This special issue is devoted to original articles that reflect recent progress in one of the most exciting developments in Political Science, the National Science Foundation's (NSF) initiative called Empirical Implications of Theoretical Models (EITM). This initiative reflects the ideas and hard work of the Political Science team there, Jim Granato and Frank Scioli, backed up by the contributions of an EITM panel that assembled at NSF in July 2001, some of whose observations we mention below. The challenge set by the EITM program is straightforward: to improve our theoretical work so that it yields more testable hypotheses and to improve our methodological work so that testing is made more effective and informative about theories. It is hard to object to this, but it also turns out to be hard to meet fully. The EITM initiative contains several components designed to close the gap between theoretical derivation and empirical test. This issue represents one component, presenting some of the most innovative work in the discipline on the current research frontier in EITM.

The Internet offers a number of advantages as a survey mode: low marginal cost per completed response, capabilities for providing respondents with large quantities of information, speed, and elimination of interviewer bias. Those seeking these advantages confront the problem of representativeness both in terms of coverage of the population and capabilities for drawing random samples. Two major strategies have been pursued commercially to develop the Internet as a survey mode. One strategy, used by Harris Interactive, involves assembling a large panel of willing respondents who can be sampled. Another strategy, used by Knowledge Networks, involves using random digit dialing (RDD) telephone methods to recruit households to a panel of Web-TV enabled respondents. Do these panels adequately deal with the problem of representativeness to be useful in political science research? The authors address this question with results from parallel surveys on global climate change and the Kyoto Protocol administered by telephone to a national probability sample and by Internet to samples of the Harris Interactive and Knowledge Networks panels. Knowledge and opinion questions generally show statistically significant but substantively modest difference across the modes. With inclusion of standard demographic controls, typical relational models of interest to political scientists produce similar estimates of parameters across modes. It thus appears that, with appropriate weighting, samples from these panels are sufficiently representative of the U.S. population to be reasonable alternatives in many applications to samples gathered through RDD telephone surveys.

This article provides a new maximum-likelihood estimator for selection models with dichotomous dependent variables when identical factors affect the selection equation and the equation of interest. Such situations arise naturally in game-theoretic models where selection is typically nonrandom and identical explanatory variables influence all decisions under investigation. When identical explanatory variables influence selection and a subsequent outcome of interest, the commonly used Heckman-type estimators identify from distributional assumptions about the residuals alone. When its own identifying assumption is reasonable, the new estimator allows the researcher to avoid the painful choice between identifying from distributional assumptions alone and adding a theoretically unjustified variable to the selection equation in a mistaken attempt to “boost” identification. The article uses Monte Carlo methods to compare the small-sample properties of the estimator with those of the Heckman-type estimator and ordinary probit.

Social scientists are often confronted with theories in which one or more actors make choices over a discrete set of options. In this article, I generalize a broad class of statistical discrete choice models, with both well-known and new nonstrategic and strategic special cases. I demonstrate how to derive statistical models from theoretical discrete choice models and, in doing so, I address the statistical implications of three sources of uncertainty: agent error, private information about payoffs, and regressor error. For strategic and some nonstrategic choice models, the three types of uncertainty produce different statistical models. In these cases, misspecifying the type of uncertainty leads to biased and inconsistent estimates, and to incorrect inferences based on estimated probabilities.

This article provides a basic report about subject recruitment processes for Web-based surveys. Using data from our ongoing Internet Survey of American Opinion project, two different recruitment techniques (banner advertisement and subscription campaigns) are compared. This comparison, together with a typology of Web-based surveys, provides insight into the validity and generalizability of Internet survey data. The results from this analysis show that, although Internet survey respondents differ demographically from the American population, the relationships among variables are similar across recruitment methods and match those implied by substantive theory. Thus, our research documents the basic methodology of subject acquisition for Web-based surveys, which, as we argue in our conclusion, may soon become the survey interview mode of choice for social scientists.

This article introduces a model of ordinal unidimensional measurement known as Mokken scale analysis. Mokken scaling is based on principles of Item Response Theory (IRT) that originated in the Guttman scale. I compare the Mokken model with both Classical Test Theory (reliability or factor analysis) and parametric IRT models (especially with the one-parameter logistic model known as the Rasch model). Two nonparametric probabilistic versions of the Mokken model are described: the model of Monotone Homogeneity and the model of Double Monotonicity. I give procedures for dealing with both dichotomous and polytomous data, along with two scale analyses of data from the World Values Study that demonstrate the usefulness of the Mokken model.

Theories that posit complex causation, or multiple causal paths, pervade the study of politics but have yet to find accurate statistical expression. To remedy this situation I derive new econometric procedures, Boolean probit and logit, based on the logic of complexity. The solution provides an answer to a puzzle in the rational deterrence literature: the divergence between theory and case-study findings, on the one hand, and the findings of quantitative studies, on the other, on the issue of the role of capabilities and willingness in the initiation of disputes. It also makes the case that different methodological traditions, rather than settling into “separate but equal” status, can instead inform and enrich one another.

The practice of using point estimates produced by the King ecological inference technique as dependent variables in second-stage linear regressions leads to second-stage results that, in general, are inconsistent. This conclusion holds even when all assumptions behind King's ecological technique are satisfied. Second-stage inconsistency is a consequence of the fact that King-based point estimates of disaggregated quantities contain errors correlated with the true quantities the estimates measure. Our findings on second-stage inconsistency, as well as a fix that we propose, follow from econometric theory in conjunction with an analysis of simulated and real ecological data sets.

Recent work in survey research has made progress in estimating models involving selection bias in a particularly difficult circumstance—all nonrespondents are unit nonresponders, meaning that no data are available for them. These models are reasonably successful in circumstances where the dependent variable of interest is continuous, but they are less practical empirically when it is latent and only discrete outcomes or choices are observed. I develop a method in this article to estimate these models that is much more practical in terms of estimation. The model uses a small amount of auxiliary information to estimate the selection equation parameters, which are then held fixed while estimating the equation of interest parameters in a maximum-likelihood setting. After presenting Monte Carlo analyses to support the model, I apply the technique to a substantive problem: Which interest groups are likely to to be involved in support of potential initiatives to achieve their policy goals?

Survey participation, electoral participation, and political interest have been given wide attention in the research literature, but no one so far has combined these three variables in one model. Taking the social isolation-hypothesis as our starting point, we developed a model with one factor, social involvement, as the common factor underlying these three types of participation. We reviewed the literature and concluded that we had to include a second underlying factor: attachment to society. Using a new data set, gathered on the occasion of the 1998 Dutch national elections and including validated voter turnout measures, we were able to test the model. After making some adaptions, we found a model with a satisfactory fit. The results show that, by including social involvement and attachment to society as mediating variables, we can reach much higher levels of explained variances of survey and electoral participation than we can with traditional models. The results also add to our understanding of the relationship between survey and electoral participation and political interest.

We develop an empirical estimator directly from an extensive-form crisis bargaining game with incomplete information and discuss its features and limitations. The estimator makes it possible to draw inferences about states' payoffs from observational data on crisis outcomes while remaining faithful to the theorized strategic and informational structure. We compare this estimator to one based on a symmetric information version of the same game, using the quantal response equilibrium proposed in this context by Signorino (1999, American Political Science Review 93:279–298). We then address issues of identification that arise in trying to learn about actors' utilities by observing their play of a strategic game. In general, a number of identifying restrictions are needed in order to pin down the distribution of payoffs and the effects of covariates on those payoffs.

In recent work, Signorino (American Political Science Review 93:279–297, 1999; International Interactions 28:93–115, 2002) has sought to test statistical models derived from extensive-form games in the context of international relations research focusing on conflict and interstate bargaining. When two or more actors interact with one another under conditions of uncertainty, Signorino demonstrates that it is necessary to incorporate such strategic interaction into the underlying model to avoid potential threats to statistical inference. Outside the realm of international relations research, however, there have been limited applications of Signorino's strategic probit model in understanding strategic interaction. In this article, I present an empirical comparison of probit and strategic probit models in the context of candidate competition in House elections during the 1990s. I show that incumbent spending deters challenger entry and factors such as minority party affiliation and redistricting significantly affect incumbent career decisions, findings that run counter to those reported in the nonstrategic model. Overall, the results illustrate that failing to account for strategic interaction can lead to biased and inaccurate estimates related to challenger and incumbent entry decisions.

We take this opportunity to comment on Herron and Shotts (2003; hereinafter HS) because of its interesting and productive ideas and because of the potential to affect the way a considerable body of practical research is conducted. This article, and the literature referenced therein, is based on the suggestions in three paragraphs in King (1997, pp. 289–290). Because these paragraphs were not summarized in HS, we thought they might be a useful place to start.

Selection bias is an important but often neglected problem in comparative research. While comparative case studies pay some attention to this problem, this is less the case in broader cross-national studies, where this problem may appear through the way the data used are generated. The article discusses three examples: studies of the success of newly formed political parties, research on protest events, and recent work on ethnic conflict. In all cases the data at hand are likely to be afflicted by selection bias. Failing to take into consideration this problem leads to serious biases in the estimation of simple relationships. Empirical examples illustrate a possible solution (a variation of a Tobit model) to the problems in these cases. The article also discusses results of Monte Carlo simulations, illustrating under what conditions the proposed estimation procedures lead to improved results.

Consumers of the National Election Study (NES) should be concerned if the survey has a bias that is increasing with time. A recent article by Barry Burden claims that for presidential elections, there is an increasing overreport bias, or turnout gap, between the NES turnout rate and the observed turnout rate caused by declining NES response rates. I show that the increasing turnout gap is an artifact of the universes these two turnout rates are based on. Reconciling the two universes shows no systematic increase of the reconciled turnout gap in presidential elections from 1948 to 2000, and furthermore demonstrates that the post-1976 rise in NES response rates (until 2000) is rewarded in a lower turnout gap. In addition, I offer another theory to explain the turnout gap. If respondents have an equal propensity to misreport that they voted when they did not, as turnout declines, the number of nonvoter respondents increases and so does the turnout gap. I show that in multivariate analysis this theory outperforms Burden's response rate driven theory, though neither theory reaches statistical significance.

Scholars of legislative studies typically use ideal point estimates from scaling procedures to test theories of legislative politics. We contend that theory and methods may be better integrated by directly incorporating maintained and to be tested hypotheses in the statistical model used to estimate legislator preferences. In this view of theory and estimation, formal modeling (1) provides auxiliary assumptions that serve as constraints in the estimation process, and (2) generates testable predictions. The estimation and hypothesis testing procedure uses roll call data to evaluate the validity of theoretically derived to be tested hypotheses in a world where maintained hypotheses are presumed true. We articulate the approach using the language of statistical inference (both frequentist and Bayesian). The approach is demonstrated in analyses of the well-studied Powell amendment to the federal aid-to-education bill in the 84th House and the Compromise of 1790 in the 1st House.

We address in Herron and Shotts (2003; hereinafter HS) an increasingly common statistical practice called EI-R, in which point estimates generated by the King (1997) ecological inference technique are used as dependent variables in second-stage linear regressions. Although HS argue that EI-R slope estimates are inconsistent and suffer from attenuation bias, this characterization of EI-R is excessively conservative. Indeed, we show here that EI-R estimates can suffer from sign reversals, attenuation bias, and augmentation bias, that inference based on these estimates can be misleading, and that EI-R is unfixable and should not be used.

This article builds a nonparametric method for inference from roll-call cohesion scores. Cohesion scores have been a staple of legislative studies since the publication of Rice's 1924 thesis. Unfortunately, little effort has been dedicated to understanding their statistical properties or relating them to existing models of legislative behavior. I show how a common use of cohesion scores, testing for distinct voting blocs, is severely biased toward Type I error, practically guaranteeing significant findings even when the null hypothesis is correct. I offer a nonparametric method—permutation analysis—that solves the bias problem and provides for simple and intuitive inference. I demonstrate with an examination of roll-call voting data from the Brazilian National Congress.

A recent article by Barry Burden in Political Analysis alerts us to a steadily increasing gap during presidential election years between self-reported turnout in the NES (National Election Studies) and “official turnout” figures based on the voting-age population (VAP), and points to declining response rates as a culprit. Changing the baseline from the VAP to the VEP (voting-eligible population) significantly changes these conclusions, and point to panel effects as a culprit. The rise in the gap was not linear, but it does emerge rather suddenly in 1996. Gaps between NES self-reported turnout and VEP estimates are higher in presidential election years than in off-years, and self-reported turnout is higher among long-term panel participants than among cross-section respondents in multielection panels.