In a two-stage cluster sample of rural and urban areas in Nepal, Rothenberg et al. (1985) found that the design effect for common contagious diseases was much higher
Histogram and normal quantile–quantile plot of body mass index from NHANES.
(a)

(b)

Than for rare contagious diseases. In the urban areas measles, with an estimated incidence of 123.9 cases per 1000 children per year, had a design effect of 7.8; diphtheria, with an estimated incidence of 2.1 cases per 1000 children per year, had a design effect of 1.9.
Explain why one would expect this disparity in the design effects. (Hint: Suppose a sample of 1000 children is taken, in 50 clusters of 20 children each. Also suppose that the disease is as aggregated as possible, so if the estimated incidence were 40 per 1000, all children in two clusters would have the disease, and no children in the remaining 38 clusters would have the disease. Now calculate deff for incidences varying from 1 per 1000 to 200 per 1000.)