Political Parties Reap Data Mining Benefits
Political Parties Reap Data Mining Benefits
WASHINGTONBoth major U.S. political parties entered into a new era in the way they use the information they collect on voters throughout the United States.
Using new levels of sophistication in how they collect, clean and manipulate the information in their massive databases, the parties are able to predict with reasonable accuracy how individual voters will vote when they get to the polls, and to target voters individually where needed.
For example, the new level of sophistication allowed the Democratic National Committee to target specific voters in otherwise solidly Republican districts to get them to the polls on Election Day.
In the past, parties would target localities or districts with voter turnout efforts, which then failed to reach some voters because they lived areas that werent targeted.
While both parties have been doing some measure of voter targeting for the last decade or so, only now have they had the ability to manipulate the vast databases of voter data quickly enough to use it tactically in campaigns.
"One of the things that changed dramatically is the amount of data that people were able to use and have confidence in," said Gus Bickford, president of Factorum Productions in Westford, Mass., a company that consults with political operations and nonprofit groups in how they use databases.
Bickford is also an elected member of the Democratic National Committee in Washington.
It was Bickford who was responsible for bringing the DNC together with Intelligent Integration Systems of Boston, which was in turn responsible for setting up the DNCs database solution in time for the 2006 elections.
"The DNC had to change its technology in terms of the size of the data, and in the time you have to work with it in terms of cleaning it and matching it," Bickford said.
"Added to that is the time required for getting it in the field and then getting responses back from the field, and determining the metrics," he said.
Bickford said that one improvement that the DNC made was to move its databases to Netezza, a database appliance thats known for very high speed.
"When youre able to run your queries, the turn around time becomes less with Netezza," Bickford said.
He said that typical questions managers might ask are, "Are these groups the right ones? Are they likely to come out to vote?"
He said that using the voter data in their database, they were able to come up with better answers faster than in the past.
"The DNC is using Netezza to analyze large amounts of data," said Ellen Rubin, vice president of Marketing for Netezza in Framingham, Mass. "Theyre working to get out the vote for the election."
Next Page: Deadline demands.
Rubin said that because the DNC was working with an unmovable deadlineElection Daythe organization depended on the analysis speed of its database so that it could move faster and react more quickly to changing conditions.
"They had over 900 fields for each one of the data records," Rubin said. "The challenge is to understand who these people are, and to figure out what kind of relationship there was in the past, and how to reach them in time for the election."
Rubin said that a lot of work was required to match voter lists with other databases to ensure that contact information was correct and complete.
"You can cause a lot of problems when you have the wrong information," she said.
Making sure the information is actually correct is a big problem.
"We have vast voter databases that we update and keep current and constantly tweaked to make sure they have optimal working value for each election cycle," said Josh Holmes, deputy press secretary for the Republican National Committee in Washington.
"The information we can retain is all valuable in determining voting patterns," Holmes said.
"Our database is not one where we feel that too much information is too much. The more information we can add and overlay is better."
This means, of course, that the data being retained by the Republican Party is as vast as it is with the Democratic Party.
Part of the reason theres so much information is that the parties both retain data that goes far beyond voter registration lists.
"It starts with basic voter registration and party registration data," Holmes said.
"Then there are things like hunting licenses and other publicly available information," he said, "and then theres consumer data information such as magazine subscriptions. Theres no one thing. Its a combination of information that gives you the picture you want to see."
One of the reasons that Bickford and the DNC decided to use Netezzas database appliance is performance.
In many cases, especially with the give and take of a close election, voter targeting at the last minute can pay big dividends.
"It gave them something theyd never had before. We were able to turn [queries] around in a much shorter amount of time," Bickford said.
Next Page: Handling data.
He also said that finding ways to handle large amounts of data, even in a high performance environment remained a problem.
Part of the answer was to call in some people with the right experience.
"The key people had worked on the genomic project and figured out how to use large amounts of data," Bickford said.
"The data is able to be massaged at the field level, and it moved us into the next generation," he said.
Holmes said that how the data is handled also has a big impact on performance.
He points out that standard queries can be performed very quickly, for example.
"It can be done just as quickly as you can sort through a spreadsheet," he said.
"If you need to redo the search item to target a specific area you havent targeted before, then its longer."
Holmes said that exactly how much longer depends on the data being searched for and the complexity of the query.
Both parties have moved beyond the simple targeting of groups.
"The piece that I would stress is the fact that its no longer that youre targeting neighborhoods, wards or districts," Holmes said.
"You used to focus on them. Thats not as efficient as what we have now. Were able to turn out Republican votes, not only in Republican areas, but in areas that dont have high Republican support."
Bickford said he had the same problem: "Who are the people who will vote, but are less likely unless you light a fire under them?" he asked.
He said that in many cases, they would be voters who might have Democratic leanings, but who didnt get the information they needed because they werent identified as Democrats, or didnt live in Democratic areas.
Bickford pointed to one New Hampshire election in which the party needed more than just the voters they always relied on.
So the DNC used data mining to identify new voters and send them mail.
"It was used in a seven-piece mail program," Bickford said. "It targeted older unenrolled men and all women," in other words, voters who were registered to vote, but not registered as being in a specific political party.
"People not registered in a specific party are the swing voters," he said.
Both parties are also continuously working on their databases.
"You can never rest on your laurels," Bickford said. "The majority of the planning you do, and continue to question whether your plans are working or not."
Holmes said that the Republican Party is already working on the databases for the next election.
"There are people who are working today," he said. "Its a 365-day-a-year job, and it doesnt take a break after the election cycle. Theyre already matching database up with the voter outcome."
Both parties drew much of what theyve learned from the business community, and they say they think that their lessons are directly transferable.
Holmes said that micro-targeting is working well for his party, and he thinks it has applications outside of politics.
Just like its important for companies to keep track of their customers wants and needs, he said, its also important to the political parties.
"It allows you to not take a single voter for granted," he said.
Check out eWEEK.coms for the latest news, views and analysis of technologys impact on government and politics.