Replications in Context: A Framework for Evaluating New Methods in Quantitative Political Science (with Jeff Harden and Anand Sokhey). Political Analysis, Vol. 27 No. 1, 2019: 119-125. Read the paper here.
How should quantitative political scientists evaluate the practical utility of new methods? Political methodologists often justify a new method’s value, in part, by reporting results from a small number of replication studies for which the new method yields divergent substantive conclusions compared to the original research. We contend that this approach encourages replication selection bias and establishes inadequate evaluation criteria, which may lead to overconfidence in new quantitative methods. We propose an alternative evaluation framework that involves preregistering a replication plan, collecting a representative sample of replication studies, and formally assessing the average difference between existing and new methods from a substantive perspective. We employ this framework in evaluating several examples of newly-introduced methods. We conclude that our approach sets a more rigorous standard for assessing whether a new method is useful to applied researchers and complements the discipline’s increasingly robust norms of transparency and easy access to replication data.
Influence in State Legislatures
This article examines the process and implications of political influence in U.S. state legislatures. I argue that legislators’ reliance on one another for information creates a distribution of informal power that can be used by any member, regardless of rank or seniority, to achieve political goals. Using ten years of temporal cosponsorship data from seven U.S. states, I test a theory of asymmetric polarization in co-partisan influence. Results show that the distribution of members’ ideology is central to the importance of ideological cues. Unlike traditional models of legislative bargaining, I find that ideologically extreme officials increasingly influence moderates, but this effect is only present in heterogenous chambers. Contextual evidence suggests that these results are driven by conservative Republicans, with influence occurring among ideologically proximate but increasingly extreme Republican dyads, with little evidence of systematic tie formation among pairs of Democrats. Together, these results imply that directional movement in party ideology may be partially achieved through spatial decisions made over a sustained period of time.
Political Institutions & Legislative Cue-Taking Networks
How does variation in state legislative institutions affect representatives’ ability to give and receive informational cues? Does limited cue-taking ability make legislators more responsive to their constituents? Cue-taking, or a legislator’s use of his or her colleagues as a source of information, is among the most powerful and consistent sources of information used by legislators (Kingdon, 1975). However, the ease and utility of this behavior likely changes across states as a result of variation in the institutions that condition the legislative environment. Term limits, for example, may constrain representatives’ ability to take cues from party leaders, given the high rate of turnover in term-limited legislatures. In turn, these legislators may be forced to rely on other, extra-legislative sources of information to make decisions, such as public opinion. This paper analyzes the extent to which state and chamber-level institutions empower or constrain legislative influence in 8 U.S. states. In addition, I test the implications of these differences for policy outcomes, exploring whether relatively less powerful legislators produce policy that better represents constituent opinion. To do this, I first infer a series of latent influence networks using temporal cosponsorship data, showing which legislators are most powerful across 101 state sessions. Then, using the resulting networks, I explore the differences among them using exponential random graph modeling (ERGM), to show how variation across institutional context affects the structure of influence. I contrast these networks using both aggregate statistics and individual-level measures. This shows, for example, how term-limited legislatures compare to those without in terms of average measures, as well as variation in the level of cue-giving power at the top of the influence hierarchy. Finally, I use a multilevel modeling with post-stratification procedure to test the congruence between public opinion and policy. This research has implications for research in state politics, political institutions, legislative studies, and public opinion.
Assessing the Value of Political Influence to Voters: Evidence from a Survey Experiment Pre-analysis Plan
Given that political influence is central to affecting policy outcomes, how do voters account for incumbents’ level of this quality when deciding how to vote? Political power, like any other candidate attribute, such as party, race, or gender, signals a connotation to voters – “correctly” or otherwise (Lau & Redlawsk, 1997) – about that candidate’s ability to affect policy changes in their preferred direction. However, this attribute primarily benefits political elites themselves; specifically, influence serves as a more accurate heuristic to similarly informed others in intra-legislative settings. Put differently, legislators look to the most influential one among their peers to serve as a cue-giver in otherwise low-information settings (Kingdon, 1973; Matthews & Stimson, 1975), while voters lack the context to utilize such nuanced candidate information. This paper examines the role of influence from the perspective of constituents with two survey experiments. First, using a conjoint design, I test whether politically influential candidates are perceived more positively than others. Here, respondents are presented with two hypothetical candidate profiles in which influence along with five additional traits are randomly varied. This test shows the independent contribution of influence to voters’ evaluations of candidates relative to other traits, such as party, gender, and race. An additional experiment puts influence into a tradeoff context to determine respondents’ representational preferences. In short, influence requires credibility and professional rapport, which must be maintained through consistent, highly visible activity. As a result, these legislators may be less able to provide certain types of representation, such as constituency service, which necessitate spending more time in the district and less time at the capitol.
Unconstrained Access: Immigration Policy Diffusion Across the European Union (with Andrea Peña-Vasquez)
Since the mid-2000s, widespread economic decline and the resurgence of populist right parties across the European Union have substantially reduced social welfare spending in most member states. At the same time, Europe has experienced a “migration crisis,” (Vespe et al., 2017) as an influx of nearly 4 million have resettled within its borders; the most vulnerable among them being the unknown number of “irregular migrants,” or those without the host country’s legal permission to enter (Vollmer, 2011) or reside there after initial entry (European Migration Network, 2014). Despite the unfavorable political and economic conditions, several EU nations, such as Italy and Spain, have nonetheless maintained programs that provide social services to their irregular residents. Using Network Inference (Gomez-Rodriguez et al., 2010; Desmarais et al., 2015) combined with qualitative case studies, we contrast the policy diffusion of immigrant incorporation in Europe. Specifically, we compare the policy “leaders” and “followers” in both instances in addition to describing how the factors associated with the initial adoption of these policies compares to their contemporary exercise.