Getting ready for tomorrow’s “Data-Crunched Democracy” conference, political “big data” has been very much on my mind — look for several data-related articles in the next
two or three few days.
But before we start digging into the big trends, let’s step back and look through a slightly skeptical lens, at least for a moment. For one thing, if you listen to Obama data guru Ethan Roeder, the 2012 campaign’s much-lauded program to microtarget political communications via data analysis wasn’t quite the flawless machine I’ve heard it described to be. “Duct tape and baling wire” held a lot of the structure together, Ethan said on a CampaignTech conference panel in April, meaning that people were still sorting some data by hand, printing out spreadsheets and relying on other labor-intensive practices the campaign’s data-integration effort hoped to avoid.
That “duct tape and baling wire” line puts the power of political data into a broader context (i.e., NO technology works perfectly), and it’s a good antidote to some of the more-breathless speculation about the future power of big data to let manipulative politicians warp the political process through hyper-targeted communications. I.e., if even the campaign widely acknowledged to have a ground-breaking data strategy STILL had to apply band-aids as it went along, how efficient will day-to-day campaigns be at delivering customized messages?
I’d argue that much of potential of data-driven communications to change our political process in some fundamental way is still theoretical — it’s an extrapolation into the future, not a crystal-clear vision of how things will be. For instance, we know that say, targeting a supporter who’s responded to your past global-warming emails with a climate-themed fundraiser is likely to yield better results than a less-targeted appeal. And, we know that the Obama campaign’s voter scoring system did a good job of identifying people likely to support Obama but less-likely to show up at the polls (we know because those people did, in fact, show up at the polls after being aggressively courted by the campaign).
But it’s a long way from these practical applications of political data to some sort of grand-scale slicing-and-dicing of the electorate to allow a politician to deliver individual messages to every individual voter, each message designed to persuade that one person. For one thing, as a panel of data practitioners pointed out at a 2012 CampaignTech panel I attended, you quickly tend to run into practical limitations: to take advantage of the ability to cut your list into 20 demographic segments, you’ll need a staff big enough to produce unique persuasive content for each of those segments (otherwise, what’s the point?). Likewise, if you apply too many demographic criteria (upper-income women in Texas Congressional District 5 who are aged 30-45, with children, are registered Republicans but who haven’t voted in the past two elections), you can end up with target universes that are too small to be useful, a great example of diminishing marginal returns. I.e., if you over-target, you can end up targeting nobody.
Finally, too much targeting can also leave an opening for a campaign’s opponents. For instance, Romney’s campaign talked a lot about microtargeted communications, but they seem to have had in mind reaching out to the same pool of voters as usual but with better messaging. Obama, by contrast, EXPANDED the electorate, in part by using data to identify people who were being left out…and then sending actual humans to talk to them one-on-one to persuade them to help the campaign.
In the end, I suspect that we as information consumers will develop countermeasures — conscious or unconscious — to help us block out much of the targeted messaging headed our way, just as we currently block out most TV and print advertising. Yes, big data may well transform the way we do politics in the U.S., but I suspect that the more dystopic predictions, (which at their most extreme involve a nation of drooling automatons spoon-fed information through targeted communications) won’t come to pass…if for no other reasons than practical ones.