« Colorado About-Face? | Main | Republican Scum Compare Ginny to Hezbollah »

Tuesday, October 26, 2004

Will the Electoral College Matter This Year?

Posted by DavidNYC

When I started this site a year ago (in fact, the one-year anniversary was last Tuesday, and I forgot all about it like a typical male!), it was out of a desire to learn more about the swing states. But my reason for wanting to was predicated on what may be a flawed premise: That the electoral college will once again be important this year. I say this because most analysts, including people like Charlie Cook, say that if the election is decided by a wide enough margin in the popular vote, the EC just won't matter.

And historically speaking, this has been true. People like to say that "every electoral vote counts," but even in the closest of elections, usually only the very biggest individual states were capable of swinging the election - and that was incredibly rare. Looking back at the closest presidential races of all time (by popular vote), this is the list I came up with:

�Ģ 1976: Carter wins the popular vote by 2% and is 27 EVs over the margin he needs
�Ģ 1968: Nixon wins by .7% and 31 EVs
�Ģ 1960: Kennedy wins by .17% and 34 EVs
�Ģ 1916: Wilson wins by 3% and 11 EVs

It doesn't always make sense to look back at the 19th century when comparing presidential elections, but I think we can do it safely here. The period of 1876 to 1892 saw an amazing string of five very close elections:

�Ģ 1892: Cleveland wins by 3% and 55 EVs
�Ģ 1888: Harrison loses by .85% but wins by 32 EVs
�Ģ 1884: Cleveland wins by .25% and 18 EVs
�Ģ 1880: Garfield wins by .02% and 29 EVs

When you finally get back to the infamous year of 1876, that's when you have an election as close in the EC as 2000 was:

�Ģ 1876: Hayes loses by 3% but wins by 1 EV

Any earlier than this, and things get too wacky. So where does this leave us? Well, as you might guess from the EV margins, not many single states could have ever made the difference. And in fact, this is correct. In order to swing the election, you need a state with the EV margin + 1, so these are the only states which could have done it in the 20th century (I'm only including states that the winner won):

�Ģ 1976: NY-41
�Ģ 1968: CA-40
�Ģ 1960: None
�Ģ 1916: AL-12, CA-13, GA-14, KY-13, MO-18, OH-24, TN-12, TX-20, VA-12

By Swing State Project standards, some of these states were close, but not many. Carter won NY by 4.5% in 1976. Nixon won CA by 3% in in 1968, but had he lost there, Humphrey would not have won - the race would have been thrown to the House because of Wallace's 46 EVs. In 1960, as in 1976, California theoretically could have provided the margin - but the loser won it both times.

The election of 1916 initially looks promising, and probably has the most similarities to 2000. There were many states that had 12 or more EVs, but most were won by Wilson. Hughes, like Gore, kept the race close by winning a much small number of big states. Nonetheless, there were still nine states which, had they flipped, would have made Wilson a one-term president. (And kept us out of WWI, and allowed Germany to conquer Europe... and I'll leave the alternate history for another time.) Wilson took KY by 5.5%, MO by a little over 3.5%, and OH by over 7.5%. The other Southern states were all Democratic landslides, particularly in the Deep South.

The one real nail-biter was, quite famously, California, which went to Wilson by .38%, or less than 4,000 votes. So that's precisely one election in the last century where a single state could have changed the outcome - where every really electoral vote did matter - prior to 2000.

The rarity of such an outcome makes me think we are very unlikely to see something like this happen again in 2004. I think the electoral college might matter, but it probably won't. What do you think?

(Thanks to PBJ Diddy for additional research on this subject. I also consulted Dave Leip's Atlas and Jim Howard's EC Calculator extensively.)

UPDATE: Reader Jonathan Katz, a political science professor at CalTech, sent me a copy of a paper he co-wrote empirically evaluating the electoral college. His conclusions - arrived at much more rigorously - are essentially the same as mine. You can read the paper by clicking here (PDF).

Posted at 02:59 PM in General | Technorati


One thing to consider is that modern methods of campaigning may make it structurally more likely that near-misses happen. Campaigning in the age of mass media is quite powerful. Couple that with the fact that a close win is as good as a win. If an underdog was barely behind in the EV, but had three states with a shot to win, with heavy campaigning, it would be foolish to divide the resources among all three, with a chance to win by many EVs, when the presidency would come with a win in any one of them (more likely if resources were concentrated there).

2000/2004 may be more different from 1996 than 1996 was from 1896... The machinery behind campaigns is far more impressive, IMO. Or, if not "1996", try "1960".

Posted by: John at October 26, 2004 03:21 PM | Permalink | Edit Comment | Delete Comment

Kerry hits 50% on Washingon Post tracking poll.

Kerry momentum continues today:

Today (Yesterday)
Zogby -3 (-3)
Rasmussen 0 (+2)
TIPP -5 (-7)
Washington Post +2 (+1)

Average -1.5 (-1.8)

Posted by: DFuller at October 26, 2004 05:06 PM | Permalink | Edit Comment | Delete Comment

I agree that a divergence in the electoral-popular outcomes (i.e. going in opposite directions as in 2000) is highly unlikely, based on probability theory and on the fact that there is a cetain amount of randomness in each state's outcome. One analogy would be football -- I'd bet that if you ranked teams by point differentials (points for - points against) that would closely match the ranking of teams by win-loss record by the end of the season (though probably not 4-5 games into the season).

However, I believe that the increased ability for advertising and campaigning efforts to be so much more geographically targeted now than in the past changes this dynamic. I think it would INCREASE the likelihood of a divergence -- but I'm not sure about that. My rationale is that a "national tide" of opinion would be less of a factor and state-specific opinion would be more of a factor.

Posted by: Jason at October 26, 2004 05:48 PM | Permalink | Edit Comment | Delete Comment

there was an interesting article in the usa today about the election.they say if people vote like they did in 2000, bush will win but if the new voters show up he is in trouble.new voters favor kerry 59-40. so bush could be toast , we just won`t know until next week.

Posted by: joel at October 26, 2004 06:18 PM | Permalink | Edit Comment | Delete Comment


That is why the Republicans are challenging 35,000 new voter registrations in Ohio. They know that if everyone votes, they will lose.

Posted by: DFuller at October 26, 2004 08:43 PM | Permalink | Edit Comment | Delete Comment

Here in Texas, early voting has been heavy, and it seems like there are more reports of Kerry votes than you would expect. Bush will carry Texas , but it will be closer than we anticipate, probably 54 /46 ..but the vote count will be impressive.

My prediction for Tuesday results:

Kerry 295
Bush 243

From 2000, Kerry picks up FL, OH, WVA, and NH
Bush picks up IA, MN and HI

We shall see !

Posted by: R Bobbitt at October 26, 2004 11:15 PM | Permalink | Edit Comment | Delete Comment

Nice work David!

I also agree that's unlikely to happen again this year, or in any particular year, for that matter, but there is still a pretty good chance that it could happen again. Correct me if I'm wrong, but I think historically, in most elections, there were few states that were not a landslide for one candidate or the other. These days there seem to be many more states where the race is close or nearly evenly split. I think that this would increase the possibility that the electoral college would make a difference.

Posted by: rob at October 27, 2004 04:36 AM | Permalink | Edit Comment | Delete Comment

This is an excellent website that is often discussed among the US expats, (including myself) living in Scotland. I just wanted to point out confusing point in this post. Correct me if I am wrong, but the national popular vote for President is little more than an interesting fact we like to know. Presidents win or lose by the electoral college alone, the popular vote does not matter per se, but it usually follows the popular result. I therefore find your reports of past elections (e.g.,1968: Nixon wins by .7% and 31 EVs) to be misleading. Nixon didn't "win" the popular vote by .7%, it's just an intersting number. Sorry to be pedantic, but I spend half my days now explaining US federalism and the electoral college to Brits who have been misinformed by continual media references to the national polls statistics. I keep saying, nation-wide statistics are only indicators, in the end they do not oficially mean anything.

Posted by: L_dog at October 27, 2004 05:49 AM | Permalink | Edit Comment | Delete Comment


I like the way you put this. "[T]he national popular vote for President is little more than an interesting fact we like to know"

That's not the way that I have ever looked at it but I think it is absolutely true. The popular vote doesn't have any direct effect on who is elected in and of itself but many Americans do look at popular vote to decide if the President has a 'mandate' when he takes.

Many people believe that the popular vote winner vote winner should be elected. (I am actually one of the few who actually disagrees with this).

Once again, well put.

Posted by: Dan Hogan at October 27, 2004 09:28 AM | Permalink | Edit Comment | Delete Comment

The problem about switching to the popular vote would be that the election would have to be federalized and it would be a nightmare. The federal government would have to set up a set of federal election laws. It would also have to set up universal standards for voter registration, early voting, and absentee voting because every voter in the country would have to be on equal ground. It would be unfair to voters a state that didn���t have early voting if Texas allows early voting for two weeks prior to the election.

To some extent, the federal government has already intruded in the right of the state to run their own election. The federal government requires states to count overseas military votes as long as they are mailed by Election Day. Many states require that absentee votes be received, not mailed, by Election Day. The federal law is unconstitutional under a couple of grounds:

A) The Constitution specifically gives state legislators the right to set state election law.
B) The law violates Equal Protection because it excludes overseas non-military absentee ballots. It puts one group of voters above another.

Note: I believe that our country should make every effort to count overseas ballots (military AND non-military). Every voter needs to get their say in the election and every vote should be counted. However, I do not believe in a election that one group of voters should be placed above another.

Posted by: DFuller at October 27, 2004 10:06 AM | Permalink | Edit Comment | Delete Comment

Yesterday on the Oreilly Factor Tony Snow said that the Kerry campaign was pulling out of Florida. Can you believe this sh%t. They are trying to discourage Democrates in Florida to not vote and stay home. Is this true because if so I tell you the polls by ARG and SUSA say they should stick around.

Posted by: godfrey at October 27, 2004 11:19 AM | Permalink | Edit Comment | Delete Comment

L_dog: Sorry for any confusion. Nixon did "win" the popular vote, but you are right - it didn't win him anything. To be super-specific, you could say, "Nixon carried the popular vote by a .7% margin."

But I disagree that the popular vote is simply an interesting number that "we like to know," precisely for the reason you point out: Electoral college results almost always track with the national popular vote totals. So those numbers have a lot of meaning. (You should read Prof. Katz's paper appended to the end of the post for more.)

DFuller: I totally disagree that a national federal election would be a "nightmare." I've been working in the area of election protection this campaign season, and let me tell you, the biggest obstacle to protecting the franchise is that we have 50 different legal systems, 50 different voter registration systems, a bajillion kinds of voting machines, and so forth. Uniform standards would eliminate this problem.

Also, people say that if we had a national popular vote, we would have had half a dozen Floridas on our hand last time. But that's not true: Al Gore won by half a million votes. I don't put anything past the GOP, but there's really no way they could contend that a nationwide recount would find another half a million votes for them. That's a full half percentage point.

Posted by: DavidNYC at October 27, 2004 11:32 AM | Permalink | Edit Comment | Delete Comment

The O'Reilly factor is just an opinion show. It doesn't even claim to have the minor journalistic ethics (pardon me for using the term ethics about Fox) of the regular Fox News operation.

Edwards is in FL today, KErry is there on Friday, Clinton is going there. No way the Kerry camp is abandoning Fl.

But you know what -- Bush seems to have largely conceded PA.

Posted by: erg at October 27, 2004 11:37 AM | Permalink | Edit Comment | Delete Comment

There is a slight error in your math. For instance, in 1976 when Carter won by 27 EVs, the Republican candidate (I was five and didn't vote that year..) would have only needed to win a state that included 14 EVs to swing the total in his favor. By changeing a state with 14 EVs from Dem to Rep, Carter would have lost -14 EVs and the Rep would have won +14EVs, thus a difference of 28. Thus much smaller states can "make a difference" than your math suggests.

Posted by: Shane Carbonneau at October 27, 2004 01:23 PM | Permalink | Edit Comment | Delete Comment

Shane, actually, the math is right. Carter didn't "win by 27 EVs" over Ford - he won by 27 EVs over the total he needed, which was 270. The actual Carter margin over Ford was 57.

Posted by: DavidNYC at October 27, 2004 02:32 PM | Permalink | Edit Comment | Delete Comment

If one uses the presidential race individual state futures prices @ tradesports.com, the data can easily translate into probabilities. For example, the current AZ price for Bush is 90 meaning that Bush has a 90% chance of winning AZ. Using all these individual state current probabilities, one can run a Monte Carlo simulation. In such a simulation, there is

Posted by: JMP at October 27, 2004 03:52 PM | Permalink | Edit Comment | Delete Comment

JMP, I like your Monte Carlo simulation, but note that the probabilities are not independent. The TradeSports lines give Bush a small chance, say 3%, of losing a state like North Dakota. A tossup like Iowa is more like 50-50. If the probabilities were independent, then Bush could plausibly lose North Dakota but win Iowa. In reality, anything that led to Bush losing North Dakota would almost certainly mean that he also lose Iowa (it would have to be an incredible scandal).

It's more the case that only a handful of states are realistically up for grabs, and if the states like North Dakota on the one hand or Massachusetts on the other swung, then I guarantee you -- there would be no electoral college tie (wink) -- it'd be an incredible blowout.

The particular scenarios leading to a tie seem relatively possible (though still way under 50-50, for sure) this year. In another year, a particular medley of swing states might make such an outcome impossible by any combination.

The possibility of pop/electoral discrepancies is something best viewed as a historical condition that could arise or not. The country can be, in effect, gerrymandered by shifting politics and shifting demographics, without anyone intentionally crafting it. If a party, hypothetically, came to dominate the states that have disproportionately more EVs than population (small states), then an EV/pop/ vote discrepancy could become not only possible but the norm. That is, if the Western states, the Plains states, Hawaii, Alaska, and New England ever became all of a like mind. If a conservative movement swept New England (small states), while liberal viewpoints gained in the South (on average, some bigger states), the conservatives could win election after election without being close to a popular vote advantage. Maybe in 100 years...

Posted by: John at October 27, 2004 04:50 PM | Permalink | Edit Comment | Delete Comment

I really do think that if there is an event that affects the likelihood of Bush winning, the "odds" would change immediately at tradesports.com (either across all states, individually or regionally - depending on the event). Tradesports (and similar futures markets) are simply a very good indicator of where things stand now--probably better than the polls. There are some state-specific issues. For example, in NV the controversy re Yucca Mountain gives Kerry a tradespot 23% chance of winning (higher than some pundits would think). There are also regional issues (e.g., "rustbelt" states - jobs overseas, worries about pensions). I think "the market" factors all of this into its futures pricing. The futures markets "swing states" are CO, FL, IA, NM, OH, MN, WI defined as either Bush or Kerry having a >30% chance of winning.

Posted by: jmp at October 27, 2004 06:00 PM | Permalink | Edit Comment | Delete Comment

Democrats should know this and be pushing the media to cover Michael Badnarik. Read now on Electoral-Vote: ���A Rasmussen poll taken Oct. 26 in Arizona puts Libertarian party candidate Michael Badnarik at 3%. When the pollsters actually ask about him, he does surprisingly well. He might end up canceling out the Nader factor by appealing to disgruntled Republicans who support a balanced budget and small government and are appalled by the current deficit and power the Patriot Act gives the government to snoop on people's lives.���

Posted by: VoteBadnarik at October 28, 2004 10:03 PM | Permalink | Edit Comment | Delete Comment