· How not to response to evaluations – The Guardian  discusses the response to an evaluation the UK government did of a mandatory work scheme, which required jobseekers to do mandatory unpaid work for 30 hours per week in order to continue getting a jobseeker’s allowance. The evaluation found this had zero impact on helping people get a job, didn’t have any long-term impact on benefit claims, and actually pushed some people into claiming a different type of benefit designed for people with health problems. Jonathan Portes has more discussion on his blog  including nice summary graphs. The Government’s response to this evaluation – argue that the evaluation covered the early months of the program when there were teething problems, so that it is already out of date – and then pump an extra 5 million pounds into expanding the program.
· Wired magazine’s Geek Dad has a nice summary  of different natural experiments used to assess the impacts of media – includes the Indonesia TV and social capital, India TV and women’s status, Brazil soap opera and fertility, and US studies.
· The UK Cabinet Office’s guide to Randomized Experiments  (via @timharford) – examples of some of the ways their behavioral insights team is using experiments to test policies and tweaks to implementation of policies.
· In Science this week – a discussion of the debate  over measuring disease mortality worldwide “for example, an international research group published the latest statistics on global child mortality in The Lancet. They estimated that 7.6 million children under age 5 died in 2010, well over half from infectious diseases. But fewer than 3% of those deaths were medically certified—assigned a cause by a health worker and recorded in an official database. For the other 97%, the scientists are forced to make sophisticated guesses.”
- links