Although canonical measures indicate the parties in Congress are increasingly ideologically cohesive, conflict within majority parties on procedural votes is rising. I explain this discrepancy with a new theory that predicts regardless of the parties’ ideological distributions, party leaders’ agenda power depends upon their ability to punish defection. If leaders cannot control resources that induce compliance, members can obstruct procedure to extract concessions without fearing reprisal. I test this theory with a new measure of legislators’ financial reliance on the party-donor network. I show obstructionist caucuses have grown less reliant on their party’s network of corporate PACs, and this financial independence predicts defections within-Congress, within-district, and even within obstructionist factions. I provide evidence these changes weakened leaders’ disciplinary capacity, as the corporate-dollar cost of voting against a GOP speaker election decreased significantly between 2015 and 2023. Governing capacity requires controlling resources legislators desire, not simply a particular distribution of ideology.

Does user engagement shape the issue agendas of congressional candidates on Twitter? We analyze over 3.4 million tweets from all candidates in the 2020 and 2022 election cycles to estimate the engagement returns to discussing different political topics and test whether those returns influence which issues candidates emphasize. We find that tweets referencing divisive national issues — such as guns, abortion, race, Trump's impeachment, January 6th, Biden, and China — receive substantially greater engagement than tweets about the issues voters care most about, including the economy, health care, education, climate, and Covid-19. Yet these incentives do not shape candidates' agendas. In aggregate and at the individual level, candidates predominantly discuss voter-relevant issues, and even those who receive especially large engagement boosts from divisive topics are no more likely to discuss them. Engagement amplifies polarizing issues online, but it does not determine the distribution of issues that candidates choose to emphasize.

Even years after the 2020 election, Donald Trump continues to claim that fraudulent and illegal votes cost him the 2020 election. In this paper we provide the most comprehensive assessment of his empirical claims to date. All of the claims we evaluate fail to provide evidence of fraud or illegal voting. Trump's claims of fraud or illegality are riddled with errors, hampered by misunderstandings about how to analyze official voter records, and filled with confusion about basic statistical techniques and concepts. Often, the claims are based on the casual impressions of what happens in a ``normal" election based on little more than intuitions. Worse yet, several claims are simply misstated by Trump's legal team or Trump. As a result, sometimes the public claims do not even match the weak evidence in Trump's legal challenges. This paper provides a resource for assessing many of the most prevalent claims made about the 2020 election and a guide to anticipating potential objections in future elections.

A large literature argues that social media encourages politicians to be polarizing, because such content receives greater user engagement. This assumes, however, that campaigns adjust their content based on what gets attention. I test this assumption with over 3.7 million tweets sent by all congressional candidates between 2019 and 2022. I introduce a reinforcement-learning-inspired design that examines whether candidates are more likely to repeat content after it goes viral. Using a matching estimator based on sentence embeddings, I compare each viral tweet to a near-identical non-viral tweet and estimate its effect on the similarity of subsequent messages. Contrary to the engagement-maximization hypothesis, I find that virality causes candidates to diverge from the content of viral tweets, both in meaning and topic. These effects are consistent across challengers and incumbents and are unrelated to ideology or electoral vulnerability --- even extreme candidates and those in safe districts do not systematically adjust content in response to engagement. These results temper concerns that user demand for polarizing or uncivil content encourages politicians to strategically supply such content.