The 2022 midterms marked the first time candidates ran in redrawn congressional districts after the 2020 census. US law requires that each of the 435 districts in the House of Representatives have the same number of people, which means districts are evaluated and redrawn every 10 years.
CNN’s interactive allowed users to toggle between two national maps to see how redistricting has changed the country after the census, based on how the residents voted for president in 2020. It also allowed the reader to view each state to take a closer look at how each one has/hasn’t changed since 2010.
Overall pageviews were 1.3 million across the entire project (combining the lead national page link and all 50 individual state links) for the year. That makes this project the No. 4 interactive on CNN.com for the entire year, and the top interactive for 2022 that was launched within that same year (our other top interactives are ongoing trackers, such as the CNN Storm Tracker).
CNN’s project provided critical analysis and information to our audiences about some of the most hotly contested congressional races in the country, highlighted by the top state pages receiving pageviews, which in order are: California, Ohio, Georgia, Virginia and Texas. Redistricting can have a considerable impact on the voting power of political parties and racial/ethnic groups, so we wanted to analyze these changes, focusing on both the political and demographic effects that will be in place for the next decade.
Every 10 years, US states redraw their congressional district boundaries to account for population changes recorded in the decennial census. This process can alter the political landscape. We used ArcGIS to compile the new boundary shapefiles, block-level voting-age population counts from the Census, and 2020 election results from Harvard Dataverse, which were provided at the voting precinct level. To allocate 2020 presidential election results to the new congressional districts, it was necessary to do a “spatial join” of the voting precinct data to the new districts. In cases where precincts were split, a second spatial join was performed to attach block-level voting-age population to the split geographies, which informed how we allocated votes among the split areas.
A similar technique was used to calculate the demographics for the new districts. We used block-level voting-age population, along with block-equivalency files for the new districts, to tabulate the demographic profile for each new district.
After the spatial analyses were complete, it was possible to tabulate 2020 vote totals and race/ethnicity profiles for the new congressional districts, which we wanted to display on maps for a visual comparison to the old districts. For the final presentation, we worked through multiple design iterations on Figma to find the most user-friendly and simple way to make the complicated process of redistricting accessible and helpful for readers, with an eye on making it as mobile-friendly as possible. The design focused on using shapes, text and color to complement the maps and avoided any imagery to keep it as lightweight, quick and accessible as possible.
We created an automated pipeline to tie the election and demographic data to the old and new shapefiles and output a compressed pre-projected topojson file for each state. This, along with pre-rendered html pages (via NextJS), enabled faster loading and display.
Context about the project:
Every 10 years, US states redraw their congressional district boundaries to account for population changes recorded in the decennial census. This process can alter the political landscape for the next 10 years, and we wanted to analyze how the political leanings and demographic attributes of these new districts might change voting power in all 50 states. But historical election data and race/ethnicity data aren’t readily available for these new geographies, so we had to do the analyses and tabulations ourselves. Over the course of several months, we calculated 2020 election data and demographics for the new districts to measure the impacts of these changes. Making the analysis as accurate as possible was in some cases painstaking and tedious work. The smallest geography we had available for 2020 election data was at the precinct level, though in many cases new districts were formed by splitting precincts. In these cases, it was necessary to do another spatial join of voting-age population at the block level to the split precincts so we could estimate the number of votes that should be allocated to each portion of the split precinct. This was often a manual process, inspecting the map for splits that the software may not have identified, or dealing with census blocks that were split by the new lines, which required analysis of satellite imagery to determine how to best allocate that block’s population. This added significant time to our analysis, but was worth the effort to provide the best estimate possible.
Breaking news often pulled us off this project so it took significant time to pull the data together as we juggled other priorities on a small team.
What can other journalists learn from this project?
We spent time considering what questions readers would have about the process and what would be helpful context for them to understand the effects, a crucial step in the process that journalists should factor in when deciding which information to show and explain. It was very important for us to add those additional layers of context so that readers could understand not only how the lines changed, but who changed them and who was affected by those changes. To that end, we added who was responsible for selecting the final maps both in 2010 and 2020 and an additional section on demographics for each district for the old and new maps to understand who lives there.