Tilman Klumpp, Hugo Mialon, and Michael Williams have posted this draft on SSRN. Here is the abstract:

Approximately one in forty adult U.S. citizens has lost their right to vote, either temporarily or permanently, as a result of a felony conviction. Because laws restricting voting by felons and ex-felons disproportionately affect minorities, and minorities tend to vote for Democratic candidates, it has been hypothesized that felony disenfranchisement hurts Democratic candidates in elections, thus helping Republican candidates. We test this hypothesis using variation in felony disenfranchisement laws across U.S. states and over time. During the 2000s, a number of states restored the voting rights of ex-felons. Using difference-in-differences regressions, we estimate the effect of laws re-enfranchising ex-felons on the vote shares of major party candidates in elections for seats to the U.S. House of Representatives. We argue that the regression estimates provide an upper bound for the true effect of restoring voting rights to ex-felons on the vote shares of major party candidates. Using this upper bound, no House majority would have been reversed in any year between 1998 and 2012, had all states allowed ex-felons to vote.