Scary graph

[unixronin]

Commentary:  Someone elsewhere commented that this needs to be a log scale, and is meaningless if it's not in adjusted dollars.

I disagree.  Adjusted dollars or not, anyone can see that this is blip, blip, blip, blip, blip, HOLY FUCK.  Sure, the dollar has tanked. But it hasn't tanked THAT much.  The dollar's lost maybe half its 1985 value, not 98% of it, which is the sort of depreciation that would be required to bring that vertical spike into scale with the other blips.

As for a log scale?  On a graph like this, the only thing a log scale would serve is to hide the data and make it look much, much less significant.  Most people do not think in terms of log scales (hell, most people don't understand what a log scale is).  If you're releasing log-scaled graphs for the public at large, you're doing it to mislead them and make the spikes look as much smaller as you can get away with.

Comments

[unixronin.livejournal.com]

OK, I should have said "if you're releasing log-scaled graphs of this type of data". There are times and audiences for which log scales are appropriate. This is not such a case.

The reason this graph scares me is that this is completely unprecedented territory. Extrapolations from previous data about how to behave, economically, aren't going to work here.

Yeah, exactly. I've seen graphs like this before, in systems where there is a significant, but finite, amount of internal ability to stabilize itself. The outwardly visible datum looks stable, with only minor corrections and fluctuations, because you can't see the system internally managing itself to keep up with the increasing deterioration. Then the system exhausts its stabilization (or error correction, or damage control) capability, and suddenly everything goes to hell in a handbasket before anyone has time to react, leaving everyone sitting around dazed and wondering WTF just happened.