Political selection is important to authoritarian regime survival. Although selection outcomes are often observed, it is difficult to decipher elites’ logic of decision-making due to their multidimensional preference and the opaque process. Employing a conjoint experiment conducted among over 300 government officials in China, this paper unpacks multidimensional elite preference in entry-level political selection. It finds that while elites comply with institutional norms by selecting candidates based on competence and loyalty, they also take into account personal preference by favoring those with political connections. Kinship ties to government increase a candidate’s chance by over 20 percentage points, even though the candidate is not deemed more competent or loyal. These findings demonstrate that authoritarian elites do not always follow the logic of regime survival; there exists an agency problem in political selection that could undermine the effectiveness of the ruling class over time.

Collecting network information on political elites using conventional methods such as surveys and text records is challenging in authoritarian and/or conflict-ridden states. I introduce a data collection method for elite networks using scraping algorithms to capture public co-appearances at political and social events. Validity checks using existing data show the method effectively replicates interaction-based networks but not networks based on behavioral similarities; in both cases, measurement error remains a concern. Applying the method to Nigeria illustrates that patronage—measured in terms of public connectivity—does not drive national oil companies appointments. Given that theories of elite behavior aim to understand individual-level interactions, the applicability of data using this technique is well-suited to situations where intrusive data collection is costly or prohibitive.

Government instability (cabinet duration) is an important feature of parliamentary democracy. Over time, the research on cabinet duration has improved in technical and theoretical sophistication. However, we note that little attention has been paid to the relationship between governments and the state itself. Our main hypothesis is that state capacity, e.g., factors such as state bureaucratic effectiveness and law and order, shape how easy it is for governments to implement the new policy and thus how well they can achieve policy objectives. We also argue that when state capacity is low, the ability to adequately respond to external shocks goes down, and instability increases. When testing this empirically we find that low state capacity does indeed help us predict an increased risk for early termination—in particular, whether the government ends through a replacement (but not by an early election). Using interaction effects, we also demonstrate that the effect of external shocks, such as an increase in unemployment, is conditional on state capacity. An increase in unemployment only has a significant effect on cabinet stability when state capacity is low, suggesting that the cabinet’s (in)ability to address the economic problems is an important factor for understating cabinet durability.

Although equality figures prominently in many foundational theories of democracy, liberal and electoral conceptions of democracy have dominated empirical political science research on topics like political regimes, democratization and democratic survival. This paper develops the concept of egalitarian democracy as a regime that provides de facto protection of rights and freedoms equally across the population, distributes resources in a way that enables meaningful political participation for all citizens and fosters an environment in which all individuals and social groups can influence political and governing processes. Using new indicators from the Varieties of Democracy project, the paper develops and presents measures of these important concepts, demonstrates their relationship to existing measures, and illustrates their utility for advancing the study of democracy in ways that more fully embrace the richness of democratic theory.

Improving geolocation accuracy in text data has long been a goal of automated text processing. We depart from the conventional method and introduce a two-stage supervised machine-learning algorithm that evaluates each location mention to be either correct or incorrect. We extract contextual information from texts, i.e., N-gram patterns for location words, mention frequency, and the context of sentences containing location words. We then estimate model parameters using a training data set and use this model to predict whether a location word in the test data set accurately represents the location of an event. We demonstrate these steps by constructing customized geolocation event data at the subnational level using news articles collected from around the world. The results show that the proposed algorithm outperforms existing geocoders even in a case added post hoc to test the generality of the developed algorithm.

To what extent do survey experimental treatment effect estimates generalize to other populations and contexts? Survey experiments conducted on convenience samples have often been criticized on the grounds that subjects are sufficiently different from the public at large to render the results of such experiments uninformative more broadly. In the presence of moderate treatment effect heterogeneity, however, such concerns may be allayed. I provide evidence from a series of 15 replication experiments that results derived from convenience samples like Amazon’s Mechanical Turk are similar to those obtained from national samples. Either the treatments deployed in these experiments cause similar responses for many subject types or convenience and national samples do not differ much with respect to treatment effect moderators. Using evidence of limited within-experiment heterogeneity, I show that the former is likely to be the case. Despite a wide diversity of background characteristics across samples, the effects uncovered in these experiments appear to be relatively homogeneous.

Political science research frequently models binary or ordered outcomes involving related processes. However, traditional modeling of these outcomes ignores common data issues and cannot capture nuances. There is often an excess of zeros, the observed outcomes for different actors are inherently related, and competing actors may respond to the same factors differently. This paper extends existing models and develops a zero-inflated multivariate ordered probit to simultaneously address these issues. This model performs better than existing models at capturing the true parameters of interest, estimates the nature of the related processes, and captures the differences in actors’ decision-making. I demonstrate these benefits through simulation exercises and an application to party behavior in Mexico.

While many scholars have moved toward using individual-level data to test theories of distributive politics, no studies have ever explicitly examined differences between individual and aggregate analyses of a distributive program. By leveraging nationwide individual-level data on both revealed voter preferences and the actual receipt of particularistic benefits through a contemporary Venezuelan land reform initiative, this article demonstrates that scholars can most effectively test and refine individual-level theories of distributive politics by combining both individual- and macro-level data. There are at least two advantages to doing so. First, comparing and contrasting findings from data at different levels of analysis can enable researchers to paint a more complete picture of distributive targeting. Second, when distributive benefits can be impacted or redirected by subnational politicians, as is common with many distributive programs, individual-level data alone can generate mistaken inferences that are an artifact of competing targeting attempts at different levels of government instead of initial targeting strategies. I demonstrate both of these points and discuss practical and simple recommendations regarding data collection strategies for the purposes of effectively testing theories of distributive politics.

When and to what extent do crises and significant events induce changes in political attitudes? Theories of public opinion and policymaking predict that major events restructure public opinion and pry open new political opportunities. We examine the effect of major events on support for public policies in the context of the Sandy Hook Elementary School mass shooting in December 2012 using a nationally representative panel survey of US adults. Across both cross-sectional and within-subject analyses, we find no evidence that Americans granted greater support for gun control after the Sandy Hook shooting. Our null findings persist across a range of political and demographic groups. We also find no evidence of attitude polarization as a result of Sandy Hook. Our results suggest that elite polarization in a particular issue area leads citizens to employ motivated reasoning when interpreting critical events, thereby reducing the capacity for attitude change. Our findings have important implications for identifying the conditions under which major events affect support for public policies and create political opportunities for policy change.

Textual data are plagued by underreporting bias. For example, news sources often fail to report human rights violations. Cook et al. propose a multi-source estimator to gauge, and to account for, the underreporting of state repression events within human codings of news texts produced by the Agence France-Presse and Associated Press. We evaluate this estimator with Monte Carlo experiments, and then use it to compare the prevalence and seriousness of underreporting when comparable texts are machine coded and recorded in the World-Integrated Crisis Early Warning System dataset. We replicate Cook et al.’s investigation of human-coded state repression events with our machine-coded events, and validate both models against an external measure of human rights protections in Africa. We then use the Cook et al. estimator to gauge the seriousness and prevalence of underreporting in machine and human-coded event data on human rights violations in Colombia. We find in both applications that machine-coded data are as valid as human-coded data.

I study how the quality of information affects politician selection in a two-candidate model where voters want to vote for the best candidate but also for the winner. Voters receive private and public signals about candidates’ relative valence. Public information has a stronger effect on equilibrium outcomes because voters use it to infer other voters’ beliefs. Contrary to what might be expected, more precise public information does not always benefit the better candidate’s electoral prospects: when voters’ private information is precise enough, improving public information hurts the better candidate’s electoral prospects. The model provides a rationale for the prevalence of large swings in voter sentiment in close elections, and for front-runner candidates’ tendency to avoid face-to-face television debates with the underdog.

Bivariate probit models are a common choice for scholars wishing to estimate causal effects in instrumental variable models where both the treatment and outcome are binary. However, standard maximum likelihood approaches for estimating bivariate probit models are problematic. Numerical routines in popular software suites frequently generate inaccurate parameter estimates and even estimated correctly, maximum likelihood routines provide no straightforward way to produce estimates of uncertainty for causal quantities of interest. In this note, we show that adopting a Bayesian approach provides more accurate estimates of key parameters and facilitates the direct calculation of causal quantities along with their attendant measures of uncertainty.

The Cox proportional hazards model is a popular method for duration analysis that is frequently the subject of simulation studies. However, no standard method exists for simulating durations directly from its data generating process because it does not assume a distributional form for the baseline hazard function. Instead, simulation studies typically rely on parametric survival distributions, which contradicts the primary motivation for employing the Cox model. We propose a method that generates a baseline hazard function at random by fitting a cubic spline to randomly drawn points. Durations drawn from this function match the Cox model’s inherent flexibility and improve the simulation’s generalizability. The method can be extended to include time-varying covariates and non-proportional hazards.