Most of Trump’s charts skew the data. And not always in his favor.

Donald Trump’s campaign is obsessed with favorable poll results, as evidenced by a graphic deluge on official social media channels. It’s a marked difference from their Democratic rival who has led polling averages throughout most of the campaign. And while neither candidate has a perfect record when it comes to spinning the data, only the Trump campaign has used a combination of cherry-picked polls and misleading data visualization to paint a cheerier picture of its prospects on Election Day.

Donald Trump’s campaign has posted more than 40 bar-chart graphics showing favorable poll results. Here’s a chart the campaign tweeted on Oct. 5. Notice anything strange about it?

How tall are the bars in the chart? Instead of ending at a clearly defined baseline, the bars fade to black. But if we add a baseline at the bottom of the graphic, the implied scale doesn’t match the numbers.

The chart visually inflates Trump’s two-point lead to over 11 points. If we adjust our scale to match the data, where would the baseline end up? In other words, how tall should the bars actually be?

Data visualization can help people understand information by representing data with graphical properties, such as size, shape or color. Bar charts represent data using length. This means that if the bars don’t begin at zero, comparisons between bar lengths will distort the data — and a reader’s comprehension of the numbers. In the extreme example below, bars A and B are nearly equal in length but we can exaggerate A’s lead over B by cutting off the scale. The distortion can also work in reverse.

Trump’s charts often distort comparisons by hiding or misplacing the baseline, mostly exaggerating his lead, but sometimes diminishing it. Here’s what happens when we put all of Trump’s charts on the same scale:

Some of Trump’s charts fade to black or otherwise obscure the baseline, while others imply a baseline that doesn’t match the data the charts purport to display.

The implied scales on Trump’s charts vary wildly even within the same graphic. The distortion is not always beneficial to Trump — some charts make Trump’s lead over Clinton appear smaller rather than larger.

One chart, from a July 18 tweet showing a Trump lead among members of the military, accurately displayed the data. The graphic doesn’t name a source, but the data appears to come from a Military Times poll of “active-duty troops, reservists and National Guard personnel,” including third-party and write-in candidates.

Update: On October 31, the Trump campaign posted a fairly accurate chart to their Instagram account. The campaign posted another chart the same day, however, that exaggerated Trump’s lead.

What about Hillary Clinton?

The Clinton campaign posts charts sparingly compared with the Trump. They use data visualization to illustrate campaign issues and policy, rather than poll numbers. For example, here’s a bar chart from the campaign’s Instagram account, posted in April.

Clinton’s bar chart, which compares wage gaps between men and women by race, also has scale issues. The chart uses light gray bars to represent 100 percent of the median wage of  white men. The bar comparing the median wage of Asian-American women to that of white men is labeled 84 percent but reaches nearly 90 percent. This scale error makes the wage gap appear smaller than the data suggests it really is.

The bottom line

Bar charts should always have a clear baseline, representing zero on the scale. Bar charts must begin at zero to avoid distorting length comparisons. Whether you are viewing or creating a data visualization, check that the chosen scale does not distort the data. For bar charts, a zero baseline is the difference between data visualization and data spin.

All of Trump’s charts