When Time Gets Flipped: Rethinking Bar Chart Order

A recent Washington Post poll about voters feelings toward President Trump’s policies found a public mostly critical of the president’s policies. As part of their write-up, they compared these poll results with responses to a similar poll conducted in February, shortly after Trump took office.

While the differences are illuminating of the shifting sentiments among the president’s base, the chart the Post used to display these is responses is a bit less clear.

I’m not sure there is an optimal way to order the bars in a bar chart, but when I read this graph, I wonder if there’s a better way. My instinct is to interpret time in this graph as going from the top bar to the bottom bar. Thus, if I look at this chart at a glance, I initially read the poll numbers as declining between February and September because I assume the top bar is the first result chronologically. In truth, however, the shares actually increased from 39 percent in February to 55 percent in September.

Visually, this is how I read the chart:

But why is my inclination to assume the chart works this way? There are two possibilities: First, when comparing values vertically, maybe people are just naturally inclined to interpret them from oldest to newest? I don’t think that’s necessarily the case. This bar chart of gross domestic product (GDP) per capita, for example, works in either way—with the most recent year at the bottom or top.

But in this example, there is only one comparison to make—per capita GDP over time—making our navigation across the chart relatively easy compared with the Washington Post chart that requires two sets of comparisons: within and across each category.

Which brings to me to the second possibility and what I think is what’s really going on. Because there are these two comparisons, we expect time to be presented left to right, oldest to newest, the way time is typically presented in a vertical bar chart (in English cultures at least). In the case of the Washington Post data, a vertical bar chart would look like this:

Reading left to right, we see the values for immigration increased, DOGE decreased, and so on. Rotating this chart 90 degrees would place the bars in the same relative position, with February bars above the September bars. Like this:

Again, I find this easier and more intuitive to navigate—top-to-bottom, older to newer.

I’m not claiming this orientation as some kind of data visualization “rule,” but perhaps—and I say this lightly—it is a better practice (not best, just better).

Let me know what you think. Did you have the same issues with this chart? Do you have a practice you regularly follow?

---

If you liked this post, consider subscribing to my Substack newsletter. It’s completely free and you get more content like this, plus links to my podcast and books, articles, and data visualizations I’m reading.