2009-2010 Fall Colloquia

ABSTRACT: Normative democratic theory praises the informed participant. Knowing about the election at hand should presumably help individuals make the best possible decision about who to vote for. Indeed, proponents of rational choice models for voter turnout have emphasized the challenge of information gathering as a costly impediment to participation. We explore the role of campaign-relevant information in large national election studies of the United States in 2004 and New Zealand in 1996 to find that information has at most a minimal influence on turnout. Further, we conducted an experiment during the 2006 Florida Primary election where we provided information about candidate issue positions in a low-information environment. Individuals exposed to issue-position information were more knowledgeable following the election, but were not more likely to have voted. The results suggest that models of electoral participation that rely on campaign information, including most current rational choice theories of participation, need to be significantly revised.

ABSTRACT: Volunteer internet panels can provide quick data at a low cost, making them more convenient for researchers in marketing and the social sciences than the slower and costlier random digit dialing (RDD) full-probability samples. Further, tests have shown that data collected over the internet vs. the phone can reduce some response biases that affect respondents who rarely take surveys or who have low cognitive skills. But can opt-in internet panel data be trusted?
Using data from identical surveys administered by 9 different vendors (2 probability, 7 internet) to 10,000 american adults, this talk addresses the validity of probability and non-probability data for three different types of research questions. First, it addresses whether the probability and non-probability samples differ on demographic benchmarks and on other attitude and behavior questions. Non-probability samples differ significantly more than the probability samples from benchmarks. Second, we address whether non-probability samples produce similar correlations, and find that about half of the time the internet panels estimate vastly different regression coefficients than the RDD samples. Third, it addresses whether effects of experimental manipulations differ across the samples, and finds that some experimental effects replicate, but others don’t.

ABSTRACT: Both academic and journalist discussions of Congressional polarization make the assumption that polarization has increased at the expense of bipartisanship. Using standard roll call analysis, this assumption is borne out; there is a one-to-one mapping between increased polarization and decreased bipartisanship. However, these findings are misleading for two reasons. First, the bills that face roll call votes are not a random sample of legislation. Second, the likelihood that bipartisanship legislation faces a roll call vote has changed over time. When an original measure of bipartisanship, bill cosponsorship, is used I find a much weaker relationship between polarization and bipartisanship. Consistent with electorally driven legislative behavior, I find that bipartisanship has persisted despite increases in polarization, particularly in policy areas that do not define the party brand name.

ABSTRACT: When measuring attitudes with questions that offer dichotomous, mutually exclusive response options, researchers sometimes ask “some/other” questions (which present the response options using stems like “some people think that … but other people believe that…”), instead of asking simpler “direct” questions (which do not couch the options in terms of others’ opinions). Two studies reported here investigated whether the some/other form yields more valid attitude reports, as some researchers have claimed, and the posited psychological mechanism of this difference. The first experiment involved face-to-face and telephone interviewing of a representative national sample of American adults (N= 1,509) and showed that the some/other question form yielded responses with lower concurrent validity. The second experiment entailed data collection via the Internet from a national panel of American adults (N= 1,002). It showed that the some/other form reduced validity, reduced respondents’ certainty about their opinions, and led respondents to think that the population’s beliefs are more evenly split. These two studies suggest that in order to maximize data quality and minimize manipulation of respondents, the direct question form is preferable.

ABSTRACT: The idealized version of democracy posits that citizens vote for candidates whose issue positions are close to their own. In this manner, citizens can improve the chances that government outcomes are to their liking. However, social psychological theories and political science theories alike contend that voters do not judge candidates on all issues equally. Rather, they weigh their overall evaluations of candidates by the personal importance that they attach to particular issues. Using data from the American National Election Studies from 1968 to 2004, this research explores the moderating effect of issue importance on issue agreement in forming overall evaluations of political candidates.

ABSTRACT: In recent years, televised political satire has gained new salience in American popular culture – riding on success of shows like ‘The Daily Show with Jon Stewart’. Using data from a web based survey experiment, we explored the impact of this newly salient form of news delivery on viewers’ political attitudes and behavioral intentions. Respondents watched either a news story from “The Daily Show”, a comparable story from CNN, or no story, and then answered questions. In the full sample, exposure to the Daily Show clip seemed to have little impact on people’s thinking. But when we explored the impact of the manipulations separately among people who have chosen to watch the Daily Show and those who have not, we found striking and interesting effects, often in opposite directions. Thus, to understand the impact that the news media have on readers and viewers requires attention to the self-selection process that determines exposure to the stimulus outside the research setting.

ABSTRACT: Data from the 2008 Exit Polls were used to demonstrate that voter turnout among young people rose slightly from 2004, but that young people were still sorely underrepresented. Indeed, the reports suggested that youth comprised only 18% of the electorate while accounting for over 21% of the voting eligible population. While this fit historical trends in electoral participation, it appeared to belie the popular notion that young people were enthusiastic in anticipation of the election. Yet the exit-polling sampling method is inherently limited by its inability to capture early and absentee voting. While the national exit polls conducted by Edison-Mitofsky use alternative methods to account for these differences, the biases introduced may be rather large. By comparing states with early voting and no-excuse absentee voting to those that only allow absentee voting with an excuse, I find that the supplemental data was highly skewed, leading to inconsistent conclusions about youth turnout. Accounting for this bias, youth voting numbers appear to reach population proportions. A number of alternative explanations could also address these findings. I explore the possibilities that young interviewers may be selecting young interviewees for the exit poll, that overall voting may be higher in early vote states thereby diminishing the youth portion of the electorate, and that differences between the states may explain this discrepancy (e.g. swing state, southern, etc.). Election results, 2004 data, and data from the current population survey are used to disentangle these possibilities. We find that single-mode studies do not vary meaningfully across states, suggesting that the problem lies with the exit poll. Indeed, a similar trend is observed in data from 2004.

ABSTRACT: We conducted an experiment testing the impact of the content of communications on the propensity for aggression in a simulated crisis game. In addition, we investigated the relationship between previous military spending and the outbreak of hostility. Finally, we examined whether the structure of the incentives might alter a player’s strategy. Our sample included male and female subjects who participated in an experimental crisis simulation game in one of three types of dyads: male-male,
female-female, or male-female. In six rounds of play, subjects made procurement decisions, took an action in response to the conflict, and wrote messages to their adversary. The study found significant effects for condition, lagged declared military spending and lagged communication.