Hi,

On p.54, in the "Comparisons, visual and numerical" subsection, the second sentence reads "For an example of a formal comparison, consider a change in the estimated support for the death penalty from [80% ± 1.4%] to [74% ± 1.3%]." However, I think the standard errors should be switched; assuming the sample size is constant, as the caption on Fig. 4.3 suggests, standard errors should be larger for proportions closer to 0.5. Alternatively, calculating reveals:
sqrt(0.8(0.2)/1000) = 0.0126 (close to 1.3%), and
sqrt(0.74(0.26)/1000) = 0.0139 (close to 1.4%).

Thanks,

Ravi