Walter Dellinger and Sri Srinivasan had a nice analysis yesterday in Slate.com of the inefficacy of voter identification laws--requirements that people show poll workers a valid government-provided ID before being allowed to vote. The principal argument in favor of these provisions is that they cut down on voter fraud and thereby increase the integrity of the electoral process. The primary argument against the provisions is that they impose burdens on voters, especially low-income and elderly voters, which reduces participation and representation in elections. Republicans tend to favor the laws, while Democrats tend to oppose them.
If we can move past the purely self-serving aspects of these provisions, the arguments essentially come down to a cost-benefit analysis. Proponents claim that there are large potential benefits in terms of reduced voter fraud and only small costs because IDs are so ubiquitous. Opponents claims that voter fraud is negligible and that the laws disenfranchise tens of millions of people who lack IDs. What Dellinger and Srinivasan bring to this debate is a nice discussion of the mechanics of these procedures and thus of the marginal impacts.
To impersonate a registered voter, an impostor would have to already know the voter's name and address and provide these details to a poll worker. Poll workers could further ask the person for the registrant's birth date (states require this data to verify that the person is actually old enough to vote). This creates a very high informational hurdle for a potential impostor.
In addition to these informational barriers, there are a couple of other practical issues. Most importantly, there are laws and penalties against voter fraud, so the impostor would need to take steps to avoid detection. For example, he or she would need to be sure that the actual voter had not already shown up at the polls and cast a ballot, otherwise the fraud would immediately be discovered. Along these same lines, there are limits on the number of times that a single individual could commit this crime (it's not like he or she could immediately get back in the queue at the same precinct and pretend to be someone else). Finally, as mentioned above, the impostor would have to have a fairly sharp mind, needing to commit all of the victim's personal details to memory (a person who needs to consult a cheat-sheet before providing basic identifying information tends to rouse suspicion).
To be sure, forms of reliable, current identification are needed in other circumstances, including the initial registration process and certainly in the case of same-, or election-day registration (North Carolina has such a procedure). However, once registration has taken place, a subsequent requirement to produce a valid government ID does almost nothing to stop fraud and therefore confers little benefit.
Given that IDs are required at the time of registration, one might claim that there also is little cost of an election day ID requirement, but this overlooks the fact that considerable time can pass between the registration and election days. During this span, people could lose their driving privileges, or especially in the case of the elderly, could simply let their licenses lapse. These people would clearly be disadvantaged by a voter-ID provision. Also, while it might be worthwhile to bring an alternative form of ID, such as a passport, one time to the registration office, it would be more costly to dig these same documents out every election day. Finally, even if one buys into the argument that the costs are low, this hardly justifies enacting a law with such little practical benefit.
Republicans regularly complain about "feel good" legislation that sounds nice but accomplishes little. That criticism applies in spades to voter-ID laws.