The $5B Midterm Elections And The Big Costs Of Big Data

Midterm Elections: The Big Costs of Big Data

Politics is not our normal beat here at PYMNTS. But the 2018 midterms has changed our thinking, and for one reason: 2018 will be one for the record books in terms of how much is being spent on the campaign trail.

According to the Center for Responsive Politics, this year will see $5.2 billion spent on the midterm elections overall. That’s less than the $6.5 billion that was spent in 2016, only $2.5 billion of which was spent on the presidential race. That suggests an increase of about a billion dollars on the Congressional races that voters will decide at U.S. polls on Tuesday.

That’s even more remarkable considering that between 2014 and 2016, the spending on Congressional races increased by only about $300 million, from $3.7 billion to $4 billion.

So why the massive uptick in spending?

One explanation, said Sheila Krumholz, executive director at the Center for Responsive Politics, is simple intensity. In highly partisan times, the midterms tend to generate a lot of energy, which translates into a lot of donations and spending activity.

And while that tells some of the story, a billion-dollar uptick in spending isn’t explained entirely by fired-up grassroots donations rolling in $5 to $100 at a time.

It seems, according to reports in CNET, that data is becoming the most valuable weapon in a campaign arsenal when it comes to reaching – and, more importantly, turning out – voters.

And gathering, interpreting and acting on that data – in short, trying to “moneyball” politics – is not cheap.

Cambridge Analytica, and a Host of Firms No One Has Heard of

Before the news broke of Facebook’s Cambridge Analytica scandal, most people thought much more about how well they knew the person they were voting for than about how well the person they were voting for might know them.

But when the news broke that the Big Data consulting firm had allegedly illegally harvested the data of 87 million Facebook users to build custom voter profiles that they claimed had helped the Trump team notch the win in the 2016 presidential election, many Americans suddenly took a great deal of interest in the subject.

Ultimately, Facebook apologized – though they noted that Cambridge Analytica was mostly at fault for disregarding the platform’s rules. In May of this year, Cambridge Analytica announced that it was shutting down. The issue more or less left the public’s consciousness.

But the rather sophisticated tracking, sorting, sifting and use of the troves of consumer data on offer on the web? That is still being done by Big Data firms that sell their insights to campaigns.

For a lot of money.

So How Much Do They Know?

Depending on how sophisticated their data operations are, it is possible your elected representatives know quite a bit about you.

Voters start their information handout to candidates before they ever actually vote: Registering puts their information into a database that each state keeps, as required by the Help America Vote Act.

That data, which is always publicly available, contains a voter’s name, date of birth, address, voting history and jurisdiction. Some states also include data like phone number or race. Access to voter data is also accessible, in all states, to political campaigns or private firms working with or for campaigns.

How the data can be used also varies widely across state lines. Some places, like Florida and Texas, have no rules in this regard. Vermont prohibits sharing election data with foreign governments and agencies (because if there is anything the international espionage community cares about, it is the voting habits of Vermont’s 600,000 citizens). California has rules about how the data can be accessed and commercialized.

But broadly speaking, this data is mostly available if one is willing to pass through the right number of steps – though by most accounts, it is not really useful on its own.

That is why data brokers exist. These Big Data professionals can take the bare-bones data from voter information and match it to the pile of digital information bits consumers leave behind throughout their normal daily online journeys, and flesh out much more detailed voter profiles that include things like property ownership status, marital and family status, likely income and more.

“When we interact with any entities online, they keep logs of our trails,” said Augustin Chaintreau, a computer science professor at Columbia University. “If you visit a site that sells shoes, and then you see similar shoes on another site, that information is shared between firms.”

Companies can also identify you by tracking the browser you use, according to Chaintreau.

But the data gathering exercise isn’t even close to finished yet. Next up is taking that data and matching it with the smartphone surveys that people take on their phones each month.

Firms like PredictWise take those surveys, which occasionally pop up while one is trying to use their smartphone, and collect them into a database, adding about 25,000 new surveys’ worth of data per month.

“We’re literally tracking 250 million Americans,” said Tobias Konitzer, co-founder of PredictWise. “We use voter data to give us details on an individual level.”

Very specific details – not only about what kind of stuff one likes to buy, but also about what level of education one is likely to have, how much of their income they save versus spend and what type of social causes they are likely to support with donations. With that information – and another big assist from AI – campaigns can then feed that data into an algorithm to make all kinds of very educated cases about what kinds of issues a voter is likely to support and how strongly they are likely to feel about them.

And yes, all of this completely legal. There is very little regulation in this area – as of now, anyway.

But Does It Work?

All of that data – gathered, crunched, re-gathered, re-crunched and formed into models – can be a very big investment for campaigns.

“It takes a lot of money to run data-driven campaigns, so that favors rich candidates,” noted Colin Bennett, a politics and privacy professor at the University of Victoria in Canada.

Or, at least, it favors rich candidates if Big Data campaigns actually do what they are supposed to do, and make it easier for campaigns to reach the voters both broadly and on a personal level.

That means better insights into how and where to push marketing efforts, where to send canvassers door-to-door to maximum effect, how best to program and dispatch things like mass mailings or robo-calls, and how to best reach out to a potential voter on social media.

Facebook has recently confirmed that during the next election cycle, they will be able to upload their targeted voter lists and show ads specifically to them. But the social network has also committed to being more transparent about political ads, making it possible for users to see how much was paid for them and who did the paying.

The $5 billion question, however, is not whether Big Data can be used to build incredibly detailed voter profile, but whether it works. Will campaigns, armed with the fat data stacks, be able to develop those truly personal experiences that draw in droves of voters?

And this, though we are not experts on politics, is where we as payments and commerce peeps feel we can provide an excellent answer – and a completely correct one.


Because though Big Data and smart AI is everyone’s favorite magic bullet, we know that whether it’s used to sell a widget or win a vote, three things will always be true.

First, all algorithms are not created equal. If it is fed with badly gathered data or programmed to find spurious correlations, it won’t be a useful tool. As Cathy O’Neil, author of Weapons of Math Destructiontold PYMNTS, no one wants to write a bad or biased algorithm, but a single mistake in reasoning can taint an AI and make all of its results highly suspect.

Second, the best data set in the world is only as useful as how well it is acted upon. If the data tells an organization to do something and they don’t do it right, that’s not the data’s fault – but it does mean the data wasn’t all that useful. Personalized experiences are great when they work, and when they don’t, they tend to feel intrusive and annoying, two things most brands and politicians want to avoid.

Third and most importantly: At some point, the product one is selling has to be worth buying. All the Big Data insight in the world can’t make a bad product a good one, whether it’s a phone, a car or a candidate for office. If a candidate is a terrible public speaker, has bizarre beliefs or keeps kissing hands and shaking babies, no amount of data insight is going to help with that campaign.

But, since everyone has worked so hard, done so much math and spent so much money, it might seem like the only fair thing to do is to show up at the polls on Tuesday (if you haven’t voted early already) and cast a ballot.

Not only is it your civic duty, but there are an awful lot of data nerds who are counting on you to make them look good.


New PYMNTS Report: Preventing Financial Crimes Playbook – July 2020 

Call it the great tug-of-war. Fraudsters are teaming up to form elaborate rings that work in sync to launch account takeovers. Chris Tremont, EVP at Radius Bank, tells PYMNTS that financial institutions (FIs) can beat such highly organized fraudsters at their own game. In the July 2020 Preventing Financial Crimes Playbook, Tremont lays out how.