« Another take on The Problem of Susan | Main | Junot Diaz reading/interview at Google »

Mind the y axis


According to this chart from HuffPo, the US Presidential race is very close. Look at the two trend lines, close together, both just a little under 50%:

But wait—according to this chart from HuffPo, Obama has an enormous lead! Look at this huge gap! Obama's trend line is nearly off the chart!

As you've no doubt figured out by reading the title of this post, they are in fact the same chart; it's just that one has a y axis that goes from 0% to 100%, while the other has a y axis that goes from 42% to 49%.

The latter is more dramatic, and shows the differences more clearly, and so people tend to show graphs more like the latter than like the former. HuffPo's default view, for example, has a y axis running from 35% to 55%:

And that's not a false view of the data; it's arguably not even misleading, given that a lead of a few percentage points is significant in this race. But I do think it's always worth checking the y axis to see just what the graph is really showing.


There's a good summary of some of Edward Tufte's taxonomy of commonly-used misleading graph techniques at http://thedoublethink.com/2009/08/tufte%E2%80%99s-principles-for-visualizing-quantitative-information/

Good stuff Jed. People that are most interested in the outcome of the race I think are focused on "likely voter" polls in swing states, or even more specifically, swing counties. And I think people who study that data have opinions that are reflected on betting sites like InTrade. And InTrade has Obama right now at a 75% chance of victory. Sure there are lots of amateurs (who don't study "likely voter" polls) who bet on sites like Intrade, and of course there are lots of factors that could change the race, but I think the 75% chance captures it pretty well at this point.

Belated thanks, Jim and Jay!

I should clarify that the point of my post here was intended to be “when you're looking at a graph, always check the y axis to see what it's showing.” I didn't really mean to suggest that any of these graphs are wrong or misleading, just that they give different impressions if you're not looking at the y axis.

But as with most of what I've posted lately, I think I didn't do a good job of saying what I was trying to say, 'cause it sounded like most of the people who commented on it (including Facebook commenters) came away with the impression that I was either complaining about the graphs being misleading or trying to figure out how big Obama's lead was.

So, apologies for my misleading phrasing/presentation—especially ironic in a post that touches on clarity of presentation.

Post a comment