H/t to Sean Parnell for the new paper by Ron Rivest and Philip Stark:
National Popular Vote (NPV) Interstate Compact [Koza et al., 2024] attempts to provide direct presidential elections in the U.S. without a constitutional amendment, through a binding agreement among “member” states comprising a majority in the Electoral College. The compact is intended to ensure that a majority of Electoral College votes go to the winner of the national popular vote. It does not succeed. The “national popular winner” has no objectively fair or correct definition if any state uses a non-plurality method (such as ranked-choice voting) for presidential elections, as two states currently do. NPV does not demand any particular method for states to determine tallies for nonplurality voting methods. NPV requires member states to accept as conclusive and correct the reported vote tallies in every state-including states that are not members of the compact. It does not require evidence that reported tallies are accurate, does not provide a way for member states to demand such evidence, and does not provide any remedy even if state-level results are untrustworthy or absurd. Even the best current state-level audits do not provide evidence that state totals are (approximately) correct. Auditing NPV would require sweeping changes to state election administration and federal legislation that ensures coordination among states. Because of these faults, any individual state, whether a party to the compact or not, could alter the outcome of the presidential election through error or malfeasance-or simply by following state law. The NPV compact undermines the trustworthiness of U.S. elections. NPV is a bad idea unless every state is required to use plurality voting and report those votes accurately in their Certificate of Ascertainment (we call this a simple direct election), has a trustworthy, organized, physically inventoried paper trail of votes and a rigorous canvass; and there is a federal requirement to conduct a rigorous, binding risk-limiting audit (at the national level) of the outcome of the presidential contest. For the foreseeable future, adopting NPV is worse than doing nothing.