A Bird’s Eye View of the Political Landscape from the Pollster API
In the next few months there is going to be an ever increasing stream of statistics, polls, infographics and visualizations headed your way, all with the same subject: The US Presidential Election. In a politically charged environment it’s hard to know which media outlets to trust. What’s more, the public will continually look to data as an unbiased source of information, whether or not that is always the case.
In a move that bolsters their trustworthiness in my eyes, The Huffington Post has come out with the Pollster API which aggregates the many election-related polls put out by a plethora of media organizations and packages them in a tidy API. What’s more they’ve taken the initiative to visualize much of this data to provide a true bird’s eye view of political polling.
These visualizations live amongst some of the Huffington Post’s other data-driven election pieces including a detailed election map and interactive primary results. The Pollster visualizations themselves are actually plotting the results of all the polls HuffPo collects through their API, and there are quite a few. Each row scrolls to the right to reveal aggregate polls for individual states on many issues. Let’s take a look at the first chart as a case study, embedded here.
Including trend lines is very helpful since there are too many individual polls to form an interpretation of a trend. The vertical line and info box showing you the values at any point along the line is a really useful interaction. The trend lines seem to mirror each other which makes perfect sense when you realize a single poll is represented by multiple dots on this chart. This is a bit of a tricky concept to grasp since the two data points aren’t visually connected in any way. The only way to confirm this is by finding the poll points with matching labels.
A great detail they’ve added, and I think one that’s pretty crucial, is the ability to actually click on any individual point to reveal the connected polls over time. These are all the points on the graph that belong to the same poll. Unfortunately this functionality isn’t easy to discover since any motion of the mouse produces a hover state with fine grained data. It’s difficult to expect a user to know there’s another layer of data available with a click.
The graphs presented along with the Pollster API aren’t necessarily designed to be used as communication devices or shown in the context of a particular piece of journalism but they do have their own audience and will inform how developers graph data from the Pollster API. With that said I think it’s fair to question whether or not this is the best way to represent the data.
The initial visual impression of the lines mirroring each other is somewhat misleading. The “phenomena” is explained by the discovery that there are two opposing data points per poll but I feel that there might be a better way to visualize the results of a single poll. What if the two data points from an individual poll were more visually connected?
In my quick mock-up each poll is represented as a column and ordered chronologically. The two ends, red and blue, are aligned to the top and bottom edge and the space between represents “undecided” and “other” responses. The vertical scale is 0 to 100, giving you a full picture of the results. Here the space between the two bars is highlighted, illustrating the a growth in undecided people or quickly demonstrating a surge from a particular candidate. The hover state for an individual poll makes the values for each answer clear and clicking on a poll shows other instances of that poll along the timeline.
Is the original Pollster visualization a better representation of the data? Do you have a different way to display this data from the Pollster API? Post a comment below or follow @thewhyaxis on twitter and tweet your thoughts and ideas.