The "Moving to Opportunity" experiment, in several U.S. cities, stands out for its remarkably ambitious random assignment research design. By lottery, participants were assigned to receive special housing vouchers that could only be used in low-poverty neighborhoods. The idea was to find out if "neighborhood effects" really contributed to keeping poor Americans from moving ahead. While many of the expected employment benefits appear not to have panned out, the experimental group lost weight by comparison to two control groups (one control group received traditional vouchers without a location restriction and the other was a true control group). There were also benefits for mental health. The authors (Jeffrey R. Kling, Jeffrey B. Liebman, Lawrence F. Katz, and Lisa Sanbonmatsu) recognize that the obesity reduction could be a fluke -- if you test enough variables, you are likely to get some spurious significant results by chance. Still, the result is fascinating and deserves further exploration. There are many reasons why moving from a highly impoverished neighborhood to a low-poverty neighborhood could affect obesity. One thinks of better food retailers and built environment, for example, and possible peer effects from neighbors.
The on-line paper (.pdf) is also nicely written, with a good exposition of how random assignment to a treatment is related to statistical approaches using instrumental variables. I will use the paper in teaching.