Election graphics can be deceiving. Here’s how to see through their tricks.
So, we meet again, dramatically zoomed-in bar chart.
Here at Popular Science, we don’t make political graphics. We’re not concerned with who’s winning in the polls. But we do care that you understand the polls and maps you’re going to see practically every day from now until well past Election Day.
There’s plenty of criticism on the internet about “bad” election graphs, so we’re not going to spend time bashing one outlet or another. While there are some hideously ugly images out there, most graphics aren’t actually good or bad—they’re just being used for something they’re not designed for.
You wouldn’t use a sledgehammer to hang a picture frame, and you shouldn’t use a map when a cartogram would work better. Once you understand the tools designers are using, you’ll find it much easier to figure out when they’re doing their job well—and when they aren’t.
Maps: Acres don’t vote, people do
Aside from poll results, most of the election graphics you see are probably basic maps showing which US states are likely to go Republican or Democrat. Maps like this are great in that they are incredibly familiar—if you want to know who Iowa or Michigan might vote for, you know exactly where to look.
The problem is that these maps are designed to be geographically accurate. They’re great at that, especially if they account for the curvature of the Earth with a proper projection that doesn’t stretch out the northern states. But as Alberto Cairo, a data visualization designer and a professor of visual journalism at the University of Miami, has pointed out: acres don’t vote, people do.
Many of the largest states by area have relatively low population densities, and thus fewer people and fewer electoral college votes. Geographically accurate maps therefore overemphasize the electoral importance of physically big states.
Cartograms, however, do a better job of displaying electoral votes. Despite the fancier name, these are just maps that usually use simple geometric shapes to represent states rather than their geographic borders. Well-designed political cartograms use one block to represent each electoral college vote, so states with more votes are bigger.
Some people don’t like cartograms because they’re unfamiliar and can seem difficult to read at first glance. It’s hard to know where exactly to find Iowa or Georgia, for instance, when the location of states is a bit jumbled. It does take some getting used to, but the more designers create cartograms, the better we’ll all get at reading them.
That doesn’t mean geographic maps are useless, though—there are ways to make them more accurate.
One is to superimpose bubbles sized by the number of electoral college votes, as The Economist has done. Doing so keeps each state in its expected location while still giving the viewer a clear idea of the number of electors at stake.
Another is to make a geographic map in which every voter is represented by a single dot. This keeps all the voters in their geographic space and manages to avoid the electoral college entirely in favor of displaying the popular vote.
Bar charts: At baseline, check the base line
America has an obsession with polls, so during election season (which now seems to last all four years) you’ll see plenty of bar charts comparing how many people say they’d vote for one candidate or the other. This means most of what you’ll see are basic charts showing percentages. But often, those charts will be strategically cropped to give you a certain impression.
Take a look at the three graphs below. All of them show that Candidate A is winning, with 49 percent of the vote, while Candidate B trails behind with 45 percent. But the three graphs tell slightly different stories.
The first one seems to show a substantial lead for Candidate A, and then each subsequent graph depicts what feels like a smaller and smaller lead. If you look at the vertical axis, you’ll see why. Graph No. 1 limits the axis to show the range between 40 and 50 percent, which makes the four-point gap appear much bigger. In the second graph the gap is smaller, because we’re actually seeing all the way from zero to 50 percent. This is why it’s important to show the baseline—or zero—on a bar chart: any amount of zoom can artificially amplify a difference.
Graph No. 3 shows the smallest difference since it’s zoomed out all the way to 100 percent. Whether the second or third graph is better is subjective—few candidates get anywhere close to 100 percent, so it may be more relevant to show only up to 50 or 60 percent so that the difference between candidates isn’t too hard to see. It just depends on what the designer wants you to take away from the chart. And that’s what you should be aware of here: what you see in a graph reflects what its maker wants you to see.
If they want you to focus on one side versus another, they’ll show you just the voters who say they want Candidate A or Candidate B. Though that’s just one part of the story. Another choice would be to show you just how many people didn’t vote in the last election, which tells a much different story:
Line graphs: Cloudy with a chance of uncertainty
Polls are often represented as binary and absolute, declaring that a certain number of people would vote for one candidate, and another amount would vote for the other. But, of course, there are many polls with many methodologies and many flaws, which is why seasoned data folks will always show you an aggregate of a wide variety of polls.
If you’re looking at a line chart that shows poll ratings over time, there’s a key piece of information you might not know to look for: uncertainty. You can see a nice demonstration of how to draw uncertainty below—it’s that pale blue or red fill surrounding each of the center lines.
That simple shading enables you to see points at which the polling data overlapped. Back in March and April, there were substantial overlaps, which meant that although the averages showed former vice president Joe Biden ahead by five or six points, there was a chance that the exact opposite could be true—President Donald Trump might have been in the lead by a couple of points.
Uncertainty is something designers have paid more attention to in the 2020 election, after polls showed Hillary Clinton likely to win in 2016. The truth was that there was always a chance Trump would win that year, but when media outlets discussed the odds that Clinton would win it was generally as a flat percentage. Pay attention to the uncertainty, and you’ll be far better informed about all the possible outcomes on election day.