The Data

(A view from inside the Obama campaign in Des Moines. - promoted by desmoinesdem)

“Don’t be too proud of this technological terror you’ve constructed.” – Darth Vader

The Obama campaign’s data-driven approach to ground campaigning has been hailed as a miracle of 21st century electioneering. The campaign, it is said, uses cutting edge “microtargeting” technology to prospect, classify, woo, and ultimately turn out supporters to vote Obama.

The subject of many fawning press accounts, the Obama database makes educated guesses about the political sensibilities of voters based on their ages, registration affiliations, and their declared allegiances to campaign workers – along with a host of other factors. The entire operation aspires to classify voters on two different spectrums – how strongly they support the president and how likely they are to actually vote. A strong Obama supporter with a sporadic voting record might get a visit from a canvasser to get registered and request a mail-in ballot. A voter the database believes is wavering might get a persuasion call from a volunteer making the case for the president.

The glowing accounts in the newspapers conjure up visions of hip young geeks parked in front of gleaming banks of computers, winning elections with the click of a mouse. Still, amidst these hymns and hosannas, I wonder if the implementation of the data-driven technique is as brilliant as advertised. My field level experience suggests that rather than having bins of voters sorted neatly by candidate preference and voting habits, the campaign has agglomerations of voters that share similar characteristics but are by no means homogeneous.

The sophisticated techniques the press describes (like using data to tailor phone pitches to voters) are things the campaign aspires to but fails to execute properly, at least in my field level view. The irony is that the “old-fashioned” way of doing things – before powerful databases – could possibly be more effective than all that tedious mucking about with low quality data.

An example: In the Beaverdale neighborhood I frequent, we have contacted those the database fingers as allies so many times, some are stale contacts. It is sometimes more fruitful to go to work on Joes and Janes off the street than ask our supporters one more time if they want to volunteer or if they want to vote early.

It is impossible (at least for me) to say which method is empirically more effective: A spirited yet flawed attempt at a targeted, data-driven approach or using more old-fashioned techniques. I can only say that the drawbacks of driving a small universe of voters insane by calling them every day are very obvious at the field level. But even that oversaturation could be just a particularly flagrant byproduct of a successful nanotargeting campaign.

Football coach Woody Hayes used to say, “Only three things can happen when you pass and two of them are bad.” And it’s true, a coach watching the introduction of the forward pass in 1906 might see an interception and conclude, wrongly, that throwing the football is a terrible way to run an offense – too many risks with that precious ball! That coach’s error would be to spot an obvious drawback but miss the advantages that a well-developed passing game offers.

The database aspires to a level of sophistication that is unattainable when its primary users are volunteers who have had perhaps ten minutes of instruction from a harried junior staff person. The data game remains a work in progress for the Obama campaign, but a potent one.

-A cross post from Six Weeks, Six Votes

  • contacting the same people

    too many times is a terrible mistake. It’s irritating to get the fourth or fifth call asking to vote early or volunteer. But on balance the data collection is surely helping more than it’s hurting.

  • Interesting post, by the way

    I’ll bump it up tomorrow morning.

  • repeated contacts

    I take solace that for most of my particular team, the contacts are of high quality – good, experienced people on the phone who aren’t reading from a script. Again, I sincerely hope this database is all it’s cracked up to be.  

    • about the same

      as what marketers dump in your inbox & mailbox. Same algorithms.

    • a good database only matters

      when the race is very close, as it seems to be in Iowa. If Obama were ahead or behind by 5 points, these contacts would be a lot less important.

      My experience doing GOTV in 2004 was that reliable voters who prefer to vote on election day got increasingly angry when volunteers kept calling them, urging them to vote early. Contacting someone’s never missed an election should not be the priority at this point.

  • Diminishing returns?

    I agree, generally, with sixweekssixvotes’ observations.

    I haven’t seen a concerted registration drive targeted at renters who have moved recently, e.g. cold canvassing apartment dwellers or talking to the manager. (It’s too late now.) That might have been more productive than continually contacting the same people.

    We may be finding out the limitations of campaigns. Dumping more and more money into TV ads surely reaches a point of diminishing returns. The same may be true for canvassing and phone calling (but I’m not going to quit).

    • tough crowd

      I haven’t seen a concerted registration drive targeted at renters who have moved recently, e.g. cold canvassing apartment dwellers or talking to the manager.

      Have you ever done this? It’s very difficult. And these days, managers aren’t going to answer questions about residents and the general lay of the land. The logistics of this type of canvassing is no different from ten years ago and no database will change this.

      I have a friend who grew up in Sioux City. She tells me her father ran off with a new girlfriend to Mexico and returns to SC sporadically for temp carpentry work. It’s not efficient for any campaign to track a potential voter like this. Or, there’s a largely rental precinct in Council Bluffs that caters to temp workers for Union Pacific. All of these people, by definition, are not rooted in “community,” so they are not going to give a hoot about down-ballot races. As far as top-of-the-ticket is concerned, the lives of Mitt Romney and Barack Obama are so far removed from their reality as well as the “issues” a canvasser might present.

      About six years ago I canvassed a trailer park on behalf of a candidate that was scrounging for these off-the-beaten-track votes — not in the system, not registered etc. I spent the afternoon talking to drunks with predictable results. Not a single one registered or voted as a result of our cold-calling. My overwhelming memory of the afternoon was watching kids playing in dirt mounds.

      According to activists, a not insignificant percentage of “potential Democratic voters” don’t have a suitable form of identification for voting. If that’s the case, you are talking about people who don’t interact with society the way the average regular voter does and tracking these potential voters is low ROI.

      The progressive perspective is either a cocooned urban liberal or staid suburban. People who cry about twenty minute commutes just don’t have a handle on those who will pick up several times a year to go where the work is. There’s a lot of blogging about the “economically distressed” who will be “receptive to our messaging.” Kind of hard to deliver it when your target is beating the recession by running out on back rent payments, i.e., the “midnight move.” The misreading of the partisanship fluctuations of registered voters comes from a similarly rooted perspective.

      Months ago, there were numerous reports that the Obama campaign was having difficulty locating even previously registered (2008) voters. This has less to do with algorithm efficacy/database quality and more with the bifurcation in American society along economic lines. Over the past 30-35 years, elections have increasingly focused on the the rooted in suburbia. It’s just easier for both sides to squabble over these spoils. These algorithms are heavily dependent of consumer preference models which excludes Americans who aren’t making themselves “heard” via their buying and social habits. They’ve gone overboard on the echo chamber that is social media. The repeat contacts, over and over, are made to those who have made themselves trackable.

      The reality is that making repeat calls/visits to the tried and true probably represents greater ROI than trying to make heads or tails of the highly mobile or those who are simply not interested in engaging with campaigns.

      • very interesting

        “The reality is that making repeat calls/visits to the tried and true probably represents greater ROI than trying to make heads or tails of the highly mobile or those who are simply not interested in engaging with campaigns.”

        You expressed this more clearly than I did in my post. I’m not  as certain as you that repeated contact to our voter universe offers better ROI than chasing some of these apartment voters, but I agree completely it’s a very real possibility.

        On the other hand, people in apartments do vote, albeit not as regularly as homeowners and other people who are highly engaged in their communities. It would be a mistake for any campaign to ignore those people completely.

        Just a week ago, I made it into a complex of studio apartments to canvass (which is somewhere between discouraged and forbidden; I didn’t inquire). The people inside were shockingly insulated from the campaign’s all-out blitz. Nobody called them, nobody hit their doors, nobody left them flyers or mailings.

        It was incredibly easy to rack up the voter registrations and absentee ballot requests. I got 14 of each if memory serves – I’d need to hit 400-500 doors conventionally, maybe even more, to get that many at this point.

        Even though as Albert points out, statistically those voters are more likely to flake, I have to think that reaching them as opposed to doing one more round of “do you want to vote early” to the homeowners was a more effective use of my time as a campaigner.  

        • they vote, but

          On the other hand, people in apartments do vote, albeit not as regularly

          unless things have changed recently, the most annoying development in canvassing w/ ‘new tools’ is the fluctuating (daily) walk/call sheets. The less reliable voters you mention are fluctuating around the daily cutoff.

          More power to you for cold-calling the apt complexes that have seen little activity. The Obama campaign claims sufficient conversion of the historically unlikely/infrequent voter, at least in swing states, for the win, even without the excitement of the 2008 race. We shall see.

  • There is a certain amount

    of inefficiency that is endemic in grassroots campaigning. The key is to use some combination of being more efficient and/or overwhelming the other side with sheer numbers to outhustle the other side too even if they are more efficient.

    I’m hoping the Obama campaign has both factors on its side in this election.  

  • Good commentary, but I bet, and would like to see...

    …someone familiar with OFA’s techniques could offer a counterargument to sixweek’s points.

    Sasha Issenberg would be good, as would someone in OFA post-election when I’m sure they’ll be forthcoming, since Obama’s last campaign will be behind them all.

You need to signin or signup to post a comment.