With a deadlocked popular vote all eyes turn toward the electoral college, where the conventional wisdom holds that the state of the race is pretty straightforward: Romney leads in North Carolina and Florida, but Obama leads in Ohio, Nevada, and Wisconsin, states worth 271 electoral votes. As an advocate of the polling average as the principal means to understand the state of the race, I don’t fundamentally disagree with this view. But this year’s battleground states are so close and so diverse that it would be wise to remain open to unexpected outcomes.

That's because the battleground states are all so different demographically. States once deemed microcosmic of the country like Michigan, Pennsylvania, and Missouri have been swept off the map for being a few points too Republican or Democratic, and they’ve been replaced with a diverse set of states that add up to a tight race. “New coalition” states like Virginia and Colorado hold exceptionally well-educated and affluent populations. More than 20 percent of the Virginia and North Carolina electorates might be black in 2012—far more than any battleground state of 2004. Nevada and the once dispositive Orlando area have been dramatically reshaped by an exploding Latino population. Northeast Ohio is full of manufacturing workers, something absent from many other battleground states.

Today, the national polls show a truly deadlocked race, and the nine battleground states each appear to rest within about three points on either side of a tie. With diverse battlegrounds, there are plenty of moving pieces that could potentially nudge a state one or two percentage points in either side’s direction without influencing the other battlegrounds. For instance, if Romney sweeps undecided and disaffected white working class voters, it won’t do him much additional good in Virginia but could make the difference in Ohio. If Obama’s GOTV efforts rejuvenate Latino turnout, it barely matters in Iowa but probably ends Romney’s chances in Nevada. If Mourdock comments cause a new round of abortion coverage, it might make a difference in Colorado without any consequence for Ohio or Florida. A strong black turnout wouldn't swing New Hampshire but could be decisive in Virginia.

Making matters worse, several of the most important questions of the election are issues that pollsters struggle with the most. While the average of polls is the best available indicator of the state of the race, the polls are neither perfect nor infallible. The RealClearPolitics average, for instance, usually misses its mark by an average of 2 or 3 points, even without systemic bias toward either party and even in years when it gets all the states right. And unlike unscientific criticisms of polls (ie: the party-ID wars), there are many substantive and defensible criticisms of the continuing accuracy of polling in an era of low response rates, including their ability to reach out to Latino voters and people with cell phones. For good measure, likely voter screens are imprecise, to say the least. While all of these issues were potentially problematic a decade ago, many were not as serious. Cell phones weren't yet ubiquitous, and an enclave of Cubans in Miami were the most important Latinos to the election. Young voters were not as critical to the Democratic coalition, since the age gap was far smaller than it is today. These issues give more reason to avoid placing too much faith in a 2-point lead in a polling average, but they could also manifest unevenly across the electoral map. If the polls miss Latino turnout in 2012, that potentially changes the outcome of Colorado, Nevada, and Florida without much of a consequence for Iowa or Ohio. And it’s worth remembering that something like this may have been responsible for Reid and Bennet’s upset victories in 2010. 

Support thought-provoking, quality journalism. Join The New Republic for $3.99/month.

None of these scenarios are particularly probable, especially individually. But one way or another, we could easily be surprised between now and the time the election is called in 13 days. The polls are pretty good, but they are not perfect, and with observers paying so much attention to the slight distinctions between Obama's 1.9 point lead in Ohio and .6 point lead in Virginia, unrealistic levels of precision may be necessary to avoid surprises. And that's before accounting for the possibility that the race could shift over the final two weeks in subtle ways that move particular demographic groups and states without similar changes in others. The nine battleground states are so close and so diverse that late movement among specific demographic groups or slight errors in the polling could easily reshape the electoral map before November 6.