[Episode 4] The 47% anecdote

In a private fundraiser during his 2012 presidential bid, Mitt Romney was filmed saying that 47% of the voters “will vote for this president no matter what”. And beyond the political backlash that this quote entailed – mostly because of something else that he said on Obama supporters, that he would “never convince them they should take personal responsibility and care for their lives” -, Mitt Romney is actually right: political campaigns are targeted towards those people that they can convince.

The question I want to ask here is if that is in line with our ideals of democracy, and whether or not there is something that we can do to fix this.

Maximizing the likelihood of being elected = Maximizing the votes-gained per dollar-spent

“The all-encompassing goal of political campaigns is to maximize the probability of victory” say Nickerson & Rogers in their 2014 Working Paper on Political Campaign and Big Data. Perhaps even more so in the United States than in France, presidential campaigns are about finding the people that your message can convince and delivering your message to these guys.

Presidential campaigns crave personal data. It is the only way for them to engineer this kind of microtargeting and lead an economically efficient campaign. “Obama’s campaign began the election year confident it knew the name of every one of the 69,456,897 Americans whose votes had put him in the White House” states this MIT Technology Review piece by Sasha Issenberg.

As impressive a piece of work as this is, it is only a third of the work you have to do, according to Nickerson & Rogers. Given 1) the people’s observed behaviors, you have to predict accurately 2) whether or not they are likely to support you and 3) how they will respond to your message delivered in a variety of different ways.

And that is a lot of data. This is where machine-learning (see here the Coursera/Stanford course) comes in handy: it is scalable (it’s easier to scale up an algorithm than a team of analysts) and it is analyst-independent (it’s not about the quality of the analyst team – which incidentally are hard to come up with -, but about the quality of the algorithms, which are much simple to build and design).

Of course this leads to very realistic (I would like to call them cynical) visions of political campaigns, involving calculating the return on investment of each of your spending in terms of votes gained. Campaigns are now data-driven, using techniques (like A/B testing as this piece by Wired’s Bryan Christian tells us or spamming via email) that previously belonged in the realm of business.

Wait. Is that a good thing?

In this current system that I just described, the underlying assumption of all campaigns is that, given that they will do better than their opponents once they are in power, they should do whatever it takes to get elected. Email spamming, collecting people’s data, raising more money than your opponent: all of this is acceptable if you end up winning because you will bring more good to this country than your opponent.

Well I, for one, believe that this is a bad thing. In an ideal world, campaigns should be about debating ideas and let the voters decide which of the candidates will bring more good to this country. The amount of money that you raise, the number of emails that you send, the private data that you collect or the quality of your machine-learning algorithm are by no means a measure of your ability to govern for the greater good of the citizens of your country. The number of votes that you get at the end of the election, actually, is. Or should be. So how do we make that happen?

Campaigns will always try to do maximize the likelihood of their victory. Candidates will always think that they will do better than their opponents – or else they definitely shouldn’t run. And in a general election, cheap and easy populistic arguments will always be thrown to gain some votes.

I think the solution is not within the political game, but in the rules of the game itself. Let us cap the budget for a presidential campaign. Let us control the data that campaigns are able to collect. Let us regulate for the campaigns to stay within boundaries that we feel are moral. This is how we’ll make the elections not about the amount of money or data raised, but about the voters’ preferences on the issues at stake in the public debate.



Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s