If you're a Kentucky Democratic Party official still reeling from the results of last year's U.S. Senate race, you might want to look away.
Forty-three percent of Democrats who voted pulled the lever for Mitch McConnell last year, according to an analysis done by the U.S. Senate majority leader's longtime pollster, Jan van LoHuisen. In some counties, that number was close to 70 percent.
McConnell's analysis of voting trends gives him a much larger advantage among Democrats than did exit polling conducted last November, which showed 19 percent of Democrats voted for the Republican incumbent instead of his Democratic challenger, Kentucky Secretary of State Alison Lundergan Grimes.
McConnell's team, looking at voting data posted by the Secretary of State's office, drew their conclusions using a method called ecological inference — the process of using patterns in aggregate data to make inferences about individual-level actions. Van LoHuisen said he relied on statistical procedures and software designed by Harvard professor Gary King to enhance the accuracy of ecological inference.
Although wildly complex, McConnell's team believes its calculations are more accurate than exit polling, where a small sample of voters, having just cast their vote, might tell a pollster they are a Republican when they are actually registered Democrats.
Josh Holmes, McConnell's top adviser during the campaign, said there are two takeaways from van LoHuisen's analysis.
The first is that the unique voter targeting and turnout strategy Holmes and his team deployed to attract conservative Democrats worked beyond their wildest expectations.
The second is that a large number of registered Democrats in Kentucky no longer think of themselves as Democrats.
"You've got a whole bunch of registered Democrats that don't even refer to themselves as Democrats in exit polls, which I think is even more alarming (for the Democratic Party)," Holmes said.
David Bergstein, the new spokesman for the Kentucky Democratic Party, was less than impressed. He said the McConnell camp's statistical conclusions will have no bearing on this year's gubernatorial race.
"Kentucky voters view state and federal elections very differently, so trying to draw a lesson about 2015 from 2014 is like comparing apples to oranges," Bergstein said. "For instance, Kentucky Democrats like Gov. Beshear and Attorney General Conway won commanding victories with strong Republican support in 2011 after a bad campaign cycle for federal Democrats — and even in a good year for Republicans they failed to win the state House."
As the gubernatorial race gears up this year, Republican candidates are locked in a battle with each other for now. But Holmes said the program McConnell's campaign created and executed could be duplicated.
"It can be replicated but it takes an enormous amount of commitment by the campaign to see it through and resources to fund it," he said. "This is not one of those things you can throw together in a month and a half and hope it works."
Beginning in the summer of 2013, Holmes and his team began building their field operation, focusing not on turning out as many Republicans as possible, but instead on targeting and turning out conservative Democrats.
"Normally, you start with the partisan lines and the registration and you try to turn out as many likely voters in your party as possible," Holmes said. "We turned that on its head a little bit."
A team of volunteers, armed with tablets and smart phones, began working door-to-door to identify voters who would fit with a program they called "Obama wedge."
As the volunteers canvassed, they found voters who agreed with McConnell on the issues, and perhaps more importantly, disapproved of President Barack Obama's administration.
"A good sign is if you've got a Democratic voter who immediately said they didn't approve of President Obama," Holmes said. "That told us they were open for business."
As the volunteers recorded what they found in the field, the information was updated to a database in real time. McConnell's campaign stopped looking at voters as Republicans and Democrats, focusing instead on voters who agreed with McConnell on the issues and disagreed with the president.
"We knew we were looking for conservative Democrats," Holmes said. "What I don't think we anticipated was the extent to which they were available."
With "McConnell voters" identified regardless of party, the senator's campaign was able to specifically target those voters with mail, television, digital and radio ads and talk to them directly about the issues they said mattered most to them.
The campaign, through differing media tailored to different voters, stayed in constant contact with them over the course of the campaign, and in the final days of the election, those Democrats were who McConnell's team focused their turnout efforts on.
The conventional wisdom in the days following the campaign was that Republicans had simply turned out in larger numbers than Democrats, propelling McConnell to a landslide victory.
But turnout was actually down for both parties, meaning that it was Democrats who supplied McConnell's nearly 16 percentage point margin.
In 2010, 52.3 percent of Republicans voted in the U.S. Senate race. Last year, that number fell to 47.9 percent. Democrats in 2010 voted at a clip of 49.3 percent. In 2014, that number dropped two percentage points to 47.3 percent.
"I don't know what the Democratic turnout would've been on Election Day if we hadn't turned out McConnell Democrats," Holmes said. "It could've been historically low."