Is this 92 all over again?

[Nichlemn]

It's a safe bet that 2012 will end up having superficial similarities to a vast number of elections, such that plenty of two-bit political pundits will write lengthy columns espousing the apparent significance in projecting the results. Of course, almost all of these pundits will conveniently overlook the vastly more numerous differences. Every election is more unique than it is similar to any other election.

[Nichlemn]

Anything is possible after the 1992 election, but I think right now, this is probably 1984. Of course, not nearly as big a victory.

Let's look at the Presidential elections beginning in 1992.

It's also a safe bet that we'll see more cherrypicking of dates in the future. Democrats like to pick 1992 as a starting point because since then, we've seen three fairly strong Democratic victories and two small Republican victories. This supposedly indicates that Democrats have an advantage in Presidential elections. But you could cherrypick in favour of Republicans by either going back further (to 1968 or 1980 for the biggest effects) or going forward further (since 2000).

Or suppose we were attempting this exercise in the past (looking back at the previous five elections). Leading up to 1992, we would conclude that Republicans had a "lock" on ~230 EVs (Ford states - Iowa and Washington, which Dukasis won in 1988) and had a large advantage in many other states. Leading up to 1952 we would conclude that Democrats had a similar advantage. We know how those elections turned out.

Those examples are obviously themselves cherrypicked to show how the "model" can fail spectacularly. But scanning all elections, it doesn't look like the results of the past five elections have a great deal of predictive power. How well did the previous five elections predict 2008? 2004? 2000? Before? My guess is that while state results predicted future voting in that state fairly strongly (they're certainly not random), states weren't "locked tight" to the extent you claimed. (For instance, Bush lost twelve states in 1992 that were "locks" by your logic. Even if you think that race was "realigning" or something - even in the EC landslide of 1988, Bush lost Oregon and Iowa, two "lock" states). And there's quite possibly a negative correlation between overall results for a party and the next election, due to desire for change.

This article by Nate Silver shows how even apparently technically advanced models can be highly flawed. The example he has is of a model that predicts House results in Presidential election years using data from 1952 onward... but if it was extended to 1948, that election would be an enormous outlier.

[Nichlemn]

Quick analysis - since 1876 (the first election in which the Democratic and Republican parties had each existed for the previous five Presidential elections), the correlation between the number of Democratic wins in the past five elections and whether a Democrat wins the next election is -0.06 - effectively nothing. No, it's not counting popular vote or the Electoral College, but I can't imagine they're going to swing it into a large positive number.

[Nichlemn]

Anything is possible after the 1992 election, but I think right now, this is probably 1984. Of course, not nearly as big a victory.

Let's look at the Presidential elections beginning in 1992.

It's also a safe bet that we'll see more cherrypicking of dates in the future. Democrats like to pick 1992 as a starting point because since then, we've seen three fairly strong Democratic victories and two small Republican victories. This supposedly indicates that Democrats have an advantage in Presidential elections. But you could cherrypick in favour of Republicans by either going back further (to 1968 or 1980 for the biggest effects) or going forward further (since 2000).

Anyone who can't see a huge difference between 1976 and 1992 can't see the trend that had taken place during three Republican Presidential landslides. Democrats used to need the South, beginning in 2000 they needed to find a way of winning without the South. (Florida  politics are so different from the politics of its neighbors and near-neighbors that it belongs in a region of its own).

Facts should create no problems; interpretations create them. I can't be certain that a Democrat can win  even as a moderate populist (Carter, Clinton)  in enough Southern states to have a chance, and it is hardly certain that southern populists have remained Democrats.  It could be that the states that I show in green are steadily drifting out of the reach of Democrats as Democrats gain elsewhere.  Does the South have a chance to churn out the populists that it used to, or has it become the ultimate preserve of corporate-friendly politicians who have cozy relationships with economic elites who wield the real power?

You must be logged in to read this quote.

Political shifts seem to occur under the cover of blow-out losses. The two-party system tends to force losing parties to attempt to form new coalitions with which victory becomes possible. I figure that Bill Clinton wasn't far from Jimmy Carter in ideology -- although Clinton was a far more adept politician. Clinton won a bunch of states that Democratic nominees for President just did not win except in electoral blowouts like 1964 and in the FDR era.  That Clinton could win states like California, Michigan, Maine, and Vermont suggested that something had changed since 1976.

Winning coalitions can also lose parts of their coalitions. The so-called "Rockefeller Republicans" who might have been uncomfortable with a Democratic Party that held many anti-intellectual populists, drifted D.

You must be logged in to read this quote.

We wouldn't have such results had there been a Republican blowout after 1988. The quality of politicians matters, and who runs matters greatly. Had the Republicans had a President more effective in getting his point across than Dubya, then numerous states would be in different colors. A shift of 3% of the vote toward the Republicans in 2004 would have knocked New Hampshire out of the pink category while shifting Oregon, Wisconsin, Michigan, and Pennsylvania into the pink category.  Who runs matters greatly, of course. I can't imagine any Democrat in the last seventy-five years winning Indiana in an election as close as 2008  except Barack Obama.  

You must be logged in to read this quote.

Obviously, any technical model that fails to account for the style and quality of a campaign, the ability of outside groups to funnel money into the contest, choice of running mate, and the personality and campaign style of a candidate will show weaknesses. It is possible that John Kerry would have delivered New Hampshire in 2000 to Al Gore and altered the course of American history greatly.  It is also possible that John Edwards, a southern populist similar in many ways to Carter and Clinton, was a perfect choice for winning over a constituency of the Democratic Party that no longer existed. Political cultures in states can change with demographic change. It could be that Indiana has been a relatively-safe state for Republicans because a strong  Democratic nominee almost never campaigns there -- but President Obama did.  Virginia used to be a reliable R state -- but not any more.  Of course either Party can run a failure against a President who cultivates significant popularity, in which case the election isn't close. Incumbents  (Hoover, Carter) can be catastrophic failures as President with predictable results.

....

So far, about every non-partisan projection shows President Obama getting re-elected with much the same states as in 2008. Five states shifted between 1992 and 1996, and three states shifted between 2000 and 2004. The states most likely to shift between 2008 and 2012 should President Obama win roughly a 53-46 split of the popular vote  are those closest  in 2008 or entail the Favorite Son effect (for example, Jeb Bush would win Florida but probably at the expense of Arizona). But should the Republicans nominate an incompetent campaigner or someone easily depicted as crazy or absurd, then how what would the 2012  election say about the political trends of the states?

Absolutely nothing. If the Republican nominee is so unattractive as to lose Texas, Kentucky, and the Dakotas, then such says nothing of long-term shifts. The 1964 Presidential election hardly foretold what 1968 would be like.

I'm not criticising the idea that there have been signficant trends - clearly there have been. The only objection I have is your apparent claim that Democrats have an advantage in having more states "locked up" by virtue of having won them each time in the past five elections. This is similar to claiming that Democrats have an advantage for having averaged a higher share of the popular vote in the past five elections.  If that wasn't your implication, then I have no significant disagreements with you.

[Nichlemn]

You must be logged in to read this quote.

You must be logged in to read this quote.

I'm not sure whether you're implying that recent electoral history implies Obama is a solid favourite for re-election or not. I mean, I agree that he is, but it's because of a whole lot of factors solely related to this election. If Clinton had won more narrowly or Bush more strong, resulting in some ruby red states becoming light blue on your map, this should have very little impact in how we predict 2012. Do you think your map would have less predictive value if you used the statewide results in hypothetical tied elections?

[Nichlemn]

You must be logged in to read this quote.

You must be logged in to read this quote.

I'm not sure whether you're implying that recent electoral history implies Obama is a solid favourite for re-election or not. I mean, I agree that he is, but it's because of a whole lot of factors solely related to this election. If Clinton had won more narrowly or Bush more strong, resulting in some ruby red states becoming light blue on your map, this should have very little impact in how we predict 2012. Do you think your map would have less predictive value if you used the statewide results in hypothetical tied elections?

Since 1992 we have essentially two sorts of Presidential elections -- one sort  in which the Republican (Dubya both times) wins by a narrow margin in electoral votes, and one in which the Democrat (Clinton twice, Obama once) wins by about 100 more electoral votes than is necessary. A relevant alternative is that the Democrat barely wins -- if Al Gore had gotten a few more votes in Florida.  Maybe I could use that model if someone can prove that some state officials cheated in Florida in 2000.  The last electoral blowout was 1988 with the meltdown of the Dukakis campaign. This model doesn't explain electoral blowouts well, but electoral blowouts need little explaining anyway.

These look like the possibilities for 2012:

1.  Bare victory by the Republican nominee  -- if the American public votes as it did in 2010. It would look something like 2000 or 2004. The Republican wins 270-310 electoral votes.

2.  Bare win by President Obama with 270-310 electoral votes. This implies the President losing most of the states that he won but neither Gore nor Kerry won (those in pink) and either Colorado and Nevada, Virginia, Ohio, or Florida and perhaps another.  

3. A win by a margin of 100 or so electoral votes (355-385) which fits the pattern of Clinton twice and Obama once. This would likely be a near-duplicate of 2008, perhaps with President Obama losing Indiana but picking up two of Arizona, Georgia, and Missouri.

4. An Obama landslide in which the President faces a weak, extreme, or incompetent opponent and starts picking off states in green

So the only basis you have for a strong Republican victory not being a "possibility" is that it would require them to win states that haven't been for five elections? As history has shown us, no such pattern has existed before. You could say "this time it's different", but that's getting into bare assertion territory.

You must be logged in to read this quote.

Again, I don't think there's any statistical significance to this relationship, especially as you got to pick the goalposts.  Although the reasoning at least makes sense theoretically, what other examples other than 2008 exist with an underdog employing a very risky strategy? Even if there are quite a few, there's so much uncertainty in the final results that it's going to be impossible to engineer a strategy that only wins narrowly or loses heavily. It wouldn't taken a lot for say, Texas to go to Nixon in 1968 and ruin this apparent pattern. Maybe it reduces the chances of Obama winning with about 330EVs, but not by a lot.

[Nichlemn]

PB2, what do you think it translates to, probability-wise? Sounds like you're very confident in an Obama win, in which case you could make good money on Intrade/wagering with people.

[Nichlemn]

Again, I don't think there's any statistical significance to this relationship, especially as you got to pick the goalposts.  Although the reasoning at least makes sense theoretically, what other examples other than 2008 exist with an underdog employing a very risky strategy? Even if there are quite a few, there's so much uncertainty in the final results that it's going to be impossible to engineer a strategy that only wins narrowly or loses heavily. It wouldn't taken a lot for say, Texas to go to Nixon in 1968 and ruin this apparent pattern. Maybe it reduces the chances of Obama winning with about 330EVs, but not by a lot.

It's my seat-of-the-pants application of game theory. For the nominee who is decidedly but not hopelessly behind -- like McCain in 2008, these are the apparent prospects


Strategy                                                             Electoral votes

Gamble and win (one chance in 50):                        275... and win
Gamble and lose but make things close                   250
Play it straight                                                          220
Gamble, but things go wrong                                   170

The consequences for 120, 220, and 250 electoral votes are much the same: a loss. Playing it straight would have meant that McCain would have shored up his strengths and abandoned any quixotic efforts to win 'back' a state that most people thought that he was losing decisively.

"Play it straight" would be a reasonable strategy for someone who thought himself a likely nominee for President four years later. In a slightly-different universe, this is what Barack Obama might have been in. But it just wasn't so.

Now what does it look like for the other side:

Strategy                                                               Electoral votes

Gamble and lose (1 chance in 50)                               265... and loss
Nickel defense                                                             290
Both play it straight                                                     330
Successful defense against a wild gamble                  370

The Nickel Defense is a method of defensive play in American football, and it worked reliably for decades.  It basically trades time that the team then behind can't afford to give up for yardage that the losing team can't use well enough. The team behind can make steady gains on running plays that devour fast-disappearing time that becomes the ally of the team in the lead. If the team behind makes a daring pass for a big gain, then the five defensive backs make an interception very likely -- and create the possibility of easy points scored on the runback of the interception or some solid offensive play from a good position in the field, making the score even more lopsided.

But both teams playing it straight means that either both teams score similar numbers of points, which is a wash -- but the team already ahead wins.  

Barack Obama may not have been a football coach, but game theory works for politics as it does for football.

.....

Now if the GOP nominee is behind in September 2012 and plays it straight, then maybe that candidate is looking to 2016 with 2012 as a dress rehearsal, perhaps making appearances on behalf of beleaguered candidates on his side.  The likelihood of that seems small... but it is still far too early to rule it out. Will someone take a not-too-bad loss to make possible a win in 2016? Good question.

      
        
          

I agree that underdogs should attempt risky plays. However, since candidates have only noisy signals of voter preferences and limited means by which to influence them, the range of possible electoral votes are going to be more diverse than you suggest. A modest win may be somewhat less likely than either a small win or big win, but not significantly.