What's driving vote preferences in the 2014 midterm elections? Plenty.

That's not always the case. There are many elections in which a single issue has dominated. The economy's the prime example; a war or other crisis is another. This year, it's a bunch of stuff.

We've modeled motivators of 2014 House vote preferences in an analysis using the latest ABC News/Washington Post national poll, produced for ABC by Langer Research Associates. First we assembled a range of seemingly disparate concerns and assessed whether they hold together as a single construct. They do, indicating that despite the different topics, they reflect a broader sense of worry about key issues the country faces and its ability to deal with them.

See PDF with data tables here.

Those items include presidential disapproval; dissatisfaction with the political system overall; economic concerns (three of them - feeling worse off under Barack Obama, saying Americans' standard of living is in long-term decline and worrying about the economy's future); worry about terrorism; worry about the Ebola virus; and opposition to last week's Supreme Court action allowing gay marriage in additional states.

We put the three economic items into a single index (alpha = .71, for wonks) and then put the whole shebang into an overall index of dissatisfaction (alpha = .77, still for wonks). (Alphas like these mean the elements hang together.) We then used this index in a statistical analysis, called a regression, to see whether it predicts House vote preferences among likely voters when other factors are held constant. Those other variables are age, gender, race, education, income, region, being an evangelical - and, most important, political partisanship and ideology.

The resulting model is a powerful one; it explains 67 percent of the variance in vote choices.

Partisanship usually is the biggest single factor in such models, and so it is here. But there's another item that comes in right beside it - the discontent index. Remarkably, it's almost as strong as political partisanship in predicting vote choices, and about twice as strong as any other element of the model.

It's worth noting that it's not presidential approval or economic discontent alone that are driving the index's explanatory power - it remains a significant predictor of vote preferences even when both of these are left out of the model. At the same time, presidential approval and the economy are substantial contributors, as customarily is the case.

We can learn a little more about all this by looking separately at those who score in the top third, middle third and bottom third on discontent in our index.

The numbers are striking. In the low-discontent crowd, 84 percent favor Democratic candidates for House. Among high-discontent folks, 81 percent favor Republicans. The deciding group is made up of likely voters in the mid-range for discontent - and they favor GOP candidates by a 22-point margin.

The mid-level discontented also are far closer to their high-level counterparts in thinking the country is headed seriously off on the wrong track. They're also somewhat closer in anti-incumbency, defined as being inclined to look around for someone new to vote for.

Demographic results are interesting as well. Women score high on the discontent index; so do conservatives, Republicans, whites and people who don't have a college degree. Groups lowest in discontent include liberals and Democrats.

(Langer Research Associates)

That may make it look like the index just describes ideology and partisanship, the two horsemen of political attitudes in this country. But it doesn't. As noted, the index independently predicts vote preferences - about as powerfully as partisanship, and twice as strongly as ideology. It means that what's motivating votes this year is not one thing, but a lot of them, working together, under the unifying banner of discontent.

TECH NOTE - A wise friend once said, "We could talk about regressions. But then we'd be… talking about regressions."

So to be brief: What we're reporting here is the result of a linear regression. A stats professor would mark us down for that - technically a logistic regression is preferred in a case like this, because its structure better accounts for the comparative lack of variability in a binary outcome. However, for our purposes it makes no difference, and linear regression is easier to work with, interpret and explain. So that's how we roll.

(Langer Research Associates)