"Democrats discover database canvassing in time for Nov. elections"

ZDNet:

In the final run-up to the November elections, campaigning and canvassing is a technology game. Canvassers use sophisticated databases and modeling software to do "micro-targeting" - identifying individuals who may be likely to cast a vote one way or the other.

The Republicans have been leading the computer revolution in politics for several elections but now Democrats are catching up. The Washington Post reports:

Campaigns have begun identifying potential voters literally one by one, even if they live in areas dominated by the opposition party. Using surveys and modeling and consumer and political data, the parties convert the electorate into subgroups that turnout specialists call by names like "Flag and Family Republicans," "Education-Focused Democrats" and "Older Suburban Newshounds."

This is known as microtargeting, and it turns traditional political mobilizing on its head by giving campaigns the opportunity to create virtual precincts of voters and poach on the opponent's turf.

But, as with most things Democratic, the road to digital electioneering has been sloppier and more contentious than for Republicans. Harold Ickes, who may work as an advisor to Hillary Clinton in 2008, has started a for-profit company called Catlist, which is building a database that will be sold to various organizations and campaigns, including all the presidential campaigns. That will put the range of Democratic institutions on a level playing field. The differentiating factor will be which campaigns know how to get the most out the data.

The private-sector Democratic effort appears cumbersome compared with the centralized operation organized under the RNC. Catalist produces the voter file, but from there the subscribers must manipulate the data to make microtargeting possible. Progressive groups are running experiments in a handful of states, including Minnesota, Michigan, Ohio, Pennsylvania and Colorado.

Emily's List surveyed 10,000 people Michigan and 9,300 in Minnesota. The next step was to build clusters of voters based on political and lifestyle characteristics. The results were then applied to the entire voter file in the states — more than 7 million in Michigan and more than 4 million in Minnesota.

In Michigan, the process produced 14 clusters — three strongly Democratic, one weak Democratic, six in the middle, two strongly Republican and two weak Republican. The modeling cost about $250,000. In Minnesota, the process produced 13 clusters, including the Older Suburban Newshounds and Rural Religious Moralists. The highly informed newshounds make up the largest of the 13, while the rural moralists are one of the smallest ones.

With this data, groups like America Votes can identify Democratic voters in the midst of heavily Republican precincts. "We have a nice chunk of people that we would not normally have known about," tech director Ed Coleman said. "We would not have blindly gone in and called people there, but now we've found a universe, and if we want to we can go call."

Naturally, the Democratic National Committee has their own database at a cost of about $8 million. "We've invested a huge amount of money and time to improve the quality of the data in the voter file," said Ben Self, a veteran of DNC Chairman Howard Dean's 2004 presidential campaign who is directing the technology effort for the party.

But Republicans still have the advantage in the datamining race. Democratic efforts this year are still seen as experimental. 2008 is the big-time and Dems hope to be ready. "We had huge voter turnout in our heaviest base areas," Coleman said. "But we cannot count on that turnout to carry the day. We're going to have to do better in these fast-growing counties outside the metro area, and this is one tool we're going to use to try to do that."