This repository incorporates statistics and code helping a BuzzFeed News article inspecting donors on campaign finance. Published August three, 2019. See below for info.DataThis information analyzes ActBlue’s recent mid-yr file to the FEC, which includes donations made via the company between January 1, 2019 and June 30, 2019. (Due to its size, the filing itself isn’t always blanketed on this repository, but commands to download can be determined in the ‘Reproducibility’ segment beneath.)MethodologyLinking donorsThe Federal Election Commission filings do not include any absolutely-precise identifiers for campaign members. So, so as identify donors who have given to more than one campaigns, BuzzFeed News constructed a donor_id, constructed from the subsequent fields:First callLast namefive-digit ZIP codeThere are some limitations to this approach:If a donor modifications their name, or misspells it every so often, this method will no longer cluster all of their contributions collectivelyIf a donor moves to a brand new ZIP code, this approach will not cluster all of their contributions collectivelyIfor extra donors inside the equal ZIP code share each a primary and ultimate call, this method will count on (incorrectly) that they’re the identical man or womanFor these reasons, the effects of the analysis should be interpreted as approximations.Contribution totals above felony restrictionThe FEC prohibits character donors from giving greater than $2,800 to any single committee. ActBlue will save you a donor from giving extra than the character restrict in a unmarried donation, but it’s miles possible for a donor to offer greater in mixture across many donations. Typically, the candidate’s committee will refund the difference at a later date which won’t be contemplated on this filing. Above-criminal contributions haven’t any effect at the analyses, which focus at the act of giving instead of how a whole lot money the campaigns have raised.AnalysisThe notebooks/codonors.ipynb pocket book includes the main analysis, written in Python. Relevant outputs may be located there, in addition to in the output/ listing.OutputThe output/ listing includesdocuments that can be of hobby to other newshounds and researchers:output/candidate-pair-counts.csv counts the number of donor_ids that gave to every aggregate ofcandidates.output/candidate-triplet-counts.csv counts the range of donor_ids that gave to every mixture of three applicants.In both files, combinations aren’t extraordinary. For example, a person who gave to four applicants could be counted for every permutation (i.e., six pairs and four triplets).ReproducibilityThe code walking the analysis is written in Python three, and requires the following Python libraries:pandas for statistics loading and evaluationfecfile for parsing the uncooked FEC filingsjupyter to run the pocket book infrastructureIf you operate Pipenv, you may set up all required libraries with pipenv installation.To down load the filing, run make filings. Warning: this could take numerous minutes and down load a five+ gb document.Run make load or execute notebooks/load_actblue.ipynb to parse the raw FEC documents.Execute notebooks/codonors.ipynb reproduce the findings.LicensingAll code on this repository is to be had underneath the MIT License. Files within the output/ listing are to be had underneath the Creative Commons Attribution 4.0 International (CC BY 4.zero) license.Questions / FeedbackContact Scott Pham at [email protected] for more from BuzzFeed News? Click right here for a list of our open-sourced tasks, data, and code.