I’ve written before about how dual axis charts can mislead and confuse (here and here), but I’m adding another post to the series because recently the US Department of Agriculture Economic Research Service (ERS) published a chart that clearly shows how these graphs can be confusing.

This ERS dual axis chart shows gross cash receipts, cash expenses, and net agricultural income in Puerto Rico from 2012-2020. Cash receipts and expenses are shown as vertical bars and tagged to the left axis and income is shown as an area chart and tagged to the right axis.

Dual axis chart with the title "Puerto Rico's gross cash receipts, expenses, net agricultural income, 2012-20." An orange area chart shows net agricultural income and a paired bar chart shows receipts and expenses.

Look quickly. Which is higher in 2012: receipts, expenses, or income?

At a glance, you might think that receipts were higher in 2012, with income higher in all other years. But when this graph is recreated using a single axis (I eyeballed the numbers, so don’t hold me to the precise values), it tells a very different story.

A redesigned ERS graph where the area chart is now a line chart and all data are on the same left vertical axis.

Because the left vertical axis starts at $400 million, it distorts the differences between the bars and the bars’ values relative to the area chart. But that’s only part of the problem. Even starting the single axis version at $400 million doesn’t show the same relationship as the dual axis version.

Same line-bar chart as above but the vertical axis now starts at $400 million.

That’s because there’s an additional problem with this graph, hidden in the chart note: “Gross cash receipts and cash expenses for 2012 are adjusted for inflation with base year of 2018. Net agricultural income has not been adjusted for inflation.” In other words, the receipts and expenses values are adjusted for inflation, but the income values are not. Even though both of the original axes are measured in millions of dollars, the dollar values aren’t actually equivalent.

The full report goes into more detail (I have bolded the phrases that are most important for this discussion).

Cash expenses for Puerto Rican farms decreased between 2012 and 2018 by 16 percent, from $594 million to $500 million. After adjusting gross cash income values downward to be comparable to cash expense values in terms of the number of reporting farms, calculated net farm income reflects the rise in revenues and decline in cash expenses, increasing from $15.5 million in 2012 to $21.0 million in 2018. That is, net farm income rose 36 percent because of both the decrease in the number of farms reporting and the increase in Government payments.

To depict how farm revenues—and therefore gross cash farm income—may have fluctuated in the years between Census of Agriculture reports, figure 5 combines data sources to show net agricultural income for 2012 through 2018. The $101 million decline in net income indicated by the solid area spanning from 2012 to 2020 in figure 5 reflects the net agricultural income reported by the Puerto Rico Planning Board. It shows a less dramatic decrease than the $133 million decline in farm cash receipts (blue bars), based on USDA, NASS survey data collected for the Census of Agriculture.

Notice the focus on change in the boldface text, something that can only be done once the data have been adjusted for inflation. Even if the analysis intended to show a decomposition in net income, because the report includes—and compares—changes in income (both percent and level), all dollar values should have been adjusted for inflation. By not doing so, the chart essentially compares apples to oranges.

In this case, because the time series is not so long, adjusting for inflation isn’t hugely important, but you can see some differences in the nominal (orange) and real (blue) series in the next graph.

Line chart with two lines--an orange line for nominal agricultural income and a blue line for inflation-adjusted agricultural income. The orange line is slightly below the blue line until 2018 until it's slightly above the blue line.

Once again, we can see how dual axis charts are problematic. They can confuse, mislead, and distort the data. In this particular case, a dual axis chart wasn’t even necessary, and it just made the data harder to read. But the lack of comparable adjustments to the data, which the dual axis chart initially masked, makes the presentation even more problematic. 

Did you like this post? You could have seen it a few weeks ago in my free bi-monthly newsletter, now on Substack!