Parties Square Off In a Database Duel

Voter Information Shapes Strategies

Fourth in an occasional series

COLUMBUS, Ohio -- The Democratic Party already knows a few things about the Victorian-style duplex on West Hubbard Street. It wants to know a lot more.

And so Mark Rutkus and Patrick Harris, armed with sturdy clipboards and cheerful smiles, are going on a data hunt. "Hi, we're here from the Franklin County Democratic Party," Rutkus begins as Linda Houston draws back her front door, tentatively at first. "Can we ask you a few questions?"

A longtime Democrat, Houston greets them enthusiastically. For the next few minutes, the local party operatives ask and she answers: Are you registered to vote? Do you reside at this address? Can you confirm the names of the other people living here? What do you know about the people living next door?

It may seem like basic shoe-leather canvassing, unchanged from the days when precinct captains kept their political machines oiled with up-to-date information gleaned from doorstep and barstool encounters. But in this year's election, there is a hidden high-tech twist. Rutkus and Harris are out to "map" the political demography of this neighborhood, trolling in the service of a quasi-science called "database targeting."

Houston's answers will bounce from Rutkus's clipboard to a computer in the state Democratic Party's offices here, and then 400 miles away to computers housed in the Democratic National Committee's headquarters in Washington.

Like rivulets flowing to rivers and rivers to the sea, this information will join an enormous data torrent streaming toward Washington from all around the country. Houston's "profile" is just one of 166 million -- or one for every registered voter -- that the DNC is constantly updating in a huge digital cache known as DataMart. The Republican National Committee tends a similar information trove, dubbed Voter Vault.

The fight for Ohio, and maybe the election itself, could come down to a battle of these databases.

'Very Powerful Tool'

The 2004 election will be the first presidential election in which both national parties use their database and number-crunching skills to shape their organizing and get-out-the-vote strategies.

Marketers have used databases to target customers for years -- they know enough about your credit history to offer you that low-interest credit card -- but the political world is just becoming acquainted. For several years, largely out of public view, the two major parties have been assembling their infobanks, each with the same daunting goal. By tracking the electorate, and employing ever more sophisticated statistical models through the field called "data mining," the parties and their candidates hope to zero in on who will vote, how they might vote, and how to persuade them to vote for Republicans or Democrats.

"You could ask me about any city block in America, and I could tell you how many on that block are likely to be health care voters, or who's most concerned about education or job creation," said DNC Chairman Terence R. McAuliffe. "And I could press a button and six seconds later you'd have a name, an address and a phone number for each of them. We can then begin a conversation with these people that is much more sophisticated and personal than we ever could before."

It is not quite that simple. Models and databases offer better-educated guesses, not certainty, about what a voter thinks and how he or she is likely to behave. But with enough computing power, enough personal details and the right search features, political database pros say they are improving the efficiency of an array of campaign decisions, including fundraising, advertising and get-out-the-vote operations.

Using little more than an off-the-shelf program and the desktop computer in his Washington office, Democratic consultant Hal Malchow shows how to predict turnout and target pools of supporters. Using the 2002 gubernatorial race in Arizona as his example, Malchow is able to match the poll responses of 5,778 likely voters against their database profiles. The program then slices and dices the data to uncover the characteristics -- in this case, middle-aged Hispanic men living in two metropolitan areas -- that defined the biggest groups of people likely to support Malchow's client but still uncertain about voting. A quick search of a voter database would return the names of those who fit this profile, making them the likely recipients of phone calls or a knock on the door by a candidate's field staff.

"This doesn't improve [a candidate's] message one bit," said Malchow, a direct-mail expert who has been a pioneer in such targeting techniques. "It doesn't change the way a candidate looks or his personality or where he started in the polls. . . . But it can be a very, very powerful tool. In the end, it's about having knowledge that allows you to use your resources in the smartest and most efficient way."