Trade unions are highly influential political actors in the Western world. Yet, it is difficult to identify their numerous effects on national election results. The main challenge is the high correlation between exogenous shocks affecting unions and shocks affecting left-leaning parties. This study leverages the unique context of the United States to address this identification challenge. A union in the US must be formed through elections regulated by the National Labour Relations Board. Regression Discontinuity Design is often used to estimate the effect of unions on various outcomes by using close unionization elections as a source of exogenous variation. However, such a design can't be used in the context of national election results. Data regarding national elections are not available at the workplace level but only on aggregate geographic units where several close unionization elections could occur in each election cycle. To deal with this challenge, I introduce the novel Regression Discontinuity Aggregation (RDA) methodology that aggregates several discontinuity events--close unionization elections--into a commuting zone level shock that measures the unions' "Luck" in each zone in each period. This shock is an attractive instrument for newly unionized workers' flow. Using this methodology, I find that, on average, a newly unionized worker is worth 1.5 new votes for the Democratic party candidate in the following presidential elections. Further analysis suggests that this significant effect partly stems from increased campaign contributions, strategic political resource allocation by unions in areas with new members, and a leftward shift in local congressional representation.

 document that during 1990-2020, the election of state female legislators in the USA increased the rate of abortions within their states. To establish causality, we introduce an application of the novel Regression Discontinuity Aggregation methodology. The effect we document is driven by Democratic women, who influence abortion rates by molding state reproductive regulation. In contrast, electing a Democratic male leader does not have a similar effect, even though legislators’ parties predict their attitudes towards abortions better than their genders. We show that this divergence between the genders stems from female Democrats granting a higher political priority to reproductive issues than their male counterparts. This finding underscores the importance of accounting for leaders' priorities when predicting policy outcomes. Finally, we employ a traditional Regression Discontinuity Design to replicate our main results for state governors.