The data trap

To hear professional political operatives tell it, winning elections is about nothing more or nothing less than mathematical calculations. It’s all about data and it’s algorithmic. You gather all kinds of data about voters, use that data to target those most likely to vote for your candidate, write a formula for reaching your “win target,” plug all the data into your formula, and out pops a victory.

Sounds great, all scientific and everything, until what pops out is a loss. The latest and most glaring example of data gone wrong is the 2016 presidential election. Clinton headquarters had the math all figured out. They shunned “persuasion” campaigning, meaning they didn’t want to waste time trying to win over voters their computers told them were not likely to support the Democratic nominee. They saw it purely and simply as a “base turnout” election. In other words, their data told them that if those identified as core Democratic supporters went to the polls and voted as expected, Hillary Clinton is elected president. In the places that mattered most, places like Michigan and Wisconsin and Pennsylvania, that didn’t happen.

What they didn’t factor into their equation was Clinton’s unpopularity and her inability to persuasively communicate reasons to support her. That left her base unenthusiastic and her opponents energized.

This is not the first time voters have confounded the political mathematicians armed with all their data and their computers, nor will it be the last. In 2014, I repeatedly heard from Democratic operatives in Wisconsin that if turnout was high in the election for governor, Mary Burke would win, and if turnout was low, Scott Walker would be reelected. Voter turnout ended up being a record high for a regular election for governor in Wisconsin, and yet Walker won.

Like Team Clinton in 2016, Wisconsin Democrats concentrated on turning out their base for Burke in 2014. If their computers said you were a likely Burke voter for one reason or another, you were hounded. You got phone calls, you got emails, you got texts, you got junk mail, people knocked on your door. You got so many reminders to vote that you were ready to scream. If the Democratic algorithm didn’t have you down as a target, you were left alone. You were given no reason to think about voting for Burke. Turns out their algorithm was wrong.

There’s good reason why political algorithms are unreliable. Elections aren’t algorithmic. Politics is more art than science. How voters make decisions can’t be reduced to mathematical equations or scientific formulas. There is a human dimension computers can’t account for.

Elections are about representation. Voters are looking for someone who gets them, someone who is saying what they are feeling, someone who reflects their own thinking and will be at least somewhat likely to act accordingly. They look at candidates differently than computers do. They look at who a candidate is, where they’re from, what they stand for. They look for someone they can relate to, someone they feel a connection with.

No algorithm can be written to produce that.

Mike McCabe