Shared publicly  - 
math fail
Matthew Yglesias originally shared:
Massive chart fail from the Heritage Foundation which tries to show that the Affordable Care Act killed jobs by referencing the second derivative of employment.
I don’t think this chart showing that private sector employment has grown after the passage of the Affordable Care Act really proves much one way or another, but it certainly proves that private secto...
Galen Pickard's profile photoMaitland Hemby's profile photoJames Clark's profile photoRoss Lippert's profile photo
You know, I never really thought it was that important for everyone to know basic calculus, until just now. Wow.
Wait, isn't that just the first derivative?
Well, they made the first derivative decrease. So they were looking at the second derivative.
The text in red should read +X jobs/mo^2. The plot is correct, if irrelevant, but the text is overtly incorrect in its interpretation of the plot.
To play devil's advocate for a moment, if the engine generating change in employment has a momentum component to it, then you could measure that momentum in this way. Failures I see here are (a) lack of evidence that there is any such momentum component, and (b) the assumption that the Affordable Care Act was the only factor affecting that momentum and not, for example, other forces in the system.

Oh, and as +Galen Pickard points out, (c) the units in the graph are labelled wrong.
Ahh, yes. Apparently, I need to take a class in chart-reading.
What text in read? I see +67,600 and +6,400 and both are in red and correctly unit-ed. Obviously (a) and (b) are issues, but I'm missing what everyone else is seeing, I guess with (c).
The chart shows jobs/month. The text corresponds to jobs/month^2.
Gosh what am I missing? I don't see the text then.
I seriously had to look at it 18 times before I got it, and my last comment was not well-phrased.

So, the first chart is straight-up, jobs vs. time. The second chart is the first derivative - delta(jobs) vs. time. All good so far. But the text in red (67,600 and 6,400), that's actually the slope of the the line in the second graph, so, the second derivative, jobs/month^2.

(Did I say that right that time? I think I said that right that time.)
No, I still don't get it. The vertical axis, the one centered at 0 is in units of jobs, with some baseline value subtracted out. The horizontal is in units of months. The two slopes are unit-ed correctly.

No wait, I am wrong. Ah yes, that isn't a baseline, zooming in on the first graph. It is the first derivative of the graph above it. The 0 crossing of the lower graph corresponds in time to the local min of the upper graph.

Yep. Fail. Would have been better off to fit a parabola to the left half of the first graph and overlay an extrapolation on the first graph. At least from the purely "make your point with graphs" issue, aside from (a) and (b).
No, I think the vertical axis is jobs/month. There was never a date in American history where there were -750,000 people working.

Oh but you mentioned a baseline value. I think there is a possible reading where they subtracted the January 2010 value out for some reason, but this reading seems unlikely since I think it is generally accepted that during 2009 employment levels decreased.
Yeah it is not a baseline. The zero crossing is at the local minimum of the top graph. That's a sure sign that it is proportional to the first difference/first derivative. That's basically what I missed.
Add a comment...