I'm familiar with this paper. Unfortunately it's deeply flawed.
The RAND health experiment studied 30 measures of health. Two of them were statistically significantly improved - vision (due to free eyeglasses, p=0.001) and blood pressure (p=0.03).
Simple math shows that 0.97 ^ 30 = 0.4. This means that the RAND experiment had a 60% chance of showing an effect in at least one particular submeasure even if there was no effect in any of them.
(The result on eyeglasses is strong enough to avoid this effect, however. So we can conclude that free glasses help people see better.)
The RAND health experiment studied 30 measures of health. Two of them were statistically significantly improved - vision (due to free eyeglasses, p=0.001) and blood pressure (p=0.03).
Simple math shows that 0.97 ^ 30 = 0.4. This means that the RAND experiment had a 60% chance of showing an effect in at least one particular submeasure even if there was no effect in any of them.
(The result on eyeglasses is strong enough to avoid this effect, however. So we can conclude that free glasses help people see better.)