Vote Share Impacts of a Candidate Postcard Campaign: Up- and Down-Ballot Results from a Partially Modified Randomization Trial in Pennsylvania Senate District 37, November 2024

Evaluations of mail campaigns to voters have focused predominantly on individual-level voter turnout, while vote share — the more consequential electoral outcome — has rarely been examined in rigorous experimental evaluations of campaign mail. This study addresses that gap by evaluating precinct-level vote-share impacts of a candidate postcard campaign in Pennsylvania State Senate District 37 (SD37), Allegheny County, November 2024. The intervention employed a partially modified two-step randomization design. First, each of 187 precincts was assigned a randomly drawn postcard penetration rate (target fraction of all registered voters to receive a card), ranging from near zero to approximately 40%. Second, individual recipients were randomly selected within each precinct from a pool of pre-identified target voters (modeled turnout probability 50–80%; Democratic support score ≥70). In 56 precincts where eligible targets were insufficient to meet the assigned rate, the dose was adjusted downward to the number of available targets. This design also simultaneously enables estimation of a precinct-level dose-response curve and a traditional binary treatment-versus-control analysis of individual-level voting turnout effects to be reported in a forthcoming study. Approximately 36,000 voters received postcards featuring Democratic challenger Nicole Ruscitto. Generalized linear model regressions controlling for prior election vote shares, partisan composition, turnout history, demographics, and legislative district boundaries yielded consistently positive vote-share effects across three Democratic candidates (Ruscitto, U.S. Senate candidate Robert Casey, and presidential candidate Kamala Harris). Under the preferred quadratic specification and contrary to the null result in a simple bivariate analysis, estimated gains from raising penetration to 35% ranged from +0.7% to +2.1%, with the largest effects observed for Harris — an up-ballot finding warranting further investigation. A notable secondary result was a negative vote-share response at low penetration levels (below ~15%), most pronounced for Casey, suggesting light-touch mailing may be counterproductive. These findings highlight the importance of confounder adjustment in precinct-level studies, the potential value of mail interventions in high-turnout races, and the advantages of variable-dose over binary designs for informing campaign strategy decisions.