Hyperlocal Biden/Trump results. To create their “Extremely Detailed Map of the 2020 Election,” journalists at The Upshot have gathered and standardized the ballot results for more than 100,000 precincts in dozens of states so far — representing nearly two-thirds of all votes cast. The accompanying GeoJSON dataset indicates the number of votes received by Joe Biden, by Donald Trump, and overall (including third-party and write-in candidates), joined to each precinct’s geographic boundaries. [h/t Kevin Quealy + Ryan Matsumoto]
Subnational COVID-19 vaccinations. Software developer Lucas Rodés-Guirao is aggregating coronavirus vaccination numbers, over time, by administrative division — each canton in Switzerland, for example, and each province in Argentina. The project, inspired by Our World In Data’s country-level dataset (DIP 2020.12.23), currently includes 20+ countries and provides links to the sources. [h/t Olivier Lejeune]
Historical night light. Researchers have harmonized two major datasets measuring the light visible on Earth at night. The first, from the Defense Meteorological Satellite Program, covers 1992–2013. The second, from the Visible Infrared Imaging Radiometer Suite, covers 2012 onward. The harmonized dataset covers 1992–2018 and polishes the data by removing, for example, “noises from aurora, fires, boasts, and other temporal lights.” Related: The Colorado School of Mines’ Earth Observation Group, which “specializes in nighttime observations of lights and combustion sources worldwide.” [h/t Milos Popovic]
Public-sector employment. The World Bank’s Worldwide Bureaucracy Indicators “is a unique new cross-national dataset on public sector employment and wages” in 130+ countries between 2000 and 2018, based on censuses and household surveys. The dataset’s measurements include the size of this workforce, its demographics, pay disparities versus the private sector, and more. Related: An introductory Twitter thread from coauthor Faisal Baig.
Super Bowl ads. Journalists at FiveThirtyEight watched 233 Super Bowl ads run by 10 frequently-advertising brands between 2000 and 2020, drawing from superbowl-ads.com’s video archive. Then they categorized each ad according to seven specific questions, which included “Was it trying to be funny?” and “Did it include animals?”