Fighting Fake News with Data Visualizations

Almost a year ago, our country was waiting wide-eyed and wondering who the next leader of the United States would become. This election in particular was unlike any from the past. On one hand, we had Hillary Clinton, a well established politician who graduated from Yale’s School of Law and went on to serve as the First Lady of the United States. On the other, we had Donald Trump, a New York real-estate billionaire who gained popularity from his successful businesses and reality-TV series.

On November 8th, 2016, the United States experienced many emotions as the country’s electoral map became more red and less blue. By the end of the night, it was finalized that Donald Trump had won the presidential election and would become our 45th president. To many, this was shocking and upsetting. For months leading up to the election, news sources and citizens of the U.S. relied on forecasted prediction polls to see where their favorite candidate stood in winning. One site, FiveThirtyEight.com, specializes in these pollster projections and on the morning of election day this forecast was released:

Sept18_Blog_Post_ElectionMap.png

In this model, we see which states are predicted to vote for whom, giving Clinton a 71.4% chance of winning and leaving Trump with a measly 28.6%. This also means that Hillary was expected to walk away with 302 of the electoral votes, surpassing the need for 270 to win. Now, almost 10 months into Trump’s presidency, we know that these outcomes were very wrong. As we close in on Donald’s first year in The White House, many people have started to ask themselves, “Should we have trusted pollster’s election projections?”.

Now 71% is significantly large than 28%, so the average person may walk away from seeing that graph thinking, “Hillary has this in the bag.” But let’s reframe this a bit:

This particular poll prediction was based on a simulation using poll data. FiveThirtyEight simulated the 2016 election outcomes thousands and thousands of times. The result? Clinton won 71% of the simulations, which was then interpreted as a 71% chance of winning the election. There are only two possible outcomes, right? Either Hillary wins, or Donald.

Let’s look at another example of chance: flipping a coin. If you were to toss a nickel into the air two times, you are left with four possible outcomes:

1 head, 1 tail

1 tail, 1 head

1 head, 1 head

1 tail, 1 tail

 

If you repeat this experiment enough you have a 50% chance of getting heads and tails, 25% chance of getting two heads, and a 25% chance of getting 2 tails. Now let’s say Hillary is heads and Donald is tails, that leave us with a 50/50 chance of either becoming president. So, after all of this, are we really surprised that Donald Trump won the election?

Perhaps a bit of the confusion came from how the information within the poll data was conveyed to FiveThirtyEight’s audience.

A good data visualization and infographic can help convey even the most complex statistical concepts to anyone. In the visualization above, we are left with little information on how these numbers were created and the probable chance that the results may yield the complete opposite (which in fact, they did).

So what makes a good data visualization? Well according to the Harvard Business Review, the three key elements to a having a good data visualization are understanding your audience, setting up a clear framework, and telling the story. In this case, our Clinton/Trump graphic fails these requirements and ended up misleading a lot of its viewers. If the graphic on FiveThirtyEight.com would have an included an additional visualization with the other possible outcome, perhaps viewers would have felt less cheated.

If you’d like to see an example of great data visualization, I recommend checking out this application called “The Parable of they Polygons”. This imaginative, interactive site demonstrates how small individual biases can contribute to large societal consequences. With its controllable features, it allows viewers to see multiple sides of the story, rather than just one perspective.

With more informative visuals and a better understanding of how polling projections work, citizens could look at these election predictions, like the one from FiveThirtyEight, and make a more knowledgeable conclusion. In a world where the internet is our main source of news, it can be difficult to not to believe the first thing you see. Keep in mind, when reading messages, consider both perspectives.