My dissertation, Influence in State Legislatures, explores the process and implications of political influence as a means to legislative information. Substantial evidence suggests that lawmakers use the behavior of their colleagues to inform their own preferences more frequently than other possible sources of the same information, such as public opinion, interest groups, or the executive. Despite being a largely latent process, the information hierarchy created by this system is a central organizing component of most observable legislative outcomes, including policymaking, agenda setting, and internal advancement. However, relative to traditional conceptualizations of political power, which tend to measure an actor’s capacity according to institutional rank or connectedness to other members, I argue that influence creates an unobserved distribution of informal power, which can be used by any member - regardless of seniority or rank - to affect political outcomes. In this framework, I argue that the ease, utility, and implications of influence vary both within and across states, as well as over time, resulting from differences in the incentive structures facing legislators in diverse political contexts. Consequently, the extent to which influence is a reliable heuristic, in addition to the interests it ultimately reflects, are likely to differ systematically across context, party, and over time.

This project is structured in three empirical chapters, each using a distinct methodological approach to comprehensively define and analyze this concept. In the first chapter, I infer networks of influence based on the timing of cosponsorship decisions across state legislatures, model the formation of ties in those networks, and compare how they have changed over time and differ by party and state context. To infer the networks, I use NetInf, an algorithm developed by Gomez-Rodriguez et al. (2010, 2012), that relies on temporal data describing the time and order in which actors engage in an activity. In this case, I use ten years of cosponsorship data from eight U.S. legislatures. With this information, the algorithm identifies a directed network for each state-chamber-biennium that can be used to partially explain legislators’ decisions across a large number of repeated events, or cascades. Here, influence is measured according to the most consistent relationships over time, using four main conditions: 1) the temporal order of legislators' behavior, 2) the length of time between them, 3) the number of intermediary observations, and 4) the relative predictive capacity of each relationship in a large number of observations. Legislators whose decisions strongly and consistently predict others’ are considered influential - even on a small number of bills - relative to those whose decisions do not.

As a whole, this measure allows for influence to be modeled as comprehensively relational — occurring among all members of a legislative chamber — rather than as an individual or dyadic trait. I then use these results to test a theory about the role of ideology and polarization in cue-giving relationships, using the inferred networks as dependent variables in two analyses to evaluate how the predictors of influence have changed over time and differ both within and across states. Here, I evaluate whether and under what conditions partisans evaluate cues spatially, accepting influence from ideologically proximate members of the opposite party, or directionally, from same-party extremists. Contrary to traditional models of legislative interaction, I find that ideologically extreme actors increasingly influence moderate co-partisans. However, this effect is only present in ideologically heterogenous chambers. Closer examinations of each state network suggest that these results are driven primarily by conservative Republicans who systematically form ideologically-motivated influence relationships, unlike most pairs of Democrats. My results suggest that party-level directional movement can be achieved through incremental spatial decisions over time. Finally, despite the representational tensions posed by this process, evidence from a survey experiment shows that Americans favor influential representation, even when informed of its potential costs.