[Image courtesy of uic.edu]
There have been a few stories recently updating states' efforts to identify noncitizen voters and remove them from the rolls. In both Florida and Colorado, widely publicized programs to use DHS lists and other data to match voter rolls have yielded 207 and 141 registered voters, respectively, confirmed as noncitizens.
These numbers prompted my friend and colleague Paul Gronke of Reed College to observe the very low "signal to noise ratio" in Florida's program, noting that only 7% of the 2,625 voters initially purged from state rolls were confirmed non-citizens.
That point prompted another colleague to make an interesting observation to me in an email, drawing an analogy to epidemiology:
In epidemiology, a false positive is the number of people who have been classified as having a trait who in fact don't have the trait. The false positive RATE is calculated by dividing the number of false positives by (false positives + true negatives). True negatives in this case would be people who are, in fact, citizens ... If they matched the full voter registration list of roughly 11.5m people, then the false positive rate is close to 0.0018%. [The similar number in Colorado is 141/3.5 million, or 0.004%] ...
The question usually asked in a situation like this is, what is the chance that someone is an ineligible voter, given that one was tagged on the first round of matching? That answer? 7% [10% in Colorado]. What if AIDS screening tests were [that] accurate? Or mammograms? Do you think we would even perform them?
I found this fascinating - not because of the statistics involved but because of what it tells us about the kind of risk that state officials in Colorado and Florida (and elsewhere) perceive from voter fraud. Judging only by actions, it would appear that these officials consider voter fraud to be a serious threat along the lines of smallpox - a highly contagious, lethal disease - and thus are willing to tolerate a high number of false positives in order to isolate and "cure" the true positives and protect the rest of the population from an epidemic. Opponents of these efforts can (and already have) suggested that such fears are overblown; in essence, that voter fraud is an isolated disease with a low degree of contagion.
Still, the depth of concern demonstrated by these matching programs - even as the number of true positives appears very small - suggests that fears of voter fraud are deep-seated and not likely to be explained or debated away.