The unfairness of the proposed California Presidential Election Reform Act is obvious: in a close election, the Act virtually assures that California's fifty-five electoral votes, which would be expected to go entirely to the Democratic presidential candidate under the traditional statewide-winner-take-all system, will instead be split, with more than a third of them going to the Republican candidate. Implementing this "reform" in the nation's largest Democratic state, but not in any of the large Republican states (like Texas), is roughly the equivalent of handing over to the Republicans the state of Illinois. What is less obvious is that the Act would be unfair and unwise even if it applied nationwide.
The Act embodies the "congressional-district system" for awarding electors to presidential candidates. That system gives the statewide popular-vote winner only two electors and allocates the remaining electors on a district-by-district basis, awarding each district's lone elector to whichever presidential candidate carries that particular district. This system sounds eminently fair to many casual observers, and it would likely have the benefit of encouraging presidential candidates to compete aggressively in a larger number of states. Indeed, even as sophisticated an observer of the political scene as Professor Bruce Cain of the University of California at Berkeley was recently quoted in the San Diego Union-Tribune as saying that, although the congressional-district system for awarding electors is "a horrible idea if it's applied only to California," the "idea itself is fine if it's applied to all states."
I believe the latter conclusion is wrong for two reasons. First, the congressional-district system not only would increase the chance that the Electoral College would generate the "wrong winner," that is, that the presidency would be awarded to the second-place finisher in popular votes, but it would do so in a way that is significantly biased to favor one political party. Second, the congressional-district system is founded on the erroneous assumption that congressional-district lines are politically "neutral" and thus well suited to functions other than electing members of the U.S. House of Representatives.
Contrary to popular belief, currently the Electoral College is not significantly biased to favor either political party. Some critics have noted that the four presidential candidates who have been denied the presidency despite winning a plurality of the popular vote nationwide-Al Gore in 2000, Grover Cleveland in 1888, Samuel Tilden in 1876, and Andrew Jackson in 1824-were all Democrats. But partisan voting patterns have changed enormously since the 19th century, rendering that point meaningless. And, even though it is true that in 2000 Gore lost the Electoral College after winning the popular tally by more than a half million votes, the 1996 and 2004 elections show that the system today is not consistently biased against Democrats. Had the nationwide popular vote been roughly tied in the 1996 and 2004 elections, Bill Clinton would have beaten Bob Dole by about twenty electors and John Kerry would have beaten George W. Bush by about thirty electors. So the same system that saved the Republican candidate in 2000 instead favored the Democratic candidates in 1996 and 2004, albeit by smaller and far less consequential amounts.
By contrast, under the nationwide application of the congressional-district system that Professor Cain and others have defended, all three Republican candidates (Dole in 1996 and Bush in 2000 and 2004) would have prevailed handily if the nationwide popular vote had been roughly tied. For example, assuming a nationwide tie vote under the congressional-district system, Bush would have won the Electoral College by about seventy-four and fifty-two electoral votes in 2000 and 2004, respectively. To see why this is so, it is useful to define partisan bias and then unpack its various components.
Partisan bias is generally defined as an electoral system's tendency to systematically favor one political party over another in translating popular votes into seats (whether in a legislature or in the Electoral College). An unbiased system treats the two major parties symmetrically. So if the two parties' presidential candidates have equal support in the national electorate, each should expect to win a roughly equal number of electors and to have a roughly equal chance to capture the presidency. Likewise, if a candidate succeeds in attracting support from well over half of the electorate, he should expect to be rewarded with more than half of the electors and thus with the presidency-regardless of the candidate's political party.
The congressional-district system would inject into the Electoral College a significant and consistent partisan slant because it combines "malapportionment bias" and "distributional bias." Under the traditional system, the presidential election effectively consists of fifty-one winner-take-all contests-one in each state, plus the District of Columbia. (For simplicity's sake, I'm ignoring the fact that two small states, Maine and Nebraska, have already adopted the congressional-district system.) Under a nationwide application of the congressional-district system, the contests would still be winner-take-all, but there would be 487 of them-fifty-one statewide contests for two electors apiece, plus 436 district-wide contests for one elector apiece (one in each of the 435 House districts and one in the District of Columbia). For the District of Columbia or any of the seven states that currently has only one representative in Congress (and thus only three electors), it is possible to re-conceptualize the contest as a statewide race for three electors, but that would not significantly alter the analysis presented here.
Malapportionment bias is the main problem with having fifty-one statewide contests for two electors apiece. This form of partisan bias comes from apportioning the same number of seats to a lightly populated area as to a heavily populated one. The classic example of a malapportioned legislative body is the U.S. Senate, where California and Wyoming each get two Senators, even though the former now has more than fifty times as many residents (and voters) as the latter. If one major party's political strength is located disproportionately in relatively large states, and the other party's strength is located disproportionately in the smaller states, the former party will be harmed by (and the latter party will profit from) a malapportionment bias.
Empirically, such malapportionment bias would exist today under the congressional-district system. Republicans, who tend to run well in rural areas, are stronger in small states, while Democrats, who tend to run well in urban areas, are stronger in large states. This explains why Bush could carry thirty and thirty-one states in 2000 and 2004, respectively, without winning a nationwide landslide: he won most of the smaller states, while his Democratic opponents won most of the larger states. The congressional-district system would effectively award the first 102 electors to the winners of the fifty-one statewide contests, regardless of population. In an election where the nationwide popular vote is roughly tied, that system would have given the Republicans something like a sixty-two to forty lead in electoral votes-even before any of the congressional-district-based electors were awarded.
The congressional-district system would be even more heavily biased when awarding the 436 district-based electors. The problem here is not malapportionment bias, as each district has, very roughly, the same total population. Rather, the problem is distributional bias. A party's support is more efficiently distributed if there are many districts that favor the party by only relatively narrow margins. Having lots of support in districts that favor the party by landslide margins is an inefficient distribution, as such districts waste votes that otherwise might have determined the outcomes in more competitive districts.
Again, there is no doubt about the current empirical situation: by any reasonable measure, the twenty-seven or twenty-eight least competitive congressional districts today are all Democratic districts. Most of them are urban districts located in the New York, Los Angeles, and Chicago metropolitan areas. And almost all of these politically lopsided districts are heavily African-American or Latino. Because these urban, predominantly minority districts soak up huge numbers of Democratic votes and very few Republican votes, the remaining 400-plus districts tilt significantly Republican in a close presidential election.
For example, under the congressional maps currently in place, Bush would have won 238 district-based electors to Gore's 198, even though Gore won a half million more popular votes. Had Bush actually tied Gore nationwide, his advantage would have risen to about 241 to 195. When that distributional bias is added to the sixty-two-to-forty malapportionment-bias-infused edge in electors chosen statewide, the Republican advantage in the Electoral College-again, assuming a roughly tied popular vote nationwide-is a whopping 303 to 235. That is a very substantial bias, roughly akin to conceding New York and Vermont to the Republicans before the campaign even begins.
Although the precise size of the partisan bias may change significantly from election to election, the underlying reasons for the bias in the congressional-district system run deep and are unlikely to disappear as voting patterns evolve. As redrawn after every federal decennial census, congressional-district lines are "political" in both intent and effect, and therefore they can never provide a truly "neutral" basis for awarding presidential electors.
At bottom, the Electoral College is a mechanism for tying control of the presidency to the voters' preferences. By contrast, our system for electing representatives to Congress is not merely a mechanism for effectuating voters' preferences about which political party should control the U.S. House of Representatives. For a term of 730 days, each representative is supposed to exercise his or her judgment about what is in the best interests not only of the nation but also of the representative's particular congressional district. An elector's job is far narrower-to cast a single vote, on a single occasion, for a single slate of presidential and vice-presidential candidates. Today, each elector is actively discouraged, and in some states expressly forbidden, from exercising personal judgment in casting that vote.
Relatedly, representatives are expected to "represent" their constituents not only substantively, in terms of sharing similar public-policy preferences, but also descriptively, in terms of sharing similar demographic and socioeconomic backgrounds. In a nation where non-Hispanic whites now constitute less than two-thirds of the population, having an all-white or nearly all-white "House of Representatives" would be worse than a misnomer; it would be an affront to our most cherished claims of democracy and equality. But while many engaged citizens know (and care about) who represents them in the House, almost no one is aware of who "represents" them in the Electoral College.
Congressional-district lines are drawn to address a variety of concerns, including ensuring fair representation of local communities and guaranteeing at least a minimally acceptable level of descriptive representation for minority citizens in the halls of our national legislature. Most of these concerns are utterly irrelevant to the Electoral College and its mission. Because congressional elections, and thus congressional-district lines, serve such a broad range of goals, they are unsuitable for choosing electors. Just to take one obvious example: viewed solely through the lens of Electoral College reform, the twenty-seven or twenty-eight most lopsided congressional districts in the nation are just a source of partisan bias, soaking up Democratic votes and tilting the electoral playing field to favor Republican presidential candidates. But viewed through the lens of the congressional elections for which these districts were created, they are a source of great diversity in the House of Representatives-racially, ethnically, experientially, and ideologically.
As other contributors to this symposium surely will note, there are plenty of solid reasons to fault our current system of electing the President of the United States. But the current system is not so badly broken that any reform would be an improvement. If adopted only by the people of California, the congressional-district system for awarding presidential electors would be grossly unfair. But even if adopted uniformly in all states, the system would inject a new source of partisan bias into our political process and would be a major step backwards for our democracy.