Time to breathe new life into my little YouTube channel, which I’m calling eagereyesTV. I’m doing so with the start of a new series I’m calling Chart Appreciation. Each episode will be on one particular visualization, news piece, or interactive. As the first one, I picked Hannah Fairfield’s Driving Safety, in Fits and Starts from 2012.
This is one of my favorite charts ever and it uses the connected scatterplot, a technique I have a special fondness for. So it seemed like a good starting point for this new series. I put a lot more work into the animations than I have ever before, and have also generally upped my game in terms of editing, as well as video and sound quality. The video is actually in 4K, so if you have the required hardware you can enjoy it at full resolution!
As usual, you can watch it on YouTube or directly here. Though I would really appreciate it if you could head over to YouTube to subscribe to the channel or maybe even give the video a thumbs-up!
Driving Safety, in Fits and Starts can be found on the New York Times website. The paper about the connected scatterplot I mention towards the end is work with Steven Franconeri and Steve Haroz from 2016.
I obviously intend to make more videos in this series, and in general (since the channel has been dormant for a while). I’d appreciate any comments and thoughts you have, positive or… constructive. I’ve put an absurd amount of work into this video, so I hope somebody finds it interesting.
Firstly, I love this video. Interesting content and great sequencing of ideas, reveals, overall production quality. Congratulations from someone who also finds himself making a lot of videos these days…
Secondly, I’ve never been crazy about this chart. Its one strength, IMHO, is that it does a great job of grabbing the reader’s attention (which, obviously has a certain value). In every other respect, I suspect that, if tested, it would perform worse than two line charts stacked one on top of the other (so that the years line up vertically). The two-line version would be much quicker/easier to grasp (it would require less explanation than what you had to provide in the video), and all of the insights that you mentioned would be more visually obvious. Not as visually interesting as a connected scatterplot, but probably more cognitively fluent and insight-generating. Another design worth testing would be a “0%-origin” line chart, which pegs the first year (1950) at 0% for both series and shows both lines in the same chart as % deviations from 1950.
Thanks! I think it very much depends on what your criteria are. We did test the connected scatterplot against dual-axis line charts for that paper I mention (linked above as well). I’m not so sure that another chart would be more “insight-generating” – a chart that makes you stop and ponder it for a while is going to give you more than one that looks like every other chart. Seeing how changes align in time is also pretty difficult with line charts, no matter if dual-axis or stacked.
Thanks Robert. I look forward to reading the paper.
After looking at this chart a little more, though, I started to get even more uncomfortable with it.
One concern is that people can only ponder the data if they can figure out how to interpret the chart, and I wonder how many people put in the time and effort to figure out how to interpret this chart (not just how to *read* it, but what it *means*). The “what it means” part took some effort for me, effort that Joe NYT reader may or may not be willing to expend.
A larger concern is that I suspect that many people –consciously or unconsciously– will interpret this chart basically as a line chart with time represented linearly from left to right. But it’s not, and this could lead to some significant misperceptions. For example, the change from 1973 to 1974 looks similar to the change from 2004 to 2011 but, in reality, the 1973-1974 change was far “steeper”. Another example: The period from 1992 to 2005 looks like a brief pause in a steep decline, but it’s not brief at all, representing almost a quarter of the entire time span shown in the chart. I’d worry that even people who consciously realize that time isn’t being represented as a linear, left-to-right scale could still be prone to these misperceptions.
Ultimately, the true pattern of change of these metrics over time (especially “Vehicle miles driven per capita”) is extremely difficult to perceive accurately, even when one correctly understands how to interpret the chart, and I see that as a major problem because the story is about the patterns of change of those metrics over time.
There are other potential perceptual problems, as well. For example, the change from 1973 to 1974 looks huge because that line segment is very long, but “Vehicle miles dirven per capita” only experienced a small change during that time. Maybe it was just me, but I had to constantly and consciously remind myself to correct for these misperceptions as I was interpreting the chart.