Visualization by Sam Learner |
| 
|
Code and data for this project lives here. You can check out a companion project on donor demographics here.
Notes
Contribution data runs through 10/15/2020.
Unique donors are determined using a combination of first name, last name and zipcode. This creates the possibility of both false positives (two Jane Smiths in the same zipcode) and false negatives (Jonathan Smith sometimes donates as Jon Smith, or Jon Smith moves homes during the election cycle). Others who have worked with donor data sometimes require a matching employer name as well. This drastically lowers the chance of a false positive (it's very unlikely that there are two Jane Smiths living in the same zipcode with the same employer), but creates many more false negatives (Jon Smith might change jobs mid-election cycle or he might mark his employer as "Smith High School" on one donation and "Smith H.S." on another). On balance requiring a matching employer felt like too stringent of a requirement that would lead to missing too many donor overlaps. As it is, the analysis very likely underrepresents the amount of overlap due to false negatives, but it misses them uniformly, across candidates.
The use of the WinRed fundraising platform among Republican candidates is relatively new and still less prevalent than the ActBlue platform among Democratic candidates. Since much of the small donor ( <$200) data comes from the platforms' filings, GOP candidates generally have fewer identified donors than Democratic candidates. Even among candidates who have adopted these platforms, many small transactions take place outside and the campaigns are only required to report contributions from donors who give more at least $200 in aggregate. This means we don't have a true donor count for any campaign, but we can generally get much closer for Democratic ones.
Sources