Two dynamic rational choice models of primaries are analyzed that demonstrate both reformers and their critics are right about primaries. Primaries can provide public instruction and a better informed electorate, and they can be poorly designed lotteries: A “recognition” and learning model describes how changing media coverage affects voters' knowledge about candidates and their subsequent voting behavior. A “strategic” voting model describes the dynamic implications of strategic voting and “horse-race” coverage by the media. We find that the recognition model has the normatively appealing dynamic of information leading to broadly self-interested outcomes while the strategy model has the unappealing behavior of a lottery the odds of which are fixed by the media's harsh judgments of who's winning and who's losing.

In coming to these conclusions, this paper illustrates a number of methodological points such as the usefulness of macromodels based upon assumptions about individual behavior, the analysis of macromodels using methods from electrical engineering, the strengths and limitations of analytical results versus simulations for understanding dynamic models, and the use of “ideal type” recognition and strategic voting models to clarify the systemic consequences of individual risk aversion and strategic voting. More generally, the paper shows how models can be used as narratives or parables for organizing disparate observations, refining our intuitions, and directing our research efforts.

This article presents a statistical model of the survey response that includes both a measure of the respondent's likely true attitude and a measure of the probability that the respondent is uncertain about her or his attitude and is “guessing” at a response. This latter possibility introduces systematic measurement error that may bias efforts to get unbiased estimates of attitudes. The model and the likely consequences of guessing are illustrated with questions about Vietnam policy taken from the 1968, 1970, and 1972 National Election Studies (NES) surveys. The article ends with a discussion of the limitations and fragility of the model.

Many recent empirical analyses of citizens' issue attitudes rely on the seven-point issue scales, which have been included in all of the biennial National Election Studies (NES) since 1968. The question format used to create these scales requires people to respond to two different issue statements simultaneously. While this approach has a number of practical advantages, it may also distort the measurement of issue attitudes in several ways. In order to examine this possibility, a 1990 South Carolina public-opinion survey had people respond to the separate issue statements. The empirical results lead to an optimistic assessment of the seven-point scales. People do seem to regard the paired issue statements in any of the seven-point questions as the opposite sides of a political controversy. This general conclusion has some important caveats: there are several distinct dimensions underlying citizens' issue judgments, and the degree of psychological distance between conflicting issue positions varies somewhat across issues. These findings have important implications for our understanding and measurement of citizens' issue attitudes.

Political scientists generally derive their quantitative methodologies from two disparate traditions: econometric or psychometric. The psychometric tradition has made a lasting impact on political analysis through attention to issues of measurement. Measurement issues are particularly troublesome to political scientists because many interesting concepts are either latent or multidimensional. On the other hand, the econometric tradition has provided political scientists with a means of coping with macrolevel phenomena. In studying macrophenomena, the problem of few cases (there is only one American economy, for example) is handled by using longitudinal analysis. Econometrics provided tools for coping with challenges that longitudinal analysis posed for regression analysis.

Using data from the 1956–60 and 1972–76 National Election Studies (NES), we replicate the issues/party identification (PID) models set forth in Franklin (1992) and Franklin and Jackson (1983). Mild constraints on the signs of the parameters in a full-information model reduce to insignificance the apparent causal influence of issue proximities on partisanship. Furthermore, alternative specifications of the issues/PID relationship yield no evidence that citizens update their partisanship based on their ideological distance from the parties.

Much research shows that individuals' economic perceptions strongly influence their political judgments, but some scholars suggest that typical survey designs may induce this influence due to the order and proximity of items in the surveys. This study explores the extent of such “contamination” of responses to both economic and political items in surveys, using an experimental design that varies survey item order across subjects. The results show that contamination is present in responses to both political and economic items, particularly when they are placed close together. Perhaps most importantly, the pervasive and deep contamination of responses to economic items suggests potentially serious implications for research on how economic perceptions influence political judgments. Ultimately, the findings are suggestive about the parameters of contamination and also the general nature (and theory) of the survey response.

We demonstrate that the expected value and variance commonly given for a well-known probability distribution are incorrect. We also provide corrected versions and report changes in a computer program to account for the known practical uses of this distribution.

The probability distribution in question, named the continuous parameter binomial (CPB) by King (1989a), has been known for at least six decades. The publications that reported the moments incorrectly, or were at least unclear about them, include a dated article (Guldberg 1931), a dissertation and several resulting published articles (Katz 1945, 1965), a popular reference book (Johnson and Kotz 1969),' work in political science on event count regression models (King 1989a), and extensions of these event count models in econometrics (Winkelmann and Zimmermann 1991). Event count regression models have become increasingly common in empirical political science research; some recent examples include Wang et al. (1993) and Krause (1994).