Parallel panel and fresh cross-section samples in recent National Election Study surveys provide valuable leverage for assessing the magnitude of biases in statistical analyses of survey data due to panel attrition and panel conditioning. My analyses employing a variety of typical regression models suggest that substantial panel biases are likely to be fairly rare in these data, even when panel and cross-section respondents have markedly different characteristics. However, two of the dependent variables considered here—campaign interest and turnout—do appear to be sufficiently sensitive to panel effects to warrant significant discounting or adjustment of panel data. I propose adjustments for panel effects in both cross-sectional and dynamic analyses, based upon variants of the “fractional pooling” (Bartels 1996) and “two-stage auxiliary instrumental variables” (Franklin 1990) methods.

The directional and proximity models offer dramatically different theories for how voters make decisions and fundamentally divergent views of the supposed microfoundations on which vast bodies of literature in theoretical rational choice and empirical political behavior have been built. We demonstrate here that the empirical tests in the large and growing body of literature on this subject amount to theoretical debates about which statistical assumption is right. The key statistical assumptions have not been empirically tested and, indeed, turn out to be effectively untestable with existing methods and data. Unfortunately, these assumptions are also crucial since changing them leads to different conclusions about voter decision processes.

This paper shows that agnostic spatial models that simultaneously attempt to estimate legislators' preferred points and ideological locations for the proposals on which they vote, such as the well-known NOMINATE model of Poole and Rosenthal, are not identified. The problem arises because the agnostic estimators inherit the granularity of the voting data and, so, cannot recapture the underlying continuous parameter space. I propose an alternative estimator that achieves identification by modeling the agenda.

Political knowledge has emerged as one of the central variables in political behavior research, with numerous scholars devoting considerable effort to explaining variance in citizens' levels of knowledge and to understanding the consequences of this variance for representation. Although such substantive matters continue to receive exhaustive study, questions of measurement also warrant attention. I demonstrate that conventional measures of political knowledge—constructed by summing a respondent's correct answers on a battery of factual items—are of uncertain validity. Rather than collapsing incorrect and “don't know” responses into a single absence-of-knowledge category, I introduce estimation procedures that allow these effects to vary. Grouped-data multinomial logistic regression results demonstrate that incorrect answers and don't knows perform dissimilarly, a finding that suggests deficiencies in the construct validity of conventional knowledge measures. The likely cause of the problem is traced to two sources: knowledge may not be discrete, meaning that a simple count of correct answers provides an imprecise measure; and, as demonstrated by the wealth of research conducted in the field of educational testing and psychology since the 1930s, measurement procedures used in political science potentially result in “knowledge” scales contaminated by systematic personality effects.

Many political science research articles that use limited dependent variable models report estimated quantities, in particular, fitted probabilities, predicted probabilities, and functions of such probabilities, without indicating that such estimates are subject to uncertainty. This practice, along with the reporting of “percentage correctly predicted,” can overstate the precision of reported results. In light of this, the present article describes a variety of measures of uncertainty that authors can include alongside estimates generated by limited dependent variable models. It also proposes an alternative to “percentage correctly predicted” and illustrates its calculations with congressional cosponsorship data from Krehbiel (1995).

Testing theories about political change requires analysts to make assumptions about the memory of their time series. Applied analyses are often based on inferences that time series are integrated and cointegrated. Typically analyses rest on Dickey—Fuller pretests for unit roots and a test for cointegration based on the Engle—Granger two-step method. We argue that this approach is not a good one and use Monte Carlo analysis to show that these tests can lead analysts to conclude falsely that the data are cointegrated (or nearly cointegrated) when the data are near-integrated and not cointegrating. Further, analysts are likely to conclude falsely that the relationship is not cointegrated when it is. We show how inferences are highly sensitive to sample size and the signal-to-noise ratio in the data. We suggest three things. First, analysts should use the single equation error correction test for cointegrating relationships; second, caution is in order in all cases where near-integration is a reasonable alternative to unit roots; and third, analysts should drop the language of cointegration in many cases and adopt single-equation error correction models when the theory of error correction is relevant.