Adam Chamberlain and Carl Klarner have posted this draft on SSRN.  Here is the abstract:

In this article, we evaluate the rationale behind partisan disaffiliation laws, which prevent a candidate from running as an independent or from switching parties if they have not adequately severed their ties to an existing party. One prominent justification for these laws is that they help prevent voter confusion, which may result in the most preferred candidate losing. Utilizing a database of state legislative elections from 1968 to 2014, we categorize independent and third-party candidates into two groups: those who have run in the past as a Democrat or Republican, who we refer to as “stealth partisans,” and “pure” non-major party candidates. The findings reveal that the latter appear to have expressive motivations for running for office, making them less strategic about where to run and unlikely to run again. In contrast, stealth partisans are much more likely to have held state legislative office, and are more likely to have run multiple times; they are also more strategic, running under conditions that are advantageous for non-major party candidates. Voters react to this, giving pure non-major party candidates far fewer votes than stealth partisans and being more apt to vote for them when “spoiling” an election is less likely. As a result, pure non-major party candidates rarely deny winning candidates majorities, while stealth partisans who have won office in the past do so more than half the time. Our findings regarding “vote stealing” do not indicate a systematic tendency for stealth partisans to take substantially more votes from the party they recently left in comparison to the other major party. Overall, our findings indicate that partisan disaffiliation laws achieve the objectives they are designed to promote.