Tuesday, October 28, 2014

Graphs that make me nervous -- UPDATED

[Here's a good link to the full report.]

A couple of quick caveats: I am never entirely comfortable focusing too much on "additional years of learning" -- it's a problematic metric -- and the link to the original report seems to be dead.

The good folks at Vox are very excited by the results of this recent experiment in merit pay. Perhaps I'm missing something obvious, but that jump between 2010 and 2009 seems really big.

Does this look odd to anyone else?


  1. FWIW, I was struck in the same way by that graph. The jump looks too big to be true. And, I, too, was unable to link to the original report.

  2. All hipster charter school data should be interpreted the same way: if the charter school serves public, minority populations in any way, what real results did it achieve? Ask how many black or hispanic students in their system passed like the AP Calculus or Chemistry exams. When the system is explicitly high school level there's no excuse for not asking for results to compare to other school systems to demonstrate progress on reducing educational inequality. Instead, things like this are usually non-stories because most touted reformist and corporate/charter school projects end up worse or no better than general public schools and only focus on exploiting provincial measures. For underappreciated though maybe surprising to some background info check here (Appendix C, or previous years results and comparable exams, tho looking at AP-like exams is generally much more valuable than SAT/ACTs due to student selection concerns with the latter) http://apreport.collegeboard.org/

    Most intra-state level high school tests or internal "value" metrics are pretty pointless so those results are just noise. If high stakes pressures do not result in explicit cheating (Atlanta) they are still often gamed.

    1. Apologies, same anon here, without tracking down the original report which had been broken I did not see that this charter school wasn't high school level. That makes it easy and likely for backers to be king weasels, they won't ever have to account for how the students do thru high school anyway.

      10-to-1 odds no one will ever see a follow-up on these students and the control group for high school attainment markers like SATs and AP exams and college enrollment. State test results from overtested younger kids certainly don't mean much long term - nothing against the kids, just the meaningfulness of the data in the first place.