Your Graph is Bad and You Should Feel Bad

Why would you flip the y-axis. When people see a line going down, they think that means something decreased. When your y-axis has 0 at the top and 1000 at the bottom, it’s basically like you want people to misinterpret it (and perhaps maybe the graphmaker did. Shame. SHAME. SHAME.)

Also what’s up with the x-axis. Why have “1990s” and “2000s” instead of marks for individual years, or every 5 years? You can have tick marks for every year but only the number label every 2 or every 5.

Telling a story: Good graphs tell a story. This graph doesn’t tell a full story even at its best. If you ignore the y-axis, the implication was supposed to be that the ‘Stand Your Ground’ law led to a decrease in murders with firearms, but there’s no other events indicated to try to explain why the trend in the 1990s was a slow decrease (which looks like an increase) or after 2007 the number of firearm murders decreased again (which looks like an increase)

Y-axis at zero, please: I almost didn’t notice this but it’s the biggest problem with this graph and many others. See how the iOS bars are generally taking up a little more space than the Android bars? Makes you think iOS has a slightly bigger market share, huh? No, not at all. The Y-axis here starts ridiculously at 75%, actually Android has close to 85% market share. This is confusing and extremely misleading, whether it was done intentionally or not I have no idea

Change over time: When your x-axis is time, you probably want to be showing change over time, huh? But there’s very little change here. That’s not a problem per se, but as a person who works with data and makes visualizations, I generally produce simple versions of a bunch of graphs, then pick the best ones and make them clear and pretty. This would not make it into my final presentation.

Percents vs values: I feel like this is going to be a recurring topic. Percents are not the solution to every data problem. Percents are not always the best way to show what’s going on. Here, I’d be very interested in the total size of the worldwide smartphone OS market. Even if the percents for Android and iOS are about the same, are they still both adding customers? It’s a really important business question as well. If the market is increasing then to increase your share you need to capture new users. If the market is steady or shrinking, then you need to focus on converting users of the other platform. That’s an actionable insight that this graph could be providing but isn’t.

Gimme a Key: Obviously we’re using emojis here to represent people, in rows of 11 for some reason. There are white men, white women, black men, black women, then I think a latinx man (Dem, 2nd row, 4th), a latinx woman (Dem 3rd row, 3rd), a muslim woman, and... not sure who the last guy in the first row Dem is supposed to be. Also latinx? South Asian? I have no idea. And I equally have no idea who’s being represented by the monocle guy in the GOP area, although it’s probably supposed to make you think “Monocle = out-of-touch rich guy = Republican”

It looks more diverse if you mix them up: What looks more diverse, this

👱👱👱👱👱👱👱👱👱👱👱👱👱👱👱👩👩👩👩👩👩👩👩👩👩👩👩👩👩👩👩👩

or this

👱 👱 👱 👩👩👩👩 👱 👩 👱👱 👩 👱 👩👩👩👩👩 👱👱👱👱 👩👩👩 👱👱 👩👩 👱 👩 👱

They both have 17 women and 15 men (same as the Dems above). Changing the order of the faces so the white women and white men are spread out instead of all clumped together makes it look more diverse. 

Hair Color: Since the “man” emoji used here has approximately the same color skin and hair, it makes the GOP side look even more homogeneous. The brunette women break up that color scheme for the Dems. If the women were also blonde the Dems would look a lot less diverse.

 Confusing Measurement: I’m not sure what we’re measuring here in a lot of ways. What do you mean by “wages” and “profits”? Are those average wages of all workers, of non-management workers, of employees, does it include contractors, does it include tips, is that adjusted for cost of living, are those reported monthly or quarterly (hard to tell when there aren’t points marked on the graph), etc etc. I could ask pretty similar questions about profits (is that all corporations or just some top number, etc)

Confusing Measurement, part 2: This gets a separate bullet because what is going on. What does “Percent change from one year ago” mean? Are all of those measured relative to one year before the graph was made? Or are they all from one year before that point in time? Also  percents are notoriously hard to interpret correctly. Is 1% growth much worse than 2% growth? It’s half as much, but it’s only 1% less; change the way you phrase it and people interpret it differently.

Colors: I’m not sure whether this is necessarily bad, but the red/blue coloring here is likely intended to mirror the Republican/Democratic color assignment in politics to suggest that Republicans care more about corporate profits (in red) while Democrats care more about wages (in blue). 

Awfully convenient: Hey, see how that red line starts going up right at the same time as the GOP tax bill was passed? Wow, what an amazing correlation and/or causation. There’s some actual understanding needed of what’s in that tax bill and how the data here was gathered before we can say that there was a causation. Did the tax regulations change immediately when the bill was passed, or would they have taken longer to kick in? Was there anything else going on with the economy that was likely to contribute to rising corporate profits? A defender of the tax bill could also suggest at this point to track wages in a different way, as take-home pay may have increased even with nominal wages the same, then maybe both lines would jump after the tax bill.

I do want to stress again that I don’t like that tax bill, and I do believe that it helped corporations and the already rich more than the working classes. I just don’t think this image is likely to change anyone’s mind, at least not without a considerable amount more research.