2009-2010 Winter Colloquia

ABSTRACT: To many commentators and social scientists, Americans’ stances on political issues are to an important extent driven by an underlying conservative-liberal ideological dimension. Self-identification as conservative vs. liberal is treated theoretically as a marker of this dimension. This research evaluates the hypothesis that conservative-liberal self-label partly represents a tendency to follow political cues when forming stances on newly politicized issues. In Study 1, conservative-liberal self-label measured in 2000 had an independent prospective effect on support for invading Iraq in 2002 and support for the Iraq War in 2004, controlling for substantive ideology, party identification, and demographics. In Study 2, conservative- and liberal-identifiers adopted stances on farm subsidy policy, and internalized beliefs in support of those stances, based on randomly varied messages about which ideological group supports which stance. Discussion focuses on the dynamic relations among political discourse, ideological and partisan identity, and substantive political attitudes over time.

ABSTRACT: While previous studies generally presume that voters engage in retrospective voting to evaluate an incumbent’s performance, another possibility is that voters are simply inclined to support the status quo when they feel personally satisfied or happy. We leverage a set of natural experiments to explore the extent to which personal happiness affects voting behavior. To do this, we examine events that government has nothing to do with and for which no government response would be expected: the outcome for the local college football team on the weekend before Election Day. We collected data on pre-election football scores between 1964 and the present, as well as local election returns for presidential, gubernatorial, and senatorial elections. We find that a win for the local college football team in the weekend before the election can lead the incumbent to receive more than three additional percentage points in the election. We also have some preliminary evidence that the incumbent party does better in gubernatorial elections when there is less cloud cover on Election Day, a phenomenon that affects voter happiness but to which no government would be expected to respond. Together, our results show that voters’ choices depend significantly on events that affect their personal level of happiness even when those events are unrelated to government activity.

ABSTRACT: Non-probability Internet samples have emerged as a widespread, rapid, and inexpensive way to collect public opinion data. Yet unlike traditional survey methods, the individuals recruited for Internet surveys are not representative of the public. Proponents of Internet samples have suggested that statistical corrections can be employed to mitigate the challenges posed by non-probability samples and thus provide leverage for accurate inferences. In this study, we compared population marginals and trends over time in one non-probability Internet survey and one RDD telephone sample. Looking at 17 variables measured over 17 weeks in both datasets, we found that demographic weighting did not consistently eliminate the differences between the two surveys for either type of inference. The inconsistent pattern of results alludes to the problems researchers experienced with quota sampling in the 1930’s and 1940’s. We discuss the implications of these inconsistent results and what they portend for the future of Internet-based sampling.

ABSTRACT: Why are some voters more likely to choose less qualified male candidates over more qualified female candidates, when virtually no voters will openly admit to sexist attitudes or behaviors? For the first time, we embed innovative but controversial “implicit” and “symbolic” sexism measures within a behavioral vote choice experiment. These measures purport to capture uncontrollable bias (in the case of implicit measures) or unconscious bias (in the case of symbolic measurements), but their predictive validity for important political behaviors like vote choice has yet to be established. First, we find evidence that implicit and symbolic sexism does indeed predict sexist voting suggesting that the electoral process may not be completely gender-neutral. After controlling for the degree to which candidates differ in qualifications, we find that voters who are more biased against women, as shown by both types of bias measures, are less likely to vote for female candidates. Gender expectations and stereotypes exist, are measurable, and can affect the evaluations of women candidates. Second, we find that as the difference between the quality of a female and male candidate increases, the effect of gender bias, as measured by the “Implicit Association Test,” decreases. In other words, there is suggestive evidence that one’s unconscious preference against female candidates can be attenuated by added qualification. Finally, we find preliminary evidence that voters have more favorable evaluations of women candidates who run in campaigns that highlight “female” issues and much lower evaluations of women who run in more “male” issue environments, showing that bias and attenuation responds differently depending on election environment. Our findings have implications for assessing how voters respond to the qualifications of women candidates, and conclude that at least with respect to gender biases, implicit and symbolic measures can predict sexist voting.