There's been some discussion recently about the possibility that Barack Obama wins the popular vote and loses the electoral college; see Mark Ambinder for a pretty good articulation thereof. However, our simulation model thinks that this is actually a bigger worry for John McCain. It assigns McCain about a 5 percent chance to win the popular vote and lose the election, to Obama's 3 percent chance. So what gives?
There are a lot of different ways to approach this problem -- but there is one fundamental that should not be ignored. Namely, the allocation of electoral votes lags behind changes in the distribution of the population. Presently, the composition of the electoral college is based on the 2000 census -- what if it were based on the population in 2008 instead?
The Census Bureau does not yet have its current population estimate out for 2008. What we can do, however, is take its 2007 estimate, and then add to it the population gain between 2006 and 2007 to come up with a reasonable estimate for 2008. For example, Texas had 23.5 million people in 2006 and 23.9 million in 2007 -- a gain of about 400,000 persons. We add another 400,000 to account for population growth between 2007 and 2008, which gives us an estimate of 24.3 million for its current population.
If we do this for each state, and then reassign electoral votes based on the new population estimates, I show the following electoral votes changing hands:- Texas gains three electoral votes.
As you can see, the gains tend to come in Republican-leaning states, and the losses in Democratic-leaning ones. We can be a little bit more specific about this by apportioning the electoral vote changes based on McCain and Obama's win percentages in each state. For example, we estimate that McCain has a 95 percent chance of winning Texas, so we assign McCain 95 percent of Texas' three new electoral votes (+2.85), giving Obama the small remainder of 0.15 electoral votes.
- Arizona, Florida, Georgia, Nevada, South Carolina and Utah each gain one electoral vote.
- Iowa, Louisiana, Massachusetts, Michigan, New Jersey, New York and Pennsylvania each lose one electoral vote.
- Ohio loses two electoral votes.
Overall, this results in a swing of slightly fewer than five electoral votes from McCain to Obama (actually 4.68), or a net change of 9-10. If electoral votes were assigned based on the current population, rather than the population in 2000, we would show Obama with tiny a 269.7-268.3 advantage, rather than the 274.4-263.6 advantage that we do now.
These new voters places like Texas and Arizona will show up in McCain's popular vote column -- but not in his electoral vote column until new allocations are made after the 2010 census. While there might be other reasons that the electoral math tends to be favorable to McCain, this starts him off with about a 10-EV handicap, which is pretty significant as far as these things go.
Saturday, June 7, 2008
How the Electoral College Hurts the GOP
-- Nate Silver at 6:35 PM
Labels: electoral math, popular vote
64 comments
Nate,
Doesn't this assume that the transient population that moves from a Dem leaning state to a GOP leaning state will vote more like their new state?
There are certainly other dimensions to this problem (some of which I hope to expose more fully in subsequent posts). But fundamentally, it means that a vote in Texas is worth less than one in Massachusetts, because you have the same number of electoral votes in Texas representing a larger population. Below is the number of electoral votes per resident based on my 2008 population estimates; you generally have a preponderance of GOP-leaning states toward the top of the list.
.Texas 759,406
.Florida 681,948
.California 666,343
.Arizona 651,119
.Georgia 648,371
.New York 622,235
.North Carolina 617,704
.Illinois 613,006
.Virginia 598,561
.Washington 594,641
.Pennsylvania 591,665
.Michigan 591,059
.Indiana 579,733
.New Jersey 576,485
.Ohio 572,791
.Tennessee 570,421
.Wisconsin 564,677
.Maryland 562,096
.South Carolina 561,771
.Colorado 552,184
.Utah 548,119
.Oregon 542,022
.Massachusetts 538,526
.Missouri 537,647
.Kentucky 534,609
.Nevada 527,047
.Minnesota 522,814
.Oklahoma 522,203
.Alabama 517,408
.Connecticut 499,973
.Mississippi 487,838
.Louisiana 477,627
.Arkansas 476,454
.Kansas 464,653
.Iowa 427,715
.New Mexico 397,046
.Idaho 383,085
.West Virginia 361,120
.Nebraska 356,162
.New Hampshire 329,190
.Maine 328,210
.Montana 323,697
.Hawaii 320,320
.Delaware 292,017
.South Dakota 270,170
.Rhode Island 262,014
.Alaska 232,301
.North Dakota 214,521
.Vermont 206,200
.District of Columbia 198,351
.Wyoming 176,885
Huh. All the 2010 reapportionment estimates I've read have said that North Carolina was going to gain a House seat. I'd never heard anything about South Carolina gaining one. Interesting.
Bedir,
There's no really assumption about the voting patterns of the new votes inherent in this model, because the states new voters are moving to tend to be solid red states, not swing states. In another decade the demographic shift could put someplace like Texas in play, but right now it's still solidly red. So a population shift effectively trades surplus victory margin for extra electoral votes.
This kinda proves the Democrats really need to get to work down in Texas to make it more of a swing state. I think it should replace Ohio as an important swing state. The keys are to activate and register the latino population and actually build a party in the state.
This year's election of Noriega should be interesting.
I'm just an undergrad student of political science, but I've found that the current system usually gives an advantage to those who win smaller, rural areas (conservatives), at least when compared to everyone's knee-jerk 'solution' to the Electoral College: national popular vote, which would then just give an advantage to those who win larger, urban areas (liberals).
The true solution is to change the composition of the House of Representatives using the Wyoming Rule: basing the number of voters per Representative on the population of the lowest state - take the population of state X, divide it by the population of the state with the fewest people (Wyoming), and give state X an amount of Representatives equal to the number you get, rounding up if it has a decimal larger than .5. Then you must allocate Electoral Votes proportionately amongst candidates who reach the minimum percentage of votes in a state - that minimum percentage being the percent necessary to earn at least one electoral vote in that state.
Might sound confusing, but it's actually pretty simple. I once laid out the 2000 Presidential Election (which would have required 324 Electoral Votes to win) on a spreadsheet along these lines, and it ended up Bush 323, Gore 320, Nader 4.
It's about as accurate AND fair as you can get - doesn't tilt to one political ideology, and it makes sure that every vote actually counts.
...That's just my two cents. >_>
Michael said: "The true solution is to change the composition of the House of Representatives using the Wyoming Rule: basing the number of voters per Representative on the population of the lowest state..."
The "true" solution? Actually, an obscure but interesting thread of debate in US history relates to the apportionment of House seats to states.
The Constitution is vague on the matter, in particular what to do with fractions, and luminaries as prominent as Jefferson, Hamilton, Madison, Webster and others weighed in on what rounding rule was "best."
Modern political scientists have proved (using an "impossibility theorem") the futility of searching for a perfect apportionment system, because there always will be the possibility of inconsistencies (weird behavior as populations shift, basically).
Google "Peyton Young" and "apportionment" and you'll find his good survey of the history of this topic.
www.census.gov/history/pdf/Fairness_in_Apportionment_Young.pdf
I'm pretty sure Michael's "Wyoming Rule" basically just works out to mirror the popular vote, but introducing a whole lot of rounding error -- similar to what the delegate allocation rules in Democratic primaries do. But there's a good reason to do that in the primaries, namely that delegates are actual people and can't be divided arbitrarily.
Maybe can explain what I'm missing, but I still do not understand why Nate uses proportional allocation of electoral votes for his projections. How does Obama have 274.4 electoral votes to 263.6 for McCain?
I'm probably missing the obvious but the analysis above used the same logic.
Thanks
If we are going to talk hypotheticals I would prefer a version of the Electoral College where all states are allocated as Maine and Nebreska thereby reducing the impact of the Electoral College, but still protecting the small states somewhat.
I think it's wrong to believe that the electoral college leads to conservative overrepresentation. Vermont, New Hampshire, Rhode Island, and Washington, D.C. are overrepresented. So are Alaska, the Dakotas, and Wyoming. Seems like a tie on the small-state front.
On the big-state front, Texas, Florida, and California are underrepresented. Tie there, too.
I absolutely agree with Anon above about getting to work in Texas. The Republican hostility toward Latinos presents real opportunities there and in Arizona, along with some other (more surprising) places like Iowa, Colorado, and Georgia. Also, the high level of unionization in the casino industry presents opportunities in Nevada, which has turned into very much of a swing state.
Nate, there are at least two other aspects of this worthy of exploration in due course.
First, as you note, there will be a reapportionment after the 2010 census, i.e., effective with the 2012 elections. Partisan control of governorships and state legislatures during 2010-2011 would make a big difference in how the district lines are drawn/gerrymandered. (IIRC, the Census Bureau sends preliminary control figures to the states by the end of the census year.)
Second, although of course we are interested in how migration affects the statewide number of congressional seats and electoral votes, we are also interested in the partisan makeup or competitiveness of the CD's, and generally speaking the effect of current migration trends is toward dilution of the GOP majority in the states and in the creation of more districts that are 'at risk' of going Democratic. For example, in Virginia, even if there is no gain in House seats, the growth of northern VA suburbs has contributed to a Democratic swing both in the northern parts and the state as a whole.
FYI, perhaps the leading expert on regional migration trends in the U.S. is the demographer William Frey. You might want to Google his name + migration. Also, here's his website: http://www.frey-demographer.org/
Brad - the decimals arise because the EV projections are averages over a large numbers of simulations.
we are also interested in the partisan makeup or competitiveness of the CD's, and generally speaking the effect of current migration trends is toward dilution of the GOP majority in the states and in the creation of more districts that are 'at risk' of going Democratic. For example, in Virginia, even if there is no gain in House seats, the growth of northern VA suburbs has contributed to a Democratic swing both in the northern parts and the state as a whole.
I could easily be persuaded otherwise, but for the moment I have to suspect that the redistricting process i