For all we've heard about Dubya's administration injecting politics into government research, its worth noting that some of their policies are backed up by independant research. During the last presidential campaign, for example, Harvard Professor Paul Peterson released a study showing clearly that vouchers improve student test performance. Though, strangely, only the performance of black students. Or, as it quickly emerged, only the black students at one particular grade level. Or something like that:

A Princeton economist, Alan B. Krueger, took the offer [of raw data], and after two years recently concluded that Professor Peterson had it all wrong -- that not even the black students using vouchers had made any test gains. ...

What makes this a cautionary tale for political leaders seeking to draft public policy from supposedly scientific research is the mundane nature of the apparent miscalculations. Professor Krueger concluded that the original study had failed to count 292 black students whose test scores should have been included. And once they are added -- making the sample larger and statistically more reliable -- vouchers appear to have made no difference for any group.

Via the indispensible dsquared, which goes on to discuss why "data dredging" is an unsound research method even if the data you're using is something like accurate (in a post to which the links are now bloggered).