Although political scientists have used spatial statistical models in their research for only about ten years, their use has quickly advanced in terms of both the breadth of substantive areas in which researchers apply them and the sophistication with which they specify and interpret them. The collection of articles in this special edition are indicative of both of these trends. They provide a strong set of examples of how scholars can bring theories about spatial relationships across units into all of the major subfields of the discipline.

This paper expands traditional predatory theory approaches to state fiscal capacity by adopting spatial analytical reasoning and methods. Although previous work in the predatory theory tradition has often incorporated interdependent external influences, such as war and trade, it has often done so in a way that maintains a theoretical and empirical autonomy of the state. Theoretically, we suggest four mechanisms (coercion, competition, learning, and emulation) that operate to channel information through interstate rivalry and territorial contiguity, trade networks, and the political space associated with regime type and intergovernmental organization membership. We test our predictions using a multi-parametric spatio-temporal autoregressive model with four spatial lags capturing the four mechanisms. Our empirical results provide support for the coercion and learning mechanisms.

This study examines whether acts of anti-government violence exhibit spatial dependence across state boundaries. In other words, to what extent can acts of anti-government violence in one country be attributed to violence in neighboring countries? Past research, which has largely focused on civil war or large-scale conflict contagion, finds that geographically proximate states are more likely to experience the cross-boundary diffusion of conflict due to action emulation. However, this assumes that actors are fully aware of conflicts occurring in neighboring countries. To address this, the article argues that the proliferation of communication technology increases access to information about events in neighboring states, thereby allowing emulation to occur and subsequently conditioning the potential for violence to spread. It tests this expectation by modeling the effects of a unique spatial connectivity matrix that incorporates both state contiguity and access to communication technology. An analysis of all acts of anti-government violence in 44 African countries from 2000 to 2011 supports the argument.

Although a voluminous literature has shed light on the relationship between economic conditions and government accountability, most studies in this literature have implicitly assumed that the actions of competing political parties are either irrelevant or that they cancel each other out. In this paper, we take an important first step toward relaxing this strong assumption. We develop and test a set of theoretical propositions from the issue competition literature about the amount of emphasis that parties place on the economy during election campaigns. We test these propositions with an estimation technique that properly situates the motivations of rival elites within the context of spatial party competition using a spatial autoregressive model. On a sample of 22 advanced democracies from 1957 to 2006, we find strong support for the proposition that parties with a greater role in economic policymaking respond to worsening economic conditions by increasing their emphasis on the economy during election campaigns. We also find strong evidence of spatial contagion effects as parties respond positively to the campaign strategies of ideologically proximate parties. This finding reveals a fundamental link in the chain of economic accountability and has important implications for the study of party competition.

Political scientists often attempt to exploit natural experiments to estimate causal effects. We explore how variation in geography can be exploited as a natural experiment and review several assumptions under which geographic natural experiments yield valid causal estimates. In particular, we focus on cases where a geographic or administrative boundary splits units into treated and control areas. The different identification assumptions we consider suggest testable implications, which we use to establish their plausibility. Our methods are illustrated with an original study of whether ballot initiatives increase turnout in Wisconsin and Ohio, which illustrates the strengths and weaknesses of causal inferences based on geographic natural experiments.

We develop a new approach for modeling public sentiment by micro-level geographic region based on Bayesian hierarchical spatial modeling. Recent production of detailed geospatial political data means that modeling and measurement lag behind available information. The output of the models gives not only nuanced regional differences and relationships between states, but more robust state-level aggregations that update past research on measuring constituency opinion. We rely here on the spatial relationships among observations and units of measurement in order to extract measurements of ideology as geographically narrow as measured covariates. We present an application in which we measure state and district ideology in the United States in 2008.

The study of international relations (IR), and political science more broadly, has derived great benefits from the recent growth of conceptualizing and modeling political phenomena within their broader network contexts. More than just a novel approach to evaluating old puzzles, network analysis provides a whole new way of theoretical thinking. Challenging the traditional dyad-driven approach to the study of IR, networks highlight actor interdependence that goes beyond dyads and emphasizes that many traditional IR variables, such as conflict, trade, alliances, or international organization memberships must be treated and studied as networks. Properties of these networks (e.g., polarization, density), and of actor positions within them (e.g., similarity, centrality), will then reveal important insights about international events. Network analysis, however, is not yet fully adapted to account for important methodological issues common to IR research, specifically the issue of endogeneity or possible nonindependence between actors’ position within international networks and the outcomes of interest: for example, alliance network may be nonindependent from the conflict or trade network. We adopt an instrumental variable approach to explore and address the issue of endogeneity in network context. We illustrate the issue and the advantages of our approach with Monte Carlo analysis, as well as with several empirical examples from IR literature.

Spatial/spatiotemporal interdependence—that is, that outcomes, actions or choices of some unit-times depend on those of other unit-times—is substantively important and empirically ubiquitous in binary outcomes of interest across the social sciences. Estimating and interpreting binary-outcome models that incorporate such spatial/spatiotemporal dynamics directly is difficult and rarely attempted, however. This article explains the inferential challenges posed by spatiotemporal interdependence in binary-outcome models and recent advances in their estimation. Monte Carlo simulations compare the performance of one of these consistent and asymptotically efficient methods (maximum simulated likelihood, using recursive importance sampling) to estimation strategies naïve about (inter-) dependence. Finally, it shows how to calculate, in terms of probabilities of outcomes, the estimated spatial/spatiotemporal effects of (and response paths to) hypotheticals of substantive interest. It illustrates with an application to civil war in Sub-Saharan Africa.

In spatial econometrics, W refers to the matrix that weights the value of the spatially lagged variable of other units. As unimportant as it may appear, W specifies, or at least ought to specify, why and how other units of analysis affect the unit under observation. This article shows that theory must inform five crucial specification choices taken by researchers. Specifically, the connectivity variable employed in W must capture the causal mechanism of spatial dependence. The specification of W further determines the relative relevance of source units from which spatial dependence emanates, and whether receiving units are assumed to be identically or differentially exposed to spatial stimulus. Multiple dimensions of spatial dependence can be modeled as independent, substitutive or conditional links. Finally, spatial effects need not go exclusively in one direction, but can be bi-directional; recipients can simultaneously experience positive spatial dependence from some sources and negative dependence from others. The importance of W stands in stark contrast to applied researchers’ typical use of crude proxy variables (such as geographical proximity) to measure true connectivity, and the practice of adopting standard modeling conventions rather than substantive theory to specify W. This study demonstrates which assumptions these conventions impose on specification choices, and argues that theories of spatial dependence will often conflict with them.

Evans and Chzhen (2016a) challenge Clarke et al.’s (Clarke et al. 2004; Clarke et al. 2009) “valence” model of voting behavior in British elections. Specifically, they take issue with results based on individual-level analyses relating subjective performance assessments to the (self-reported) vote measured concurrently. The argument is a basic one: there is reason to think that evaluations of performance are caused by—and not causes of—the vote. That this may be true is supported by a large and growing body of research in political science, much of which focuses on economic voting (e.g., Campbell et al. 1960; Kramer 1983; Wlezien et al. 1997; Bartels 2002; Anderson, Mendes and Tverdova 2004; Evans and Andersen 2006; Ladner and Wlezien 2007; Evans and Pickup 2010). Estimates of valence effects in Clarke et al.’s individual-level analyses thus may be correspondingly biased. It is an important issue for the study of valence and also other electoral research relying on perceptual and attitudinal predictors.

The influential valence model of voting developed over the last decade by the British Election Study (BES) team assumes that party and leadership performance evaluations have a causal impact on party choice. An alternative perspective argues that such performance evaluations are instead the consequences of party choice. This article examines the analytical and empirical underpinnings of the BES valence model and compares it to the party-driven approach. To do so, it estimates cross-lagged structural equation models of the association between Labour Party preference and evaluations of the Labour government's performance during the 2005–10 British electoral cycle. It shows that party preference has a stronger effect on performance evaluations than vice versa; performance evaluations have no significant effect on party preference toward the end of the electoral cycle. The study also finds that, contrary to claims made concerning their merits as simplifying heuristics, performance assessments have no impact on short-term movements in party choice for less politically attentive voters. To a substantial degree, evaluations of party performance express—rather than explain—party choice, and would appear to have limited merit as simplifying heuristics.

This paper responds to Evans and Kat’s critique of the valence politics model of electoral choice. Their critique is deficient in several respects. First, the authors do not test the valence politics model, which is motivated by a theory of voting rather than a claim about the relationship between generalized measures of “party preference” and “party performance.” Second, Evans and Kat do not provide theoretical grounding for partisanship, which they claim is strongly exogenous to other variables of interest. Third, there are several specification and testing problems with their structural equation model. We study the properties of the valence model using a vector error correction model of aggregate monthly survey data gathered throughout the New Labour Era. Consistent with theoretical expectations, key valence politics variables constitute a powerful cointegrated system in which the dynamics of partisanship are endogenous to other variables in the system.

Whiteley et al. criticize our re-analysis of the valence model of party choice. In reply, we argue that they are mistaken with regard to their understanding of some of the claims made in our paper and, in one instance, the variables included in the analysis. We also point to flaws in their interpretation of the distinction between weak and contemporaneous endogeneity and the assumptions of their vector error correction model. Their argument privileging aggregate analysis introduces predictable inferential problems and also, it might be suggested, potentially casts doubt on their own research, which uses individual-level analysis to make the case for the valence model. For these reasons, we see no reason to moderate the negative implications of our analysis for the validity of the model.