Michael G. DeCrescenzo and Kenneth R. Mayer have written this article for ELJ. Here is the abstract:

How much did Wisconsin’s voter identification requirement matter in 2016? We conducted a survey of registered nonvoters in the counties surrounding the cities of Milwaukee and Madison to estimate the number of registrants who experienced ID-related voting difficulties in the 201 presidential election. We estimate that 10 percent of nonvoters in these counties lack a qualifying voter ID or report that voter ID was at least a partial reason why they did not vote in 2016, and six percent of nonvoters lacked a voter ID or cited voter ID as their primary reason for not voting. Theoretically, we argue that voter ID requirements ‘‘directly’’ affect voters who lack qualifying IDs but also ‘‘indirectly’’ affect voters who are confused about their compliance with the law. We find evidence of such confusion, with many respondents mistakenly believing that they did not have th necessary ID to vote when they actually did. Our analysis permits us to calculate bounds on the possible turnout effect in 2016. Most of our credible estimates suggest that the voter ID requirement reduced turnout in these counties by up to one percentage point.