Vlad Kogan, Stephane Lavertu and Zachary Peskowitz have posted this working draft. Here is the abstract:
There is considerable debate about how election timing shapes who votes, election outcomes, and, ultimately, public policy. We examine these matters by combining information on more than 10,000 school district tax referenda with detailed micro-targeting data on voters participating in each election. The analysis confirms that timing influences voter composition in terms of partisanship, ideology, and the numerical strength of powerful interest groups. But, in contrast to prominent theories of election timing, these effects are modest in terms of their likely impact on election outcomes. Instead, timing has the most significant impact on voter age, with the elderly being the most over-represented group in low-turnout special elections. The electoral (and policy) implications of this effect vary between states, however, depending on whether seniors are sheltered from local taxation.