Joshua Sellers and Roger Michalski have posted this draft on SSRN (forthcoming Vanderbilt Law Review). Here is the abstract:

Democracy is expensive. Voters must be registered, voting rolls updated, election dates advertised, voting technology purchased and tested, poll workers trained, ballots designed, votes counted and verified, and on and on. Despite the importance of election expenditures, we have a shamefully inadequate amount of information about how much our elections cost. This Article, based on a novel and painstakingly hand-coded data set, provides much needed information on election expenditures across multiple years from four states: California, Arizona, Texas, and Florida. These states, given their unique characteristics, provide a compelling sample set.

In what we believe to be a completely novel approach to the collection of election expenditure data, we supplement our hand-coded data with predictive machine learning. This allows us to estimate average annual election spending across multiple government units. Our findings, unsurprisingly, reveal great variation both across and within states. However, our findings also reveal that much of the variation is seemingly unconnected to poverty, race, and other traditional explanations of electoral disadvantage. This brings into question many basic assumptions legislators, courts, and scholars harbor about election expenditures. Our findings implicate not only policy discussions about election funding, but the limitations of doctrinal interventions and judicial remedies that are divorced from issues of resource allocation.

The Article proceeds in five parts: Part I provides background on election funding, including why elections are expensive and what the most common funding sources are. This Part also discusses election law doctrines and how they do not directly consider election expenditures. Part II outlines our data and methods. Our novel methodological advancements permit us to draw conclusions that previous researchers have been unable to support. Part III presents our main findings. Part IV responds to the findings and explores potential doctrines under which election expenditures might be considered. Part V weighs the pros and cons of several non-doctrinal reform proposals.