My Election Predictions: They Were Pretty Awesome

I made a prediction in all 51 Presidential state-level contests and all 33 Senate contests that were held last night. Of those 84 races, I was correct in 82 of them, including going 51-for-51 at the Presidential level. I correctly called 332 Obama, 206 Romney as the final electoral tally, with Obama winning all his 2008 states except Indiana, North Carolina, and Omaha's single elector, but winning everything else, including Florida. I also correctly predicted 55 Democratic Senators, assuming Angus King does the obvious thing regarding caucus choice. My only errors were in North Dakota, where I was honestly a bit surprised that Heidi Heitkamp managed to pull it out, and Nevada, where I thought the legendary Democratic turnout machine would deliver a victory for Shelley Berkley. I was wrong as to both of those, but they canceled out in terms of overall partisan balance.

My predictions compare very well to Nate Silver's. Like me, he nailed the Presidential race, even having Florida as pretty much a tie with Obama the narrow victor. Like me, he got two Senate races wrong, both in the west. Unlike me, however, he did not call the final Senate balance of power correctly, because his errors did not cancel out; he had Heitkamp losing in ND, but also had Jon Tester losing re-election in Montana. I had observed a while back that ND and MT were two very sparsely-polled Senate contests, and the FiveThirtyEight model was giving an awful lot of weight to the "state fundamentals" in them. In other words, Nate's model was assuming that the state partisan gravity would assert itself rather strongly against polling that suggested really close races. I was always a bit skeptical that it would, and indeed it didn't. In fact, the reason for thinking Berkley would win in NV was that the state's partisan gravity would outweigh the rather substantial volume of polling in that state. I'm a bit embarrassed, in retrospect, to have bought into that, although the fact that it has happened before in that state makes it a bit more understandable. Overall, however, Nate and I and really anyone who was willing to take the polls at face value did really well in forecasting the state-level contests.

One thing to mention, by way of tooting my own horn, is that I've been saying for at least ten or eleven months now that I thought the Senate races would go very well. Specifically, since Olympia Snowe's retirement I've been pretty damn confident we'd hold the Senate, and even before than I thought it was pretty likely. So I've been correctly predicting what happened last night since a time when pretty much everyone thought we were going to lose control of the chamber. I don't remember if I wrote a blog post to that effect, but I definitely have been saying it in person to my friends and family. Part of why is that the Republican tendency to throw away multiple winnable races per cycle has just become, in my opinion, predictable at this point, so I was always sort of assuming from the start that stuff like that which occurred in Indiana and Missouri would happen somewhere. And I enter the 2014 prediction season, which I admit is looking to be a bloodbath at the Senate level, with that same assumption in mind. I don't know where the Republicans will cough up winnable seats, but I'm pretty sure that they will do it somewhere.

Another thing I'd say is that my method seems to have been pretty well vindicated. That method, specifically, is to get all my polling data from looking at what PublicPolicyPolling says, and what FiveThirtyEight says. Both sources aced the election; many other sources bungled it rather badly. PPP are just clearly the best in the polling game right now, and Nate does the best job I know of at taking the vast bulk of less-awesome polling data and turning it into a useful commodity. I plan on continuing this method going forward, and I expect to continue being vindicated in it.

As for the House...? Well, I'm a bit surprised that Democratic gains were that meager, or maybe disappointed is a better word than surprised, because this was always what people were saying. It looks right now like we might get circa 200 seats when all's said and done, leaving us with a further 18 to take back the House in future elections.

But on the whole, given that my predictions were pretty good ones, it's good to be right.

Forthcoming later today will be a piece of general reactions to the election, and then a piece previewing the next couple of years, both of which could be very long. It's snowing right now, just the perfect little icing (literally!) on the cake of the past 24 hours, so I might not get to them for a while; staring out the window at the snow takes precedence. But they will get written.

Jim Cramer's Crazy Election Prediction

So, it has achieved a fair bit of notoriety that Jim Cramer predicted that Obama would win 440 electoral votes. This is pretty patently ridiculous. How ridiculous, you ask? Well, I've taken the 538 state popular vote estimates from a few days ago (when they provided a convenient chart of that figure for all 50 states in one place), and given Obama states in descending order of his margin in them until he hit 440. Here's what the map looks like:

That's, um, ridiculous. It's also 451 electoral votes, which is not 440. In fact, I can't see any good way to hit 440 on the dot. You could subtract an 11, like Arizona or Indiana, but c'mon, we're giving Obama Texas, there's no way AZ/IN are marginal states here. Or you could take away South Carolina and one of the 3's, probably one of the Dakotas, but that's -12, not -11, giving 439. I suppose you could add the next state on the list, West Virginia, which gives an extra 5, and then try to subtract 16, probably just by taking away Georgia. That's 440, but the world in which Obama wins WV but not GA is a really weird one.

Although I suppose that's not such a disqualifier, since we're talking about a world in which he wins Texas, so "weird" is kind of a given. The point here is really that it's clear beyond a reasonable doubt that Jim Cramer did not have a map of states won and states lost in mind when he made his prediction. He couldn't have. Not only does the magnitude of his ridiculousness become apparent when you have to draw up a map, actually getting to that specific number is really, really hard. This is not so much a prediction, in other words, as a blancmange--no, wait, that's something different. Not so much a prediction as an expression of the sense that Obama is going to win, probably by a decent margin, but with no idea what that actually means.

I should add that there were times during this campaign when I thought a map like that one might be possible. If, say, Rick Perry or Michele Bachmann or Herman Cain had been the nominee. Maybe if Newt had won the primary, though I'm a bit skeptical that such an enormous landslide would've materialized. In any event, we've had six close elections in a row now, with neither candidate getting 400 electoral votes unless Mr. Cramer knows something the rest of us really don't. That's weird. As best I can tell it's only happened once before that six consecutive elections have been highly competitive, in roughly the 1876 through 1896 period. I have a feeling it has to end eventually, though I'm not sure what will make that happen. But I think the above map is an interesting one to look at as a possible configuration of a latent Democratic landslide. It would be a really fun map.

The Smallest Winning Map Ever

A few days ago, Matt Yglesias put up a blog post in which he came up with the smallest possible winning electoral college map under the current vote distribution. He calculated this by sorting the 51 voting areas (i.e., states plus D.C.) in descending order by population density, and tacking states onto the map in that order until he had 270 electoral votes. D.C. was first, obviously, followed by New Jersey and then various other states. In the end he got a map with 627,421 square miles making up exactly 270 electoral votes and beating out the 268 electoral votes of the other 3,166,662 square miles of the country, just 16.5% of the country's land mass constituting a winning coalition. Actually, Alaska all by itself is bigger than these states which constitute an electoral majority. It's a majority of the people, though, with 166,439,539 living in the Yglesias victor states against just 142,306,179 in the losing states. That's a population density of 265.3 people per square mile in the winning states, and 44.9 people per square mile in the losing states.

Okay, cool. But the American population is a good deal less clustered in the Northeast than it used to be, so I got to wondering whether the smallest winning map might have been even smaller in the past. Using the 1960 census apportionment figures, i.e. the first batch after the addition of Alaska and Hawaii, I was able to craft an even smaller winning map:

Those blue states take up just 559,605 square miles, with 3,234,478 square miles of red territory, and an exact 270-268 electoral margin. My calculation method was a bit different from Yglesias'; instead of calculating 1960 census population density, I simply calculated electoral vote population density, and added until I hit 270. California was the tipping-point state, and since it got me to 284 EV's, South Carolina and West Virginia, only slightly more vote-dense than California, were superfluous. Incidentally, using this same method for the 2010 figures got me to the same map as Yglesias, though through a slightly different method. Whereas he added on the densest states until he hit Michigan, and then had 282 EV's, took Michigan back out to get down to 266, and added on New Hampshire to hit 270, I added on the most vote-dense states until I hit North Carolina, which got me to 282 EVs, and then subtracted off South Carolina (9) and Vermont (3). Somewhat confusingly, it doesn't look to me like Michigan is even in consideration; unless I'm really missing something, it is not, as he says, the 18th most densely populated state.

That's as far back as you can go, obviously, with Alaska and Hawaii still included, and since one of those two states in particular changes the geographic footprint of this country quite substantially the direct comparisons sort of end there. But just for fun, let's also look at the figures using the 1920 census apportionment, the first after all 48 continental states were added to the rolls. Here, we can get 268 electoral votes (enough to win, since there were only 531) with just 481,569 square miles, with 2,638,247 square miles taking the other 263 votes (the total is smaller, because we're missing Alaska):

Frustratingly, if you swapped out South Carolina and Virginia for Missouri, which would remove a bit more than 5000 square miles, you'd end up with 265 electoral votes in the blue states, and 266 in the red states. So I think this is probably the smallest possible winning electoral map since the continental United States got itself filled out. For whatever that's worth.

Census Reapportionment Not Mattering in Obama/Romney Contest

In the last census, northeastern industrial states lost a bunch of electoral votes, and Sun Belt states gained a lot of electoral votes. Democrats tend to do better in the former than the latter. That's the kind of thing that can only matter in a close election, but we're in a close election, so the electoral math matters. Will this shift of electoral power from the left-leaning to the right-leaning parts of the country matter this time around? Well, I suppose we don't know what the actual map will look like just yet, but we can get a sense of how the states are lining up. If we make a "path of least resistance" map, i.e. assign states to each candidate in descending order of how likely (per 538's current forecast) they are to carry those states, we find that Ohio is the tipping-point state right now, i.e. for both the Obama and the Romney path-of-least-resistance winning map Ohio is the last state added to the tally. The map looks like this:

In this map, Obama has 253 electoral votes, Romney has 267, and Ohio's 18 are the tipping-point. If we give them to Romney, he wins 285-253, and if we give them to Obama, he wins 271-267. So the question is, would either scenario be changed by reverting to 2000 census figures?

Answer: no. Of the blue states on that map, Massachusetts, New Jersey, Pennsylvania, Michigan, and Illinois each lost a vote, while New York shed two electors and Nevada and Washington gained one apiece, for a net loss of 5 electors under the new figures. So Obama's 253 electoral votes in these states were 258 votes before the 2010 census; not enough for a victory. But, of course, if you tack on the 20 votes from Ohio, you've got 278, and an Obama win. So the path-of-least-resistance calculus would be identical with the 2000 numbers: if each candidate won every state that FiveThirtyEight currently considers them to be more likely to win than any state they do win (i.e., if each candidate takes a path of least resistance given the current 538 odds), it all comes down to Ohio.

We can monkey with this calculus a bit, however. Suppose Obama loses Ohio, but goes looking for backup options. His best bets are Iowa and New Hampshire, worth a combined 10 votes now and 11 votes last decade. Under the new lines, that gets him to 263, which is no good. Under the old lines, that would have been 269, good for a tie but then, almost certainly, a loss in the House of Representatives. Or what about Virginia? It's got 13 votes under both census counts, so adding Virginia to the blue states on that map gives 266 and a loss under the current lines, but 271 and a win under the old lines. Colorado's 9 votes under both censuses would get Obama to 262 or 267, neither of which is sufficient; however, under the old lines, Colorado plus New Hampshire would have done it.

Another wacky scenario would be if Obama were to win Florida, while losing some other states that look easier for him as of this instant. Tack Florida on to the map above and you get 282 electoral votes under the new lines, and 284 under the old lines--since Florida gained two votes, this reduces the difference. Obviously either would be enough for a win, but we can then peel off Obama's other marginal states: taking away Nevada gets us down to 276 new votes and 279 old votes, and then removing Wisconsin gets us 266 new votes, and a loss, and 269 old votes, and a loss-through-tie.

So in the scenario where Obama loses Ohio but makes up for it by winning either Colorado or Virginia, the 2010 census will have changed the result of the election, but in pretty much any other scenario, it won't have mattered.

"State Fundamentals" And the 2012 Senate Contest

According to the forecast from FiveThirtyEight, Nate Silver's famous election forecasting site, Democrats are currently expected to hold 52.4 Senate seats after this election on average, a decline of just 0.6 Senators from their current position, and are given 86.4% odds of retaining the chamber. Those are some nice numbers for a year when Democrats were expected to get pretty well hammered. They've got Democrats leading solidly in MA, RI, NY, NJ, PA, DE, MD, OH, WV, MI, MN, CA, and WA, with rather robust leads in HI, NM, MO, WI, VA, CT, and FL, plus a tiny lead in IN. Republicans have modest leads in MT, ND, NV, and AZ, with a stronger lead in NE and locked-down leads in TX, UT, WY, TN, and MS. Independent candidates are projected to win in Vermont and Maine, both rather solidly. If you were to just "call" each state according to who's favored to win it, however slightly, you'd get 53 Democrats, including Bernie Sanders and Angus King, maintaining the status quo in overall caucus membership.

But digging into the numbers a bit, we find that there are a lot of important states where FiveThirtyEight is assuming that partisan gravity will play a really substantial role. In Arizona, for instance, the "adjusted polling average" has Republican Jeff Flake leading Democrat Richard Carmona by just 0.4%, but the "state fundamentals" say it's an R+8.3% state, so overall they've got Flake winning by 1.9%, and sporting a 62% chance of victory. In Connecticut, Democrat Chris Murphy's lead over Republican Linda McMahon is just 1.7% in the adjusted polling, but the +19.1% Democratic fundamentals number turns the whole race into Murphy +4.4%, 74% chance of victory. Indiana adjusted polling gives Democrat Joe Donnelly a 2.7% lead, but the fundamentals say it's Republican Richard Mourdock by 1%, so Donnelly's lead is just 0.3% in the forecast, and just a 52% chance of winning. In Montana, Democratic Senator Jon Tester holds an 0.6% lead in the adjusted polling, but Republican Denny Rehberg is favored by the state fundamentals by 8%, giving him a 1.6% lead overall, and Tester just a 38% chance of holding his seat. In North Dakota, adjusted polling shows a tied race between Republican Rick Berg and Democrat Heidi Heitkamp, but Berg takes a 4.9% lead in the forecast on the strength of a 10.5% lead in the fundamentals, and Berg is given 78% odds of victory.

That's a lot of key Senate contests being very heavily affected by our a priori assumptions of the race. That's not to say that FiveThirtyEight is doing anything wrong by adjusting for state fundamentals, and we can see that states with more robust polling are giving less weight to the fundamentals. But it is interesting to observe. As I read the landscape, the battle for control of the Senate is being fought on Democratic terrain in the East and Midwest, in Wisconsin, Missouri, Ohio, Florida, Virginia, Connecticut, and Massachusetts. But the battleground in which Democrats will seek to expand their majority is Western Republican terrain: Montana, North Dakota, Nevada, Arizona (okay, and let's give Indiana a special exemption, even though it's not western). If you're a believer in partisan gravity, then you would expect Democrats to squeak through the one blue-state defensive contest where the polling's been a bit scary, namely Connecticut, and to pull through the toss-up races, but not to pick up the Republican-held seats out west. If partisan gravity fails to show up, though, Connecticut will be very close, although probably will still go for the Democrat, but Montana, North Dakota, and Arizona become very interesting. Tack on those three states, plus a slightly more solid lead in Indiana, and Democrats could get to 56 seats without even winning in one of the more Hispanic states in the nation, Nevada.

In one way it will be an interesting test of modelling assumptions. In another, it's a demonstration of the need for more polling. These Senate contests in small states with non-competitive Presidential contests are important. People need to be polling them. That they are not doing so is not okay.

Democrats: We Care About People

A recent NBC/Wall Street Journal poll asked which party people thought was "most attuned and sensitive to issues that affect" various groups in society. (My apologies if the link doesn't lead anywhere that looks appropriate; I saw this poll on a poll-aggregating site, not on its own direct web page.) They asked this question for the following groups: religious conservatives, men and women in the military, small business owners, retirees, you and your family, stay-at-home moms, the middle class, working women, young adults under 30 years old, Hispanics or Latinos, and gays and lesbians. Interestingly, though this isn't my main point, most of these are groups that it is typically considered a self-evident good to be attuned to the interests of, but, notably, a whole lot of people would not feel that way regarding religious conservatives, including me, and a substantial but declining number would not feel that way about gay people. That probably explains why those questions produce such imbalanced results, because one of the political poles is proud to declare itself as attuned to that group's interests and the other pole is happy to accuse their opponents of being so attuned. But, as I say, not my main point.

One Political Rule to Rule Them All

Harry Enten had a column in the Guardian recently about political "rules." You know, like "incumbents never win re-election with approval ratings below 50%." Until, of course, they do. The basic conclusion of the article is the very sensible one, namely that, since in each election one can find scores of rules saying that both candidates simply must win, the rules are basically bunk. Like I said, this is basically sensible, but there's one rule of which I'm rather fond: no Republican has ever won the White House without winning Ohio. I like this because it makes sense, and it illustrates the one long-term constant of American partisan politics. Ohio is the state which, perhaps more than any other, embodies the business interest. Other industrial Northern states tend to contain bigger, denser cities, leading to more sympathy for the workers and liberalism in general; Southern and Western states typically have non-industrial concerns. But in Ohio, businessmen, business owners, and capitalists generally tend to be the dominant political group. And the only thing that's been a fixed point in the American political scene is that the Republican Party is pro-business. They were pro-business under Lincoln, pro-business under McKinley, pro-business under Coolidge and Hoover and Nixon and Reagan and Bush. It makes sense that if they couldn't even convince Ohio, the natural dwelling-place of their basic historical constituency, to vote for them, they wouldn't have much luck convincing the nation as a whole.

Adventures in Weird: Romney Handicap Edition

Regular readers, if there are any, may recall my notion of the "Romney handicap," a figure calculated by subtracting President Obama's net job approval from his margin over Mitt Romney in a general election trial heat in any given poll. Well, here's an oddity. A recent CNN/ORC poll showed Obama with 48/47 approval numbers, for a +1 margin. They also show general election results of Obama 53%, Romney 41%. That's a +12 margin, for an 11-point Romney Handicap. But the weird thing is, that same poll showed Mitt Romney with 44/43 favorability numbers, an identical margin to Obama's approval rating. Weird, right? Why, if he's equally popular, is Mitt Romney giving Obama so much of a handicap?

Well, it's because this is an apples-to-oranges comparison, favorability versus approval. Obama's favorability numbers in this same poll are 56% favorable, 42% unfavorable. So he's got a +14 margin on favorability, Romney's got a +1 margin, and when you run them against each other Obama's got a +12 margin. Sounds about right. This makes me wonder whether the focus on approval rating might not be misplaced. Perhaps we should be looking at favorability ratings; after all, you can't have an approval rating for a challenger. Maybe the true rule isn't that Presidents need 50% approval to win re-election, but rather 50% favorability. Or maybe I'm just taking one slightly kooky poll result and drawing way too many conclusions from it. But it's a phenomenon to keep an eye on, if you like keeping an eye on esoteric polling phenomena.

I Don't Get It

PublicPolicyPolling recently asked voters nationwide whether they had a favorable or unfavorable impression of each of the 43 men who have been President of the United States (with the option, of course, of "don't know"). Now, it can be a little tricky to compare these ratings since name recognition, which I define as the percentage giving an answer either way, varies from a low of 16% for Franklin Pierce to a high of 95% for Barack Obama. Is a 12%/9% approval rating better or worse than a 42%/39% one? But looking at the numbers can show us some interesting, and at times perplexing, things:

The Insanity of an Economically-Determined Electoral World

In his morning "Wonkbook" post rounding up interesting political happenings from recent times today, Ezra Klein spends the first paragraph talking about Andrew Sullivan's recent article praising and defending the Obama Administration. Then he says this:

But Obama's reelection won't be decided in the pages of newsweeklies. It will be largely decided by the state of the economy. And the state of the economy will largely be decided by events in Europe. And Europe's not looking so good.

Now, I'm not here to quibble with the factual accuracy of that statement. If we interpret the word "largely" loosely, then it's basically true: Nate Silver's analysis at FiveThirtyEight suggests that macroeconomic performance may account for about half of the variance in Presidential election results, and Europe will certainly be quite important to American economic performance in 2012. No, my point is different. This idea, that Presidential elections are determined largely or almost entirely or, at the extreme, entirely by crude macroeconomic results, may be true, but if it is true it tells us some very bad things about our voters.

Comparing Nate Silver and Ladbrokes

From time to time I like to look at the odds that inTrade, the big political gambling/prediction site, is giving various Republicans to win the nomination. But I think I actually prefer the odds at Ladbrokes, a major British betting site. They don't just have numbers on the nomination and general election: they've also got things like Iowa caucus odds. And, by a happy coincidence, FiveThirtyEight just released their first Iowa prediction, where the bottom line is percent chance to win. So what happens if we compare them? Well, for any given odds you can calculate an implicit percent chance to win, so if we do that with the Ladbrokes odds and compare them to 538's numbers we find that... Ladbrokes likes every single candidate more than Nate does. This, of course, is because the bookies like to give you shorter odds than you deserve, so that they make money in the end. Their implicit probabilities of winning Iowa add up to 124.8% between the eight candidates. But if we divide out by this number, and scale it down to a total of 100%, we find a fairly strong correlation between the two methodologies. Nate gives Newt a 49.6% chance of winning, compared to 28.2% for Ron Paul, 10.6% for Romney, 5.2% for Rick Perry, 4.1% for Bachmann, 1.6% for Santorum, and 0.7% for Huntsman. Ladbrokes, by contrast, gives Newt 46.4%, Paul 26.7%, Romney 11.4%, Perry 5.2%, Bachmann 4.7%, Santorum 3.2%, Huntsman 1.2%, and Gary Johnson 0.4% (he's not in the 538 forecast, I think). The big takeaway is that Nate likes the front-runners, Gingrich and Ron Paul, a lot better than Ladbrokes, giving them a combined 77.8% chance to win against 73.1% at Ladbrokes. Conversely, Ladbrokes is slightly overselling the odds of a Romney, Perry, or Bachmann win, and vastly overselling Santorum and Huntsman (giving them 5% combined instead of 2.3%; that's more than double!). Do not bet on Rick Santorum to win Iowa on Ladbrokes; they're giving you awful odds.

Of course, you shouldn't actually bet on Newt or Ron Paul either, because Nate only likes them better than Ladbrokes' oddsmakers once you scale down to a sum of 100%. Overall, though, it's a pretty impressive alignment: the correlation between Nate's odds and the Ladbrokes odds is around 99.9%. (This is impressive for Ladbrokes, by the way, since we assume Nate knows what he's talking about in advance.)

Why I Can't Support the Constitutionalization of the Right Against Gerrymander

I'm a huge fan of Baker v. Carr, and all its progeny, the cases which for the first time in our nation's history subjected the apportionment process to constitutional scrutiny, and declared the fundamental principle of "one man, one vote."* I also think that there is roughly-speaking zero good reason for anything resembling the kind of gerrymandering that's standard fare these days. I'm pretty sure that if you plugged a population-distribution map into a computer, and gave it a few parameters about respect for political divisions, respect for demographic distinctions, etc. versus the desire for compactness and simpleness of form, that computer could spit out a map of equipopulous districts rather quickly. I think we ought to do that, rather than let committees of the politically-powerful draw maps to achieve various ends. And yet, I just cannot bring myself to support the idea that anything like independent, non-gerrymandered districting is required by the Constitution.

The fundamental logic behind the idea that this is a constitutional requirement, or should be anyway, is that we want the voters to choose the government, not for the government to choose the voters. A scheme of partisan gerrymandering, where the opposing party's voters have their votes diluted much as blacks had their votes diluted in old Southern maps, seems like an obvious case of one party using power to perpetuate its own power. I'm open to the idea that this is not acceptable, constitutionally speaking. But lots of gerrymanders aren't like this! They are, often, for the purposes of 'incumbent-protection.' In some states, the two parties achieve a state of detente, in which each is allowed to protect all of its incumbents, reducing the number of 'swing districts' and locking in the status quo. This seems uncool, right? We want The People to retain the power to vote their government out of office.

But here's my problem: they have that power! Voters can just vote for someone else if they want to! If a district is drawn in which Barack Obama captured 70% of the vote, with the intention of creating a "safe seat" for some Democratic incumbent to cushily occupy for all eternity, the voters of that district retain a complete right to kick that incumbent out of office if they so choose! Given that we have a fairly universal open primary arrangement in this country (and I do think there may be some sort of right to an open primary, though I'm not sure what I think about that issue), those Democrats can just vote for some other Democrat in the primary, as happened to Emmanuel Celler and Adam Clayton Powell in deepest-blue New York City. Or they could vote for some independent candidate. Or for the Republican. It's up to them. That it is highly unlikely a majority of them will decide they prefer the Republican challenger to the incumbent Democrat does not mean that their right to elect whomever they damn well please as their representative has been in any way lessened.

To put it another way, I just can't bring myself to accept that there's a right to live in a swing district. Hell, if we could define swing districts precisely that would seem to imply that voting preference couldn't change, which would make the whole notion of a right to vote kind of silly. The fact that I, and the vast majority of my fellow inhabitants of the First Congressional District of Rhode Island, have a fairly well-settled preference for the Democratic Party over the Republican Party, does not mean that our right to vote has been infringed!!!  I take about as much offense at an incumbent-protection map, like the one that governed California for the last decade, but c'mon people, it's just not an infringement of the right to vote, as malapportionment is. It's not an attempt to remove political power from members of one race, or one political faction. It's just the creation of relatively more districts which, we expect, will be won by large margins as opposed to small margins. There's no right to have roughly fifty percent of the people in your district favor each political party. So however much I might favor independent districting, I just can't conclude that it's a constitutional requirement.





*That's how the principle was phrased at the time; obviously, it should now be rendered "one human, one vote." (And yes, that's better than "person," because we don't give non-humans any votes, now do we?)

While We're At It, the Gingrich Handicap

See the original post on the Romney handicap. The Gingrich handicap is exactly the same idea, just for Newt Gingrich instead of Mitt Romney. Well, what do we find when we look at the Gingrich handicap? It's a lot bigger than the Romney Handicap. The recent PPP national poll found Gingrich polling three points worse than Romney, for a 12-point handicap. That's about the smallest gap anywhere. In the state surveys I classify as "recent," in which Romney displays about a 10-point handicap on average, Gingrich is spotting Obama a 17-point edge. That's enormous. Honestly, a 10-point handicap candidate is pretty awful to begin with, and should make us reconsider the idea that Romney's a strong candidate. But -17 is just abysmal. Note that Lyndon Johnson's approval rating around the 1964 election was probably well above the record 61% of the popular vote he received, i.e. Mr. Catastrophically Weak Candidate Barry Goldwater actually posted a negative handicap, by this metric, relative to President Johnson.

Getting back to Newt, there's a pretty broad band of states, both in recent surveys and in older ones dating back to last winter, where his handicap is in the realm of 17 to 23 points. Then there's a handful with a dramatically smaller handicap. Among recent surveys, that's Nevada (12), North Carolina (also 12), and Mississippi (7). Nevada and North Carolina are swing states, and pretty good places to be relatively strong (although there's the caveat that Romney's handicap in that NC survey was also anomalously low, and I suspect that may be random variation rather than anything else). But this is some serious unelectability. If we assume zero undecideds, then Obama's approval rating would need to be 41.5% in order for Gingrich to tie him nationally, under these numbers. If we assume as many undecideds as we see in the latest pollster.com trendlines, we'd need to see an Obama approval of 39% before Gingrich could start thinking about winning. I'm pretty sure the Obama people are counting on an approval rating greater than 39% on election day (not that they might not get it, but that if they don't it won't be surprising when he loses). If they nominate Newt, they're throwing this election away, unless something massively changes over the next year. Go for it, guys! It worked so well in the 2010 Senate races, after all.

Romney Handicap Continues to Grow

I've defined something I call the Romney Handicap, which for any given poll is Obama's net margin over Romney in the general-election trial heat minus Obama's net approval rating in that same poll. The idea behind it is that it tells us what approval rating Obama needs to beat Romney: if the Romney handicap is 5 points nationally, then Obama should be able to tie Romney nationally even if his net approval rating is -5. When I originally wrote my post on the subject, the average Romney handicap from all the most recent Public Policy Polling state-level surveys was around 6 points, maybe a little bit less. But shortly thereafter I noted that more recent surveys had shown a larger Romney handicap, and indeed, several weeks later that trend has continued. Of the twelve PPP state polls I designate as "recent," a category that starts a month or two ago and includes all polls taken since that time, the average Romney handicap has been 10.25%.

The range has run from 17 points (Ohio, a bad place for Romney to be at his very weakest) down to just four points, in North Carolina. PPP polls NC every month, and the month before the most recent the Romney Handicap there was much larger. Mississippi clocks in with a 5-point Romney handicap (which I don't think Obama will much mind as one of his relative weak spots!). Other than these three outlier states, the range is from a 7-point to a 14-point handicap. Meanwhile, while the previous PPP national poll showed a Romney handicap of just 6 points (a tie vs. Romney but a -6 approval rating), this month Obama has opened a three-point lead on Mitt while retaining the lousy approval rating, for a 9% handicap. All of this suggests that, over the past couple of months, Mitt Romney has become a weaker candidate, by somewhere from three to five (net) points nationally. That's a big deal. An extra three points of margin, nationally, would've made John Kerry President. In a close race, the damage Romney's sustained of late could make the difference. By the theory of this model, the break-even point for the Obama vs. Romney match-up has shifted from a -6% Obama approval rating to a -9% or -10% one. Pollster.com tells me that right now Obama's national approval is -6.8%, similar to what PPP found. If the election were held today, therefore, Obama would be re-elected over Romney, but only because of the growth of the Romney Handicap over the past few months. This continues to be a trend worth watching.

Narrowly Defeating an Incumbent

Here's a question that occurs to me in light of my previous post: how often has the incumbent President narrowly lost a Presidential election? Well, let's run down the list. When Bill Clinton beat George Bush, he did so by a 5.5% popular-vote margin, equivalent to a 6.8% margin among the two-party vote, and scored 2.2 times as many electoral votes as Bush. Moreover, Bush managed only 37.5% of the overall popular vote, due to Ross Perot's third-party bid. Ronald Reagan beat Jimmy Carter by nearly 10 points, and got 10 times Carter's electoral tally. In 1968 Lyndon Johnson didn't even make it to the general election (this has happened other times). Franklin Roosevelt won 57.4% against Herbert Hoover, who fell under 40% and carried just eight states worth 59 electors. In the wacky three-way of 1912, William Howard Taft won a stunning eight electors on the "strength" of his 23.2% popular-vote showing. In 1828 Andrew Jackson crushed John Quincy Adams 178-83 in the electoral college, and 56% to 43.6% in votes.

That leaves us with just a few examples. In 1840, William Henry Harrison (who had come within 4000 votes in Pennsylvania of becoming President four years earlier) beat Martin Van Buren by a 6-point margin popularly and a roughly 4-1 margin in the electoral college. In 1800 Thomas Jefferson and the Democratic-Republicans won 73 electors to President John Adams' 65, a very narrow margin; however, in the handful of states irregularly allowing popular votes, Jefferson won over 60% of the vote. Then you had the Cleveland-Harrison festival, where Harrison "beat" Cleveland by a "margin" of negative 0.8% in the popular vote, and just 233-168 in EV's, and then Cleveland beat Harrison by a 3% margin, taking 277 EV's to Harrison's 145. I'm tempted to throw those elections out, since Cleveland won the popular vote in both of them. Likewise, since in Jefferson's day the popular vote didn't really exist, I'm tempted to throw that out, too.

We're left with the 1840 and 1992 elections, which featured popular-vote margins of roughly 6% and much more landslide-style electoral college performances. In other words, it's rare. When an incumbent President loses, they tend to lose big. And I may be a bit of an optimist here, but I doubt Obama is going to lose big in 2012. Maybe he will: maybe the economy will get that much worse, and Romney will somehow run a strong campaign. Interestingly, I also once found the converse, i.e. when incumbent Presidents win re-election they usually increase their original margin of victory. I guess close elections featuring incumbents are just pretty rare. This makes sense on the prevailing theory, that when times are good or getting better you just re-elect the incumbent and when times are bad and getting worse you dump them for the other guy. The 2012 election is weird, then, in that times are bad, getting better at an unacceptably slow pace, and the alternative is considered, well, awful. So perhaps there just isn't that much precedent on the subject. In any event such historical precedents don't actually have much predictive value, since there are always lots of them on both sides, but it's an interesting result nonetheless.

Romney's Handicap

I mentioned in my last post that each different Republican candidate appears to give Obama a certain 'handicap' in his re-election bid. That is to say, you can roughly think that every change of 1 point in Obama's net approval rating will cause a 1-point change (in the same direction, obviously) in his margins over his Republican opponents. But for each different opponent, there's a different calibration. In one poll Obama might lead Romney by 3 points, Perry by 8, and Palin by 17, all obviously with the same approval rating. So each candidate is giving him a different handicap. I thought I'd take a look at trying to quantify the Romney handicap.

Warning, Republicans: Presidential Approval is Volatile

Barack Obama's job approval ratings are not very good right now. Pollster.com's average puts it around 43% nationally. And yet, in this environment, we have the scary-good polling firm PublicPolicyPolling showing that Obama is tied nationally with Mitt Romney, his most likely general election opponent and by far his most formidable potential foe, and leads Romney by 4% in Iowa. Moreover, that's in an Iowa poll where Obama's approval rating had declined by six net points since August and his lead over Romney had declined by... six net points. In other words, as I've long been suspecting, Obama's general-election vote share looks a lot like a direct, point-for-point translation of his approval rating, plus some handicap depending on who his opponent is. And notice that Romney's trailing Obama by four points in a state where Obama's net approval rating is -9. That's a 13-point net handicap. That's enormous! So it's not so much that Romney's a strong candidate, but rather that Obama is sufficiently vulnerable that even a very weak candidate like Romney is competitive against him, and all of Romney's primary competitors are incredibly vulnerable. Okay, fine, that's the situation. It looks pretty decent for Republicans, actually.

The Curious Case of West Virginia

And no, I'm not talking about the weird way in which it got admitted to the Union. I'm talking about the strange pattern it followed as it gradually stopped thinking that it was a northern state, and started thinking like a southern state. Here's my analysis: West Virginia is closely connected to three states, Ohio, Kentucky, and Virginia. I don't think the connection to Maryland or Pennsylvania is nearly as strong. So, let's look at the difference between its presidential voting and that of those three states in each election since its creation. If we do, a very interesting pattern emerges. From 1864 through 1936, the closest of these three states to West Virginia was Ohio twelve out of nineteen elections. On average the WV-OH difference was just 8.8% in this span, compared to a 14.9% difference with Kentucky and a 19.9% difference with Virginia. During this period, West Virginia thought it was a northern state. Then something flipped. From 1940 through 1980, Kentucky was the closest state to West Virginia nine out of eleven times, differing by an average of just 5.2%, compared to a 12.4% differential for Ohio and a 16.0% gap for Virginia. Now Kentucky was thinking like an inland southern state, not a northern state.

Since then it's gotten a little bit complicated: we've had three Ohio elections, two Virginia and two Kentucky. Meanwhile, the average gap is 9.8% for Kentucky, 10.7% for Ohio, and 14.3% for Virginia. I'm not sure exactly what's been going on, but part of the confusion is doubtless because the party of northern liberals nominated a southern moderate during two of those elections. In 2008 West Virginia was right alongside Kentucky in radically opposing Obama, differing by just 3%, while Ohio and Virginia supported Obama and had gaps more like 18% or 19%. One other thing I note is that there was a bit of a transition period in the flip I observe in the previous paragraph. From 1924 through 1956, the difference between Kentucky and West Virginia was under 10% in every election, averaging a paltry 2.4%. And while Ohio 'won' all six elections from 1892 through 1912, and Kentucky 'won' the five consecutive elections from 1940 through 1956, they split the six elections from 1916 through 1936 three apiece.

So what we see here is that, after seceding from the Confederacy and joining the Union, West Virginia originally thought it was a northern state. It wanted nothing to do with its Confederate progenitor, and didn't even really agree much with its neighboring fellow border state. But then, at around the time the shift from Southern conservative Democrats and Northern liberal Republicans to the other way around was just beginning, West Virginia started to realize that it wasn't, actually, a northern state. It still wasn't Virginia, but now it was perhaps more aligned with Kentucky, so similar in demographics and geographic situation. More recently, things have gotten a little bit more confusing, as (among other things) Virginia proper has stopped being so certain that it's a Southern state. Right now it feels like in the long term Virginia may wind up being the most liberal of these states, with Kentucky and West Virginia the most conservative. It's a funny path this odd little state has taken, I'd say.

Splitting the Electoral College

I know the Republicans aren't playing this game in good faith or anything, but just for fun let's look and see what the results of the 2008 Presidential election would have been if every state had allocated its electoral votes in the Maine/Nebraska manner, i.e. 2 for the winner of the state and then 1 for the winner of each Congressional district. As it happens, Obama would have won 301 electoral votes, compared to 237 for John McCain. That's a lot narrower than his actual 365-to-173 win. However, it represents Obama taking a slightly larger percentage of total electoral votes, 55.9%, than he received in actual votes, where he got just 52.9%. It's even an increase over his two-party-vote performance; dropping third-party votes, Obama won just 53.7% of the remainder. So a split electoral college in 2008 would've produced an overall result very similar to the popular vote numbers, with a slight boost to the winner, instead of the lopsided 68% win that the (mostly) winner-take-all version gave him. I quite honestly have no problem running things this way instead of the way we do it, but the point is that we need to do it uniformly. Also, in this case I suppose I would have even more of a problem with gerrymandering than I do currently; in fact, I even think switching to this kind of system would make me more interested in the arguments that Baker v. Carr and its progeny would then require an extension to prohibit gerrymandering. When I have time later, I hope to take a look at the '04 and '00 elections and see if a split electoral college would've made any difference (I bet it would have in 2000).

Of course, the real solution is just to ditch the electoral college altogether, either through the National Popular Vote Compact or, preferably, through a Constitutional Amendment. I wonder if the spectacle of Republican state legislatures screwing with their electoral votes might be enough to give one or both of those projects the boost they need to succeed.

Home State Winnings

Here's a fun fact: only two people in history have won a U.S. Presidential election without carrying their home state's electoral votes. In 1844, James Polk of Tennessee lost that state by 0.1% to Henry Clay of Kentucky. Then, in 1916 (a strange election to begin with, given Woodrow Wilson's bizarre three-way victory in 1912 and the fact that it's one of only two times that a President's re-election margin has been narrower than his initial victory) President Wilson lost New Jersey rather healthily to Charles Evans Hughes of New York, though he did capture a higher percentage of the popular vote in New Jersey than he had four years earlier. Polk is the only person ever to do this without already having been President, of course.

As for Presidential losers? Well, there have been 60 losing major-party Presidential candidates in our history, by my count, including multiple losers like William Jennings Bryan. Together they have won their home state just 33 times, or 55%. There appears to be some streakiness here as well: through 1848 73% of losers won their home state. Then, through 1884, just 45% of losers won their home state. As if that weren't bad enough, from then through 1956 just 25% of losers were successful at home. Since Adlai Stevenson's double defeats to Eisenhower, though, only George McGovern (who lost a landslide) and Al Gore (who won the election) lost their home states, a loser home winning percentage of 86%.

So, what does this tell us about the 2012 elections? Well, I think it's kind of a reasonable conclusion that if the Republican can't win their home state, they aren't very well favored to win the election. Sure, there's no particular reason to say this must be, and after all Al Gore did essentially win his election, but the fact remains that it is incredibly historically rare for someone who isn't currently President to win a Presidential election without the support of their native state. Let's look at the people FiveThirtyEight ranked recently as either a member of the "top tier" or a "wild card," as opposed to a "long-shot." Let's be honest, I don't care about guys like Buddy Roemer in this analysis. So, here goes: