What’s going on in these final week election polls?

What in blazes is going on with these swing state polls? Clinton has stabilized a narrow lead in national horserace polls after a week of uncertainty following the FBI’s discovery of emails. However, various swing state polls are still showing erratic results that are unlikely to settle down by Election Day.

The uncertainty in polls is a by-product of the way swing states have transformed the electoral system for the presidency, and now also control of the Senate. Since 1988, nine states have oscillated between the parties, are usually decided by under five percentage points, and are bellwethers that usually swing with the national result. These states are Colorado, Florida, Iowa, Nevada, New Hampshire, New Mexico, North Carolina, Ohio and Virginia. The candidates visit these states, set up voter contact operations there, and bombard these states with advertising the week before the election. The list of swing states is short and is very stable, although as I have argued elsewhere, Wisconsin, Arizona and Georgia may soon start swinging. Virginia, Ohio and New Hampshire may soon exit the swing club.

In the other 41 states, this election is basically over, and voters are reduced to anxious spectators awaiting the verdict of the undecideds in the swing states. But it wasn’t always this way. Of course the Electoral College means the election is decided in the states, but more states used be competitive. Between 1924 and 1960, only ten states were consistently uncompetitive: Arizona, Arkansas, Georgia, Kansas, Nebraska, Oklahoma, Utah, Vermont and Virginia. The 1960 election, determined by 0.2 percent nationally, had 20 competitive state elections decided by under five percentage points, with 16 blowout contests with a 10-point or more winning margin.

We are in a new phase of American political history, where elections are decided by a dwindling number of tightly competitive swing states. In each successive close national contest since 1960, the number of truly competitive states has decreased, even in close elections.

Obama defeated McCain in 2008 by a solid 7.2 percent nationally. Yet 35 states were blowouts and only six states were closer than five percentage points. Obama won in 2012 by a tighter 3.9 percent national margin, yet there were even more blowouts – 36 in all – and only four close states.

What does this all have to do with erratic poll results in the final week of the campaign? Plenty.

In a political world dominated by swing states, the decisive swing voters in these states are all-powerful and hard to survey. Swing states do tend to have far more independent voters than other states, but independents still have implicit leanings toward one of the major parties. They tend to “come home” in swing states when exposed to intense grassroots campaign activity, saturated with campaign news and bombarded with advertising. In practice, swing states actually have a smaller percentage of true swing voters than other states. Even when the candidates are disliked or there are viable third party options, there are still not many persuadable voters at this stage in the campaign.

Some swing voters are angry and alienated. Others are young. Still others are ethnic or racial minorities whose motivation to vote is hard to assess. Blacks had record high turnout in 2008 and 2012 because of the Obama candidacy, and it difficult to know how high turnout will be this year. Latinos, especially the young, show signs of high turnout tilted toward the Democrats as a vote against Trump, but we don’t know how much nor exactly how many will turn out.

That is why we see polls giving us differing snapshots of the swing state “likely voters.” Each poll must make an educated guess to predict which registered voters will actually cast a vote on election day, and whom they will support. Not all polls use the same likely voter “model” to make that estimation. Many use a conservative model based on screening questions to measure each individual’s level of intention or enthusiasm to vote, degree of attention being paid to the campaign, and past history of voting. Some polls in addition attempt to guess which demographic groups are more or less likely to turn out this year compared to other years, and then weight these groups accordingly. Some polls will consider first-time voters, others will not. With swing voters in flux, determining the likely electorate is more art than science.

Under such conditions, we shouldn’t be surprised that the swing state polls are varying widely.

If you really want to know which swing states are extremely close, don’t just watch the polls, watch the campaigns. They have state-of-the art techniques to identify and mobilize those elusive swing voters. See which states the candidates and running mates visit in 48 hours before Election Day. Watch where the advertising dollars are poured.

Polls in past elections have missed the mark on predicting likely voter turnout in recent tight elections. Some polls will get it wrong on Election Day 2016 too. The trouble is, you don’t know which ones are wrong. Make sure to stock up on antacids, November 8 might be a longer night than expected.

McLean is Professor of Political Science at Quinnipiac University. He is a contributing author in Presidential Swing States: Why Only Ten Matter, edited by Stacy Hunter Hecht and David Schultz (Lexington Books, 2015).

The views expressed by Contributors are their own and are not the views of The Hill.