Visualizing Electoral Data 

 Andy Eggers and I are currently working on a project on UK elections. We have collected a new dataset that covers detailed information on races for the House of Commons between 1950 and 1970; seven general elections overall. We have spent some time thinking about new ways to visualize electoral data and Andy has blogged about this  here  and  here . Today, I'd like to present a new set of plots that we came up with to summarize the closeness of constituency races over time. This is important for our project because we exploit close district races as a source of identification.  

 Conventional wisdom holds that in Britain, about one-quarter of all seats are 'marginal', ie. decided within majorities of less than 10 percentage points. To visualize this fact Andy and I came up with the following plot. Constituencies are on the x axis and the elections are on the y axis. Colors indicate the closeness of the district race (ie. vote majority / vote sum) categorized into different bins as indicated in the colorkey on top. Color scales are from  Colorbrewer . We have ranked the constituencies from close to safe from left to right. Please take a look:  

   

 The same plot is available as a  pdf  here. The conventional wisdom seems to hold. About 30 percent of the races are close. Also some elections are closer than others. 

 A long format of the plot is available  here . It allows to identify individual districts, but requires some scrolling. We are considering developing an interactive version using javascript so that additional info pops up as one mouses over the plot. Notice that both plots exclude the 50 or so districts that changed names as a result of the 1951 redistricting wave.  

 Finally, Andy and I care about districts that swing between the two major parties. To visualize this we have produced similar plots where the color now indicates the vote share margins as seen by the Conservative party: ((Conservative vote - Labour vote)/vote sum). So negative values indicate a Labour victory and positive values a victory of the Conservative party. We only look at districts where Labour or the Conservative party took first and second place. Here it is: 

   

 The partisan swings from election to election are really clear. Finally, the   long format is here . The latter plot allows to easily identify the party strongholds during this time period.  Comments and suggestions are highly welcome. We wonder whether anybody has done such plots before or whether we can legitimately coin them as Eggmueller plots (lol).