[rcthompson]

I think there are one too many dimensions to this data. I know they collect statistics on both approval and disapproval, but for the most part approval + disapproval = constant, so I think it would be more informative to simply plot something like (approval - disapproval) on the y-axis and time on the x-axis. Then you could see the trends and compare different presidents without having to watch the animation once for each president and then trying to remember where that president went while you're watching the next one so you can compare them.

[xianshou]

Actually, for several of the presidents, approval + disapproval != constant. For example, Obama moves straight right (up on disapproval) for his first 200 days before descending in approval. There seems to be a substantial difference between the effects of actions that anger some (presumably of the opposing party) and those that disappoint all.

[mbostock]

There's a good analysis of how to design a visualization that shows both approval and disapproval here:

http://eagereyes.org/blog/2012/embracing-uncertainty-two-lin...

[CWuestefeld]

There's a problem, but not with the approval/disapproval axes. It's with the time axis.

Note that we're looking at days 0 - 1,000, that is, less than a 3-year range. This completely overlooks the second term of 2-term presidents. So, for example, we're seeing GWB's post-9/11 boom but not his later ignominy, and we're not seeing Reagan after he got the economy sorted.

[colonel_panic]

But let's face it, the reason we're having our attention drawn to this right now is that we've got a president running for reelection, so it gives us exactly what we want to know.

[redthrowaway]

If you did that, though, you'd have to plot each president as a line. That's not necessarily a bad thing, but it's a different visualization entirely.

This reminds me in many ways of Hans Rosling's TED talk on visualizing global health outcomes and other data:

http://www.ted.com/talks/hans_rosling_shows_the_best_stats_y...