Don’t Use Stacked Bar Charts
When I was an undergrad, my Finance professor required the class to subscribe and read a newspaper. I chose the the Wall Street Journal and continue to read it daily. I’ve collected a handful of graphs from the daily news for a data visualization review.
Example 1: Parsing the Risks of Covid
The below graph attempts to summarize the results of a survey on attitudes toward the COVID-19 vaccine. The survey is a good idea. When the vaccine becomes available, who is going to take it and if not, then why? These answers could have a profound impact on policy decisions. Take a look at the graph and I’ll offer my analysis below.
The good: The bar charts follow a similar format throughout. Each graph starts with the survey question, shows the key and colors of the answers, and graphs the results as a percentage of the total responses. The color key order matches the order on the bars. Orienting it a horizontal bar chart helps with readability.
The bad: In general, it is difficult to compare stacked bar graphs in a meaningful way. Yes, we can compare the left-most and right-most values in each chart, but it is very difficult to compare the values in between. It looks like India and the U.K. both equally “somewhat agree” on taking the vaccine or do they? How do I determine the percentages without adding totals and subtracting parts? The graph doesn’t communicate the benefits of comparing the results by country and fails to direct the viewer to the important information. What is the intended takeaway from the first chart? Should I know the percentage of people that strongly agree to take the vaccine, strongly disagree, or live in the middle? What first appeared as an impressive way to communicate survey results is actually a confusing mess.
Suggestion: Don’t try to do it all in one graph. Isolate the numbers that are significant and present it in a way that’s easy to read. If the viewer needs to know the percentage of people that will take the vaccine, then show it in one graph.
Example 2: Bank Earnings Boosted by Investments
Three investment banks, JPMorgan, Citigroup, and Wells Fargo, reported better than expected earnings. That’s great news, especially after the economic destruction of 2020. Again, take a look at the chart and I’ll offer my opinion below.
The good: At first glance, the graph does a good job of showing quarterly net income over the last eight quarters and the viewer can quickly see that after a rough first half of 2020 the banks are back to 2019 levels. All this is shown in soothing shades of blue.
The bad: Unfortunately, that’s all the graph can communicate. It’s hard to see which bank contributed to the rebound and by how much because the values are stacked. To make things even more confusing, how is the viewer supposed to evaluate 2Q 2020? Wells Fargo reported a loss, which means that 2Q 2020 had the lowest earnings and should be the lowest bar. Instead, Q1 2020 and Q2 2020 have the same bar height which suggests they have equal total earnings and ultimately misleads the viewer.
Example 3: A Good Example
The bar chart below shows the weekly new unemployment claims from March to December.
There is an elegance in simplicity. The chart is clear in its message: unemployment spiked with lockdowns and continue to trend above pre-pandemic levels. In one glance the viewer can understand the trend and weekly numbers.
Example 4: Complex Can Be Clear
Are you curious what your federal income tax bracket is for 2021? The below charts shows just that.
These charts lay out the progressive tax rates along with the taxable income bracket. It uses an economy of space with no distractions. The large font makes the graphic legible, even on a small screen. For clarity, the tax rate is on the same line as the applicable income range. Instead of combing individuals with joint filers, the author makes two charts with the same format. The visualization makes an intimidating subject easy to understand.
