Read Bruce D. McCullough and Ross McKitrick's new paper, "Check the Numbers: The Case for Due Diligence in Policy Formation". (I thank Professor McCullough for the pointer.)
From the beginning, where the authors classify replication in the social sciences as an under-provided public good, to the last sentence, "Disclosure of data and code for the purpose of permitting independent replication in no way intrudes on or imperils academic freedom; instead, it should be seen as essential to good scientific practice, as well as a contribution to better public decisionmaking," it is superb.
In between, the authors discuss some recent horrible examples, including the following:
--The Harvard Six Cities study
--The Boston Fed study
--The "hockey stick" graph
--The U.S. obesity epidemic (JAMA paper)
--The Bellesiles affair
--The JPE file sharing paper
Calls for more replication in social science research are now astonishingly old--I was in graduate school or shortly out of it when I read a paper by the then editor of Economic Inquiry calling forcefully for more, and his wasn't the first. There's been some improvement since then, but shamefully little.
This paper deserves a wide audience.