Volunteers sent fully handwritten postcards using colored markers to 3280 female voters in Wisconsin, encouraging them to vote for specified state senate candidates during the 2018 midterm elections. Their turnout was 2.5% lower than 1333 who received no postcard. Card writers worked sequentially through 4 different voter lists with different mailing date windows, but they did not have time to finish all lists, which the authors characterized as “quasi-random assignment” to treatment vs. omitted subject groups. The groups did not differ on voting propensity in earlier elections.
The authors proposed several potential explanations why turnout impact may have been negative, the third of which they felt was most likely. First, voters may resist with backlash when pressured to vote. Second, some voters may procrastinate if they receive a reminder of an election that is several weeks away. Third, the postcards about state senate candidates may have distracted voters from attending to other more salient races at the top of the tickets.
(Bankston & Burden, 2021, Bankston & Burden appendix, 2021; these are PDF files in a folder on the FLRN Google drive) [Friedland, L.](Newby, R)
Comment by David Salkever: Buried in the supplemental material or appendix is the important factor of when letters were sent. Much research has confirmed that get-out-the-vote efforts are most effective closer to election dates, except for messages that are late enough to miss early voters. Here were the 4 voter lists, in order of mailing. Also listed are descriptions of the impacts for each mailing group, for which the article gives clues but not exact numbers:
Number who Number who did Weeks before election Weeks before election Impact on
received cards not receive cards that mailing started that mailing ended turnout
274 9 8 7 medium negative
246 63 7 6 large negative
1449 555 6 3 small negative
962 706 3 1 large positive
The earliest mailed cards to the smallest number of voters contributed the bulk of the negative impact. Mailings closer to the election had small negative impact then large positive impact. Time of mailing matters, though these data do not answer how much each potential explanation may have contributed to the overall negative effect.
A tip from David for those of us who read original research papers: “A suggestion for all careful readers of empirical papers. Read the tables and numbers first (whether in the appendices or not) regardless of the reputation of the authors. The text is story-telling; the truth (or at least half-truth) is in the numbers.”