“She has identified 35,750 persons who voted in North Carolina sharing a name, birth date and Social Security number with someone who voted in another state in 2012. “
This is mistaken. The report identified 35,750 people who voted in NC sharing a name and birthdate (not a social security number) with voters elsewhere. There were around 750 people who matched name, birthdate and social security.
Were the social security numbers for the other 35,000 people compared and found not to match up or were they not compared at all? The article doesn't state which of these possibilities is the case.