Everything you think you know about test prep is wrong. Specifically, test prep does almost nothing for test scores, there is little variation in test prep by income, and black people use it more than whites.
The AAA Tranche of Subprime Science
Surprising facts about income inequality
Barratry (about a historical system of government)
What do you do after a major policy change? Alter your measurement technique in order to make "before" and "after" results incomparable. By a strange coincidence, "we are expecting much lower [better] numbers" - policy wonks will of course recognize this is "just because of the questions and how they are asked", but somehow I doubt innumerate reporters and the general public will.
How Biju Dominic reduced pedestrian deaths on Bombay railway tracks. Very interestingly, he didn't try to dissuade people from crossing - he just tried to make safe crossing easier.
On the topic of HFT (which I know I'm beating to death), Felix Salmon reblogs Joe Stiglitz, and the original Joe Stiglitz. Unlike Michael Lewis, this is a critique worth reading, even if I disagree with some of the underlying premises (part of Stiglitz critique of HFT is that markets are not valuable, therefore HFT improving markets is not).
A nice Comparision of different concurrency models: Actors, CSP, Disruptor and Threads. Nothing detailed, but nice overview.
Nassim Taleb On the Biases in the Estimation of Inequality Using Bracketed Quantile Contributions. A very interesting paper which shows that for a fixed level of power-law inequality, improvements in the ability to measure it will look like increases in inequality. This is a very important bit of math. I don't know why I didn't see it before.