Ranking presidents has been a pastime among American historians ever since Arthur Schlesinger's 1948 survey set the prototype. Although these rankings reliably draw news coverage, there are issues to keep in mind. Whose opinions should be surveyed? Liberals and conservatives, political scientists and economists, Americans and non-Americans might all have very different ideas about what makes a good and effective president. Also, how do we combine multiple rankings? Economist Kenneth Arrow showed in 1951 that there is no valid way to amalgamate multiple rankings into a single one. What to do? Visualization methods provide a way to tackle these problems. We begin by computing distances between individual rankers (the Kendall distance.) Then, with multidimensional scaling, we use these distances to plot the rankers and see what patterns develop. This gives a way to see the answers to such questions as - How much difference actually is there between different rankers? - Do political scientists and economists differ much in their rankings? - Do rankings change over time? These methods, of course, are used to visualize many different sorts of ranked data—from restaurant reviews to Hollywood movies. With this in mind, we use an R dataframe of movies with American presidents as characters to answer one last question: - If we left presidential rankings up to Hollywood, rather than political scientists, would anything change?
