There are two major data-driven arguments when it comes to voter identification. The first is from supporters: voter identification laws are needed to prevent (widespread) fraud, and requiring individuals who show up for in-person voting at the polls to present identification, photo or otherwise, will prevent that fraud. The second is from opponents: voter identification laws tend to disenfranchise (many) voters who lack voter identification, which disproportionately affects low-income and minority voters.

The problem is that neither side has much data; hence, (widespread) and (many). 

Is voter fraud a problem? In particular, is the kind of voter fraud that voter identification laws prevent a problem, much less "widespread"?  There isn't much data to support that. There are few criminal indictments or reports of such fraud (but other forms of fraud do exist and are more prevalent). Supporters say it's a "perfect crime"--it's hard to detect or catch anyone who shows up claiming to be someone else, and, even if it's discovered, there's no way to find the culprit. But, lacking data, it's hard to say.

Do voter identification laws suppress turnout or disproportionately affect minority or low-income groups? Early studies in states like Georgia and Indiana suggest, perhaps not.

That's an important fact in Pennsylvania at the moment. After some extensive litigation that delayed implementation of the commonwealth's new voter identification law in the 2012 election, the judge in the case is asking for the data.  One would think the parties would have had this information available, rather than relying on speculation and conjecture. Now, the court seeks it, and we'll see the lawyers' best efforts to address this issue head-on.

Whatever the merits of a voter identification law are in the first place--and whether it's even necessary to prevent any fraud--it's hard for opponents to point to a harm to voters without this data. (Admittedly, one could argue that the onus should be on the state to establish why even a nominal burden is required before a voter votes, but that's an issue I'll save for another day.) And as the lead Pennsylvania plaintiff in last year's litigation obtained identification shortly after the conclusion of the case, the practical barriers may be harder for the plaintiffs to establish than originally thought.