As a data analyst, I do this sort of thing a lot. I'm given a list of id's, whether it be SSNs or medical card IDs.
First, from the article:
Only in situations where no driver’s license can be provided should the State check their identity through the Help America Vote Verification (HAVV) interface provided by the SSA. The 43 States that participate provide the last 4 of the SS#, the applicant’s name, and DOB. The interface responds if a match was, or was not found, if they are deceased, and so on.
First, what was the result when they went by license number? Second, is the SSN the only criteria that was used to match?
The first problem I see with this is that this database is incomplete, by about 20%--so approximately 20% of all Americans aren't in this database. The second problem I see is that if they are only using the last four digits of a SSN, this could lead to duplicate matches--which usually means that the program will pick the first person with the combo requested and return that. Duplication can lead to false matches or no matches depending on the program you're using.
The other thing that should be pointed out is this:
In fact, Arizona weekly processing has one of the highest Non Match percentages in the country, typically between 50%-65%. In most other States this ratio is from 15%-25%.
This indicates to me, not that there is anything nefarious, but that there is something wrong with AZ's data--and it's most likely due to some sort of data hinkeyness--for instance, leading zeros in one, no leading zeros in the other, leading or trailing spaces in one that does not happen in the other, deduping errors, or my favorite, the format type error (one is stored as a character format and the other is stored as numeric--that will give you a no match every day of the year). To me, it looks like a data issue, and as for them running checks on it, that's exactly what I would do if I found a data issue.