the 2012 presidential race

A short note, and then a long note.  I have not blogged in a month; how then have things changed in the presidential race?  In September, President Obama was trending to run close to his 2008 numbers, perhaps a point or two behind and winning somewhere between 332 and 347 electoral votes, in essence, one less state than 2008 (Indiana).  Over at 538, Nate Silver notes that the race has returned to where it was in July; I suppose it is possible that the race has moved back and forth since then, with Obama moving forward on his convention bounce and Mitt Romney’s remarks on the 47%, and then back down again after his Denver performance.  Perhaps it’s also possible that the race hasn’t moved at all.

Pollsters, like political reporters, want political horse races to be dynamic and exciting.  Sure, Secretariat’s great win in the Belmont would still be remembered, just as Nixon’s victory in 1972, but if the win in the Belmont had not been for the Triple Crown, it would be far less famous.  Beyond the interest that in polling and stories that comes from an exciting race - and that interest does result in money; the effect is not trivial - beyond that though, people who pay for polls are far more interested in polls that show movement.  Every poll has a margin of error.  It is to be expected that two polls on the same race will diverge and that several polls are needed to establish a baseline with confidence.  And while I am skeptical about the number of undecideds in a presidential race, most political races do have a capacity for movement, even late in the race.  It is a kind of Pascal’s wager:  a poll that (correctly) damps down the swings of enthusiasm and shows a steady lead is more valuable than a sensitive poll that shows somewhat improbably swings in enthusiasm.  The damped-down polling might be fine in a presidential race where things really are unlikely to change, but in a senatorial race, the consequence of polling that fails to spot a trend against you is an unexpected loss; the consequence of a too-excitable poll is sleepless nights and extra campaign work.

That the Obama lead largely disappeared after the first debate, seemingly due to  Obama’s disastrous performance but perhaps due to an expected decline, at least according to economic forecasters.  His lead though, however much smaller, is still a lead.  The results, at least by some pollsters, showing a national lead for Romney, are interesting but as likely to be wrong as they are irrelevant.

A victory in the national popular vote is irrelevant in that the presidential race is decided by the Electoral College and unlikely in that the Electoral College does respond to changes in the popular vote.  It is dragged along, so to speak, and not completely disconnected.  Obama has consistently maintained a lead in the Electoral College in the last month, sometimes dipping below 280 votes, sometimes creeping back over 300.  I think that an Obama victory in the popular vote of 2% to 4% ought to translate to 303 electoral votes; between 4% and 6% should add Florida.  Ohio and Nevada, with 24 electoral votes, seem to his surest route to victory, even if he barely wins the popular vote.  Iowa and New Hampshire are likely very close as well, with Colorado and following next.  In 2008, Obama won by more than 7% and won in states that no one expected him to win six or eight months before.  His popular vote numbers have consistently polled lower in 2012; if that holds up, then his election totals are likely to be down in most states.  In a state like California, that means a win by smaller margins.  In a state like Indiana, that means he loses a state that he previously won.  That several national tracking polls have shown leads of several points for Romney is improbable because Obama has consistently lead in Electoral College projections, and on sites that project every (or most) states, usually exceeded 270.  Why do I say that the state polls showing electoral victory are likely to be more accurate than national polls?  Simply the number and general consistency of those polls:  that after all is one the principles of statistical analysis, that confidence grows as population size grows. 

And I’m not even going to try to get in pollster bias.

Updated:  a link as to the math behind preferring the state polls