I think the law would better serve the rights of voters if it were a reasonable match. It’s foolishly optimistic to think that there would be no errors in the myriad databases — believe me, I know, I work in a medical billing office. Errors happen all the time, especially with names that are unfamiliar or that adhere to a different naming custom.
Take a name like Michael Smith — unlikely to be misspelled in the database.
Then take a name like Miguel Gonzalez-Santiago. He might be in there under Miguel Gonsalez, Miguel Gonzales, Miguel Gonzales-Santiago, Miguel Gonsalez-Santiago, Miguel Gonzalezsantiago, Miguel Gonzales Santiago, Miguel Gonzalez Santiago, and so on.
Take any name in Chinese, Japanese, Greek, Russian, or Arabic. They don’t romanize well. In addition, a common name like Phuong Nguyen might be in the computer as Nguyen Phuong, to make it adhere to the given-name-surname order of Western names.
We don’t track race data in our computer system, but as an unofficial eyeball of the naming trends, people with unusual names with more unique spellings (Uneeque, Jakkyb, Mickel, Shaniqua) tend more to be on welfare.
Any kind of “exact match with zero errors” system will favor voters with easy-to-spell names that are fully fluent in English and who have been long-time residents. I understand the desire to make certain that illegitimate votes are not counted, but throwing away 1,000 legitimate votes to catch 1 felon seems contrary to the principle that it is better that 1,000 criminals walk free than to convict 1 innocent man. An “exact match” system just feels like a de facto English literacy requirement.