentially a check to see whether you have the condition of being here illegally, it means it too is subject to false positives.
After pulling some numbers for Maryland and doing a Bayesian inference I found the following:
Using the most generous numbers I could for E-Verify supporters, 6.3% 6.1% of positives in Maryland would be false positives, equating to 354,916 343,649 people.
However, as I said, that’s being generous. I also ran a more realistic version of the numbers, using the changes listed below:
- My source for Maryland population was the U.S. Census Bureau. Their figure is 5,633,597. According to Help Save Maryland the number of illegals in the state is 250,000. In my generous calculation I treated this as a subset of the total state population. However, realistically, illegals are unlikely to be part of Census numbers so I have added them to the population number for a total of 5,883,597.
- For my generous calculations I was operating using a standard of 99.9% test accuracy for finding illegals as illegals. For my realistic one I downgraded to the 99.5% accuracy cited by the Center for Immigration Studies (still pretty generous in my opinion).
- In the first calculation I relied upon the .3% false positive rate cited by E-Verify supporters. As Jim Harper has noted, a number this low is suspect considering E-Verify’s previous track record, so I moved it up to 1%, still short of his 1.24%
So, running the numbers using these adjusted figures, the rate of false positives increases from 6.3% to 18.6%, or 1,094,349 people incorrectly being denied work (1,047,849 if you apply the percentage to the unadjusted population number).
When the stakes are as serious as they are with E-Verify, an error rate that impacts so many people is inexcusable. Maryland has no business mandating E-Verify and I encourage everyone to reach out to your state representatives and let them know that.
UPDATE 1:
As my good friend Tim Andrews has pointed out, I really should include the calculations.
First version:
.999(.044)/[.999(.044)+.003(.956)]=.939
1-.939=.061=6.1%
Second version:
.995(.042)/[.995(.042)+.01(.958)]=.814
1-.814=.816=18.6%
The .999/.995 refers to the accuracy of the test in identifying illegals as illegals.
The .044/.042 refers to the percentage of illegal immigrants in the Maryland population.
The .003/.01 refers to the percentage of false positives.
The .956/.958 refers to the percentage of persons in Maryland that are not illegal immigrants.
Hope that clears up any confusion.
UPDATE 2:
It has been pointed out to me by a friend far better at statistics than I that I mis-applied the result of the Bayesian inference and that it should have been applied to the number of total positives and over-sized the total sample by using a portion of the population to approximate the working portion (his recommendation was half). So here are my corrected calculations and results (all data say the same except the general population number has been halved and the illegal immigrant population is 84% of its previous total; this was derived by applying national demographic info from Pew’s Portrait of Unauthorized Immigrants in the United States report to state numbers.)
.999(.075)/[.999(.075)+.003(.925)]=.96
1-.96=.04=4% of all positives being false positives
.995(.069)/.995(.069)+.01(.931)]=.881
1-.881=.116=11.9% of all positives being false positives
If we accept the 5.8% rate of positives cited by the DHS is accurate (I’m skeptical the system won’t turn up more positives, especially false ones, as more strain is put upon it), that still comes out to 6534 Marylanders being unjustly denied employment because of E-Verify under the most generous model and 20,891 under my realistic model (18,951 if you use the realistic percentage but the generous population number).
Obviously this isn’t as strong a point as I initially thought I had, but I think even 6534 people improperly denied employment because of E-Verify is too many – and when you consider that there’s no reason to think things would be as ideal as that forecast, the problem becomes even more pressing.
Recent Comments