Hey folks,

I really enjoyed teaching a one-day, introduction to ggplot2 workshop last week. It was a lot of fun - I enjoyed teaching the principles behind ggplot2. I’ve been noticing many learners (and teachers) focusing on making templates that they can recycle to make variations on a common plot type. This is how I often teach ggplot2 and the rest of the tidyverse - it’s also how I learned R. In the most recent workshop I was testing a hypothesis that teaching concepts would yield more long term learning gains than the template approach. I’d love to work out some of the kinks and teach it again. Let me know if you’d be interested in learning by this alternative approach.

If you’re a late-Gen Xer like me, the word “waterfall” will instantly queue up in your mind the song “Waterfalls” by TLC. Whenever I see a waterfall chart, I think of this song. Sorry. Not sorry :)

Anyway, last week I found this waterfall chart in a Washington Post article on the 0.3% drop in the US GDP during the first quarter of 2025.

Waterfall charts are helpful for depicting the cumulative effect of positive and negative components. For the GDP, personal consumption, private investment, government spending, and exports typically count in the positive direction and imports in the negative direction. For the first quarter, government spending was down a smidge, leading to a similar decrease in GDP. I was able to gather these data from the data linked through the Bureau of Economic Analysis.

R library(tidyverse)
tribble(~category, ~change,
    “personal_consumption”, 1.8,
    “private_investment”, 21.9,
    “gov_spending”, -1.4,
    “exports”, 1.8,
    “imports”, 41.3
)

I thought this would be a great plot to share with you all. I instantly started thinking about how I would create this in R. This plot has a few cool things going on.

First, there are vertical line segments with arrows. I would create these using geom_segment() with the arrow argument. Each category would have its own x and xend arguments, but these values would be the same within the same category. They would also have a starting and ending y position argument; the ending position having the arrow head. Thinking of the data I shared above in the tribble, I would likely calculate the cumulative change to create y value and lag that to generate the yend values.

Second, there are the bars. Normally when I see these rectangles I think, “bar plot!”. But bar plots start at 0 on the y-axis. These bars start at different locations for each category similar to the arrows. Instead, I’d either use geom_rect() to draw rectangles or I’d generate another line segment, but make them very thick to match the rectangle. Hmmm. I wonder which would be easiest to try out.

Third, each rectangle has a solid black line at either end to indicate the top and bottom edge. I’d likely do this with geom_segment().

Let’s pause here for a moment… Could I generate this waterfall chart using only geom_segment()? I would use the function in three different ways. I think this sounds like a cool challenge.

Next, there’s annotation for each piece of the waterfall where the category label is bolded and colored like the bar. The rest of the text is in a regular black font. The labels vary in how the text is justified. Because of the combination of font faces and colors, I’d likely use geom_richtext() from the {ggtext} package.

Finally, there are also the grid lines. At first appearances they look normal. But, the grid line that intercepts the y-axis at +1 is on top of the “Personal consumption” segment and behind the “Imports” segment. All the other grid lines are behind their segments… why?! I think this is silly. If I wanted to follow this faithfully, I’d likely use the background grid lines that are controlled with the theme(panel.grid) arguments. Then I’d either draw a fourth segment on top of the “Personal consumption” segment. Hey, that’s a fourth geom_segment()!

Let me know what you think of this type of plot. Have you seen waterfall charts in your work? I’d love to see more examples.

​