There was an interesting article in Politico yesterday, titled [gulp] “Honey, I shrunk the Obama data machine.”* The article discusses next steps for the Democratic data machine in the leadup to the 2013 and 2014 elections. The big question: can the Obama analytics tools translate to the state and congressional levels?
The answer (to paraphrase): “yes, but only some of them.”
When people talk about the #Demdata advantage in campaigns, they’re really talking about (at least) three distinct phenomena. Two translate well to smaller campaigns, the third doesn’t. The dividing line is something that I call the analytics floor.
(1) One of the biggest advantages Democrats hold over Republicans is the rich voter file that Democrats have developed. Republicans are working to build their own national database, to sometimes-comedic ends. That voter file can be exported to congressional campaigns, special elections, governor’s races, etc. OFA alumni like Dan Wagner of Civis Analytics specialize in just this sort of data modeling. Obama invested millions in developing the voter file and built a network of hundreds of experts in combining the voter file with polling data to produce much clearer maps of the electorate. As those experts turn to consulting and expand their reach outward, the price of these services will become more affordable over time.
(2) A second advantage comes in the form of lessons learned through persuasion and turnout experiments. The Analyst Institute was very busy during the 2012 election cycle, running tests to determine what sort of techniques and appeals can best sway undecided voters and motivate disinterested supporters. These lessons in political behavior are transportable from one election to another — if they’ve determined that voter “report cards” drive people to the polls, that’s a lesson that can improve off-year elections as well. Democrats have invested in cutting-edge social science, and are in no rush to share their findings with Republican competitors. This advantage will echo into 2014 and beyond.
(3) The third facet of #DemData is what Daniel Kreiss calls “computational management.” Computational management refers to the day-to-day role that analytics can play in campaign management, and the “culture of testing” it promotes. The Obama campaign tested everything. It tested e-mail subject lines. It tested font sizes. It tested niche television spots. Data settled arguments and maximized investments. Here’s where things get dicey.
Day-to-day inputs aren’t going to be available and/or useful to smaller campaigns the way they were to the Obama campaign. If you’re running a mayoral race in Hartford, there will be one or two polls conducted *at most*, and they’ll probably come from a relatively unknown firm. That’s exponentially less data than the Obama “cave” was working with. If you’re running a Rockville City Council race, there may be no polling available. And the number of people visiting your website/receiving your emails/reading your tweets is so small that you can’t run tests to find out which messages/frames/asks are most effective. You need scale for computational management. The analytics floor is the dividing line between large-scale and small-scale.
Computational management is a solution to large-scale problems, though. Honestly, running for city council just isn’t that complicated. Talk to your neighbors, earn the endorsements of community leaders, place a table at community events. The districts are small enough that you will mostly be relying upon personalized political communication anyway. The Hartford mayoral race is a bit more complicated, so data and modeling play a modest role. Think of that as a rule: as we increase the size of the electorate, the power of the office, and the (resultant) money being spent on the election, the size and complexity of the campaign apparatus increases as well.
The Obama campaign’s biggest managerial innovation was using multiple forms of data to improve decision-making in this complex environment. Analytics is a solution to the problems introduced at massive scale. Below the analytics floor, the tools are less useful, but they’re also less necessary.
*C’mon Politico, a 24 year-old Rick Moranis reference? You’re better than that.