Division of the Humanities and Social Sciences California Institute of Technology Pasadena, California 91125 Empirically Evaluating the Electoral College

The 2000 U.S. presidential election has once again rekindled interest in possible electoral reform including the possible elimination of the Electoral College. Most arguments against the Electoral College have either been based on ancedotal evidence from particular elections or on highly stylized formal models We take a very different approach here. We develop a set of statistical models based on historical election results to evaluate the Electoral College as it has performed in practice. Thus, while we do not directly address the normative question of the value of the U.S. Electoral College, this paper does provide the necessary tools and evidence to make such an evaluation. We show that when one preforms such an analysis there is not much basis to argue for reforming the Electoral College. We first show that while the Electoral College may once have been biased against the Democrats, given the current distribution of voters, neither party is advantaged by the system. Further, the electoral vote will differ from the popular vote will only when the average votes shares are very close to a half. We then show that while there has been much temporal variation in voting power over the last several decades, the voting power of individual citizens would not likely increase under a popular vote system of electing the president. The 2000 U.S. presidential election has once again rekindled interest in possible electoral reform. While most of the popular and academic accounts have focused on balloting irregularities in Florida, such as the now infamous “butterfly” ballot and mishandled absentee ballots, some have also noted that this election marked only the fourth time in history that the candidate with a plurality of the popular vote did not also win the Electoral College. This “anti-democratic” outcome has fueled desire for reform or even outright elimination the Electoral College. Division of the Humanities and Social Sciences, D.H.S.S. 228-77, Pasadena, CA 91125; [email protected], http://jkatz.caltech.edu. Department of Statistics and Department of Political Science, Columbia University, New York, NY 10027; [email protected], http://www.stat.columbia.edu/∼gelman. Center for Basic Research in the Social Sciences, 34 Kirkland Street, Harvard University, Cambridge MA 02138; http://gking.harvard.edu, [email protected]. This is not the first time that such controversy has surrounded the Electoral College system. Perhaps the most scandalous presidential election in U.S. history was the 1876 election between Samuel J. Tilden, a Democrat, and Rutherford B. Hayes, a Republican. The nation was deeply divided; caused partly by a deep economic recession as well as a seemingly endless number of scandals involving graft and corruption in the incumbent Republican administration of Grant. Making matters more divisive, there were also a number of third parties that contested the election. Similarly to the 2000 elections, the outcome hinged on resolving potential vote count problems in Florida, Louisiana, and South Carolina. These states were so divided, as was the rest of the country, that they sent two slates of electors each to Congress—one set for Tilden and one set for Hayes. The Congressional procedures for resolving the disputed set of electors had expired, therefore Congress established a 15 member commission to decide the issue of which set of electors to use. After much intrigue, the commission narrowly voted to use the electors for Hayes from all three disputed states thereby giving the election to him. Hayes won the election despite that the consensus was that Tilden had won 51% of the popular vote to Hayes’ 48%. Another case in which the Electoral College vote went contrary to the popular vote was the 1888 election between the incumbent President Grover Cleveland, and his Republican challenger, Benjamin Harrison. Cleveland garnered huge majorities in the 18 states that supported him, whereas Harrison won slender majorities in some the larger states that supported him. In the end, Cleveland won the popular vote by about 110,000 votes. constituting less than one percent, but lost the Electoral College. The other election in which the popular-vote leader did not become President was in 1824, when Andrew Jackson won a plurality of both the popular and electoral votes. But, because Jackson did not win an electoral-vote majority, the election was decided by the House of Representatives, which voted for John Quincy Adams. The popular vote was even closer in the 2000 election. Gore won the popular vote by approximately 541,000 votes, or about about half a percentage point of the total vote cast but, as with both Cleveland and Tilden before him, lost the Electoral College. Most arguments against the Electoral College have either been based on these particular elections or on highly stylized formal models (see, for example, Banzhaf 1968). We take a very different approach here. We develop a set of statistical models based on historical election results to evaluate the Electoral College as it has performed in practice. Thus, while we do not directly address the normative question of the value of the U.S. Electoral College, this paper does provide the necessary tools and evidence to make such an evaluation. There are two fundamental ways that the Electoral College could potentially be flawed. First, it may be biased in favor of one party. That is, the distribution of votes could have a party’s candidate winning the popular vote but losing the Electoral College. For example, if it is likely that the Democratic candidate is to win with overwhelming majorities in a few states, then this will boost the overall Democratic vote share but