The point this case study is trying to convey is that correlations based on averages are usually too high when applied to individuals. The reason is that averaging across a lot of people, in this case taking the percentages of people in each county who were immigrants and who favored Hiram Johnson, tends to smooth out a lot of individual-level variation in the data, which makes it seem like there's less noise than there really is and therefore gives you too high a correlation. The correlation the political scientists got (0.5) is almost certainly quite a bit too high as an indication of how strong the person-level relationship was between immigrant status and voting for or against Johnson.