The issue of gay marriage is a contentious one, and we recently had yet another glimpse at just how controversial it could be when North Carolina residents got the chance to vote on it on May 8, 2012. North Carolina state constitution Amendment One, which defines marriage as between one man and one woman and bars the state from recognizing domestic partnerships or civil unions (same sex or otherwise), was passed by popular vote: 61.05% for; 38.95% against. Same sex marriage was already illegal in North Carolina; the amendment added the ban to the state constitution. After the amendment was passed, several maps began circulating social media services, pointing out some pretty telling correlations. Unfortunately, these maps were not sourced well, the data analysis was not necessarily trustworthy, and there were a few features that would improve them. We felt that many of these maps only told part of the story behind Amendment One. In order to investigate the vote further, we pulled in some census and state university data to produce an interactive visualization for comparing North Carolina counties data sets. (Click on the image below to go to the interactive on Visual.ly.)
Before we get into the details of the project, let’s look at some of the maps that inspired this project, and how they could have been better. The map that kicked off the whole process was WRAL‘s interactive election map. It was updated as soon as preliminary results came in, feeding the media’s desire to provide timely (if altogether inaccurate) information.
So: the map seems straightforward enough, what’s wrong with it? Binning a whole county into “for” and “against” is potentially a serious problem. If every green county was 49% against, 51% for, the map would look like it was a landslide despite being fairly close. (This didn’t actually happen, but the map is still hiding valuable information.) In addition to the binning issue, red and green was a poor color choice. First, colorblind people will have some serious issues trying to read the map. Second, red is typically seen as bad in our culture, while green is seen as good. This attaches a bias to the meaning of the map. For many cases this would be okay, but in this case, since voting for amendment one (green) actually means a vote against gay marriage, the red and green seems to be anti-gay marriage, something an impartial news organization should avoid. The early preliminary results from WRAL’s interactive were captured by someone in a screenshot and major universities were pointed out on the map. Unfortunately, since these were preliminary results, New Hanover county and Guilford county were incorrectly marked as being against, while Dare county was marked as being for the amendment. When the final tally was in, the counties were actually tipped the other directions. (This early reporting would be less of an issue if the map showed true percentages instead of binning into for and against.) In addition to the inaccuracies, the map has been very selective of the universities that were included. North Carolina has 17 state universities, and while the 8 that are pictured are the largest, the other 9 are still very relevant. Another map that appeared shortly after the election shows the number of people who completed a bachelor’s degree. There are a few problems with the bachelor’s degree map. First, the data is not sourced. Second, if we assume the map uses census data, it is excluding people with higher than a bachelor’s degree. The way the survey is taken, people respond with the highest attained education, so individuals with higher than a bachelor’s degree would be excluded. Third, the map bins the percentages instead of using a continuous color scale. This results in some minor ambiguity in reading the percentages.
After seeing the existing maps and the problems they had, we decided we could do better. To start with, we settled on a blue-yellow color scheme for the amendment maps. These colors do not have the semantic positive-negative associated with red-green, so there is no bias built into the visualization. We like the simplicity of the binned view showing the swing of a county, but since it hides details, we also included another view showing percentages. This view has a diverging color scale so that blue is still counties that had more votes against while yellow had more votes for. Diverging color scales can be difficult to compare to non-diverging color scales, so we also included a straight white-yellow linear scale of the same data. This scale is good for finding correlations between maps. The university map had some potentially good points, so we included a revamped version of it showing the count of state universities in each county. We also included data about degrees, but we added graduate and bachelor degrees together to show the data more accurately. Since choropleth maps always have problems due to population density, we also included a map showing population density. We felt that there were other categories of data that might be relevant to the issue. We also decided to include: marriage, separation and divorce, income (there are monetary benefits for married couples), female to male ratio, or veterans (due to the recent abolition of the Don’t Ask Don’t Tell policy) all have relevance to the marriage issue. We also decided on a side-by-side layout to allow easy comparison between any two maps. Hover interaction provides details on demand, showing the specific numbers for every dimension for the current county, and the county outline helps to compare the current county in each map. We welcome any criticisms and suggestions, and we hope you enjoy exploring the maps and coming to your own conclusions! Lane Harrison and Drew Skau are PhD Computer Science Visualization students at UNCC.