Political Behavior Blog

« January 2006 | Main | March 2006 »

28 February 2006

More Journals?

Across the AAPOR discussion list comes an announcement of a new journal called Survey Research Methods. Published by the European Survey Research Association, the journal will deal with issues of "survey design, sample design, question and questionnaire design, data collection, nonresponse, data capture, data processing, coding and editing, measurement errors, imputation, weighting and survey data analysis methods." But aren't these matters already covered nicely by POQ, Political Analysis, Political Behavior, and other general journals such as AJPS? The only way it seems that the new SRM will differ is that it will be published on-line and can handle more interactive content. Personally, I'd rather that we put our scholarly efforts into elevating the journal Political Behavior to a more prominent place among journals. Let's hear your thoughts in the comment section.

Posted by Barry Burden at 10:05 AM | Comments (2)

PPBW: Hainmueller & Kern on Incumbency

In the next session of PPBW, Jens Hainmueller and Holger Lutz Kern will present "Party Incumbency as a Source of Contamination in Mixed Electoral Systems." The paper uses a cool regression discontinuity design to estimate the effect of candidate incumbency on party vote shares in the German mixed-member system. Although the effect turns out to be modest in size, it is a nice demonstration of a potentially quite powerful technique. Here's what the abstract has to say:

We identify party incumbency as a new source of contamination in mixed electoral systems. We argue that incumbency not only boosts incumbents' SMD vote shares; it also has a positive effect on vote shares in the PR tier of mixed electoral systems. This effect has been overlooked in the contamination literature so far. We test our hypothesis with data from German federal elections and employ a regression-discontinuity (RD) design in order to overcome some of the obstacles to reliable causal inference in observational studies. The RD design exploits the random variation in party incumbency status that occurs when a district race is close and thus allows for causal inference under a weaker set of assumptions than the regression models commonly used in the electoral systems literature. We find that party incumbency results in a gain of about 1 to 1.5 percentage points in PR vote share, which is sufficient to potentially trigger significant shifts in Bundestag majorities.

Posted by Barry Burden at 10:00 AM

27 February 2006

APRW: Burden & Hillygus on Biased Voter Learning

This week's American Politics Research Workshop (APRW) features a paper by me and Sunshine Hillygus titled "The Devil You Know: Voter Learning, Polarization, and George W. Bush's Reelection." Among other things, we show that no president should ever pledge to be "a uniter not a divider" as Bush did; public attitudes toward the president always polarize. Incumbency, polarization, and learning are natural consequences of a president's first term in office. The abstract says more:

Over the course of a president’s first term, the electorate acquires new information about the incumbent. This happens in two ways. First, among the uninformed, passive learning about the president and his policies typically improves his evaluations because it reduces uncertainty relative to the challenger. This contributes to a presidential incumbency advantage. Second, among those with some information and predispositions, active learning causes polarization of opinion. As a result, opinions toward the president almost always diverge during his first term in office, and this has been compounded by the secular rise in polarization in recent years. In light of this theory, George W. Bush’s reelection win was not that different than the reelection wins of other recent presidents. His presidency was unique, however, in that the learning process was accelerated by 9/11 and the Iraq war. Despite the initial unifying effect of 9/11 on opinions of Bush early in his presidency, evaluations of Bush polarized over these very issues of security and terrorism while contributing to his reelection.

Posted by Barry Burden at 2:10 PM

21 February 2006

Gerber on State Fiscal Policies and Local Government Inequality

Last week at the American Politics Research Workshop, Professor Elisabeth Gerber of the University of Michigan presented a paper entitled "The Impact of State Fiscal Policies on Local Government Inequality," co-authored with Michelle Woolery. Using data from a sample of Michigan cities, the authors find that limitations imposed by the state on localities' ability to tax exacerbate disparities among communities.

According to Professor Gerber, many states enacted tax and expenditure limits (TELs) as a result of the tax revolt of the late 1970s. Michigan, for example, passed the Headlee Amendment, which limits the amount of property tax that can be levied by localities. Further restrictions were imposed following the passage of a ballot proposal in 1994. Professor Gerber is interested in studying on how the impact of these limitations varies across communities, focusing especially on community capacity (such as wealth), demand factors (such as community demographics), and the local political system.

Professor Gerber found that wealthier counties are much less likely to be "constrained" by TELs, largely because they start with a greater tax base to begin with. Somewhat surprisingly, more racially diverse counties seem to be more constrained, despite the hypothesis that racial tensions might dampen demand for expenditures. Finally, local governments with partisan elections were relatively less constrained, presumably because partisan competition forces taxes to be low.

This area seems to be a promising one for future research. Given the variety of local institutional forms in the United States, one can imagine numerous studies examining the effect of the political system on public policy.

Posted by Ian Yohai at 12:09 AM

18 February 2006

Tolbert on Ballot Initiatives and the 2004 Elections

On Friday February 17th, the Political Psychology and Behavior Workshop sponsored Professor Caroline Tolbert’s talk, “Direct Democracy, Agendas, and Presidential Vote: Gay Marriage and the 2004 Election�. Prof. Tolbert, presenting her ongoing work with Professors Todd Donovan and Daniel Smith, offered compelling evidence that initiatives banning gay marriage on the ballot in thirteen states during the 2004 election had a strong influence on the voter’s calculus.

Using a unique data set, Prof. Tolbert argued that having a same-sex marriage ban on the ballot made voters who were concerned with gay marriage and a resident of a state with a ban of gay marriage on the ballot substantially more likely to vote for Bush than similar residents in states without these initiatives on the ballot. Prof. Tolbert’s analysis also suggested that these initiatives had substantial power in setting the agenda for an election. Counterintuitively, these effects are primarily concentrated among secular voters. Her paper and talk concluded with a conjecture that these ballot initiatives potentially represent a new electoral tool for campaigns and politicians.

The audience response to her presentation was largely positive. One issue addressed by several members of the audience, and a problem for political science more generally, is the selection of states used to test the effects of these ballot initiatives. Many studies attempt to exploit variation among the American states as an opportunity to discern how context influences individual level political behavior. Unfortunately, states are not randomly assigned to have ballot initiatives or other policies, and as a result it may be difficult to ascertain the extent to which the ballot initiatives were driving Prof. Tolbert’s results.

Posted by Justin Grimmer at 10:04 AM

16 February 2006

Sensibility Wars?: What is the white working class?

In a lively exchange, Larry Bartels and Thomas Frank have argued over the role of the “white working class� and that group's allegedly waning support for the Democratic Party. A few weeks back, we were lucky enough to hear Bartels present his paper “What’s the Matter with What’s the Matter with Kansas� to which Thomas Frank has written a rejoinder (Bartels has incorporated Frank's criticisms into his latest version of his paper).

One of disputes between the two gentlemen is the definition of the “white working class.� Bartels outlines some possible criteria: educational attainment, occupation, subjective identification, and income. As Bartels describes, each of these have their drawbacks. In an earlier draft of the paper he used income, but in response to Frank’s critique, uses educational attainment in the draft presented at CAPS. In an empirical analysis of Frank’s thesis, Bartels finds little support for the idea that working class whites have abandoned the Democrats.

This debate raises the question: What does it mean to be working class in America? Is working class a meaningful classification?

This question might go back to the oft cited David Brooks article “One Nation, Slightly Divisible� from the Atlantice Monthly published in December 2001. In it, Brooks argues that we are a nation divided by our different sensibilities and tastes; and like a high school cafeteria, we tend to form cliques or communities. Might “working class� be a construct based on perceptions of the sensibilities or tastes of one group (not necessarily class) by another? These sensibilities are reflected in the electoral map, but they are not predicted by variables we associate with working class. We need to explain where these sensibilities arise and how they translate into political behavior.

Maybe we should have new taste and sensibility measures in the National Election Survey. Heineken or Bud Light? James Taylor or Lynrd Skynrd? Culture war indeed.

Posted by Andrew Reeves at 10:17 AM | Comments (1)

13 February 2006

(More on) predicting election outcomes

I've been interested for a while in how to try and predict the likely outcomes of this fall's midterms, as well as how best to predict elections in general. The Swing State Project just posted some information about how well "political experts" (including campaign consultants, political scientists, pundits and the like) predicted the 1994 midterm elections.

The answer? Not well... Few of the 21 experts predicted the GOP's victory that year - and no-one came close to predicting the extent of their takeover. As DavidNYC puts it,

Bottom line: Beware predictions, especially from high-profile experts. That's not to say we shouldn't try to make any - hell, that's half the purpose of the political blogosphere. It just means we all have to maintain a strong skepticism - and keep that salt handy.

Fair point, although presumably the advice applies equally to low-profile, non-experts. Feel free to refer me back to this post the next time I make a prediction!

Posted by Phil Jones at 4:50 PM

Tolbert on Same-Sex Marriage Ballot Initiatives in 2004

This week's session of PPBW features a paper by Todd Donovan, Caroline Tolbert, and Daniel Smith titled "Direct Democracy, Agendas, and the Presidential Vote." The paper argues that same-sex marriage bans on state ballots influenced the outcome of the 2004 presidential election by affecting the issue agenda and priming voters to consider the candidates in terms of the views of gay marriage. This is a more specific -- and testable -- hypothesis than the general debate that has taken place about "moral values" in 2004.

This study draws from agenda-setting and issue priming theories to develop hypotheses about how state-level ballot propositions may affect presidential elections. We propose that ballot measure prime voters to evaluate candidates in terms of policy questions placed on their state ballot. We test if concerns about gay marriage in 2004 had greater salience for voters in states where proposals to ban gay marriage were on the ballot, and test the impact of gay marriage on presidential vote choice. We find that gay marriage was more likely to be cited as an issue use to evaluate candidates by some residents of states with marriage ban propositions. We also find respondents who were concerned about gay marriage who resided in states with the issue on the ballot were more likely to vote for Bush. The research is among the first to show that state ballot propositions may have subtle but important effect in presidential elections.

PS - Elisabeth Gerber is speaking Wednesday in the American politics research workshop on "The Impact of State Fiscal Policies on Local Government Inequality."

Posted by Barry Burden at 10:03 AM

9 February 2006

Predicting election outcomes

Predicting the future as well as explaining the past is often a major goal of political science. Every four years, we're treated to various statistical models going head-to-head to predict the presidential election outcome using various quantitative measures. The recent elections in Canada, however, saw a different kind of test, as different methods, not just different models, of predicting the outcome were put to the test.

Although several polling outfits put out predictions of the seats each party would win based on their final days of polling, of greater interest are two other, more unusual methods. First, the UBC Election Stock Market acts as a financial market, where traders buy and sell contracts representing the political parties competing in the election. The ultimate value of each party's stocks are then used to predict the number of seats each party is likely to win in the election.

Second, the Election Prediction Project solicits information and opinions from the public, which is considered by a panel of political pundits, before making a somewhat informed guess about the likely winner of each riding.

So how did they do?

The Election Stock Market helpfully shows their results compared to other methods. Polling firms differed largely in their results, with the four polls shown having an average error in party vote shares of around six percentage points. In contrast, the Stock Market incorrectly predicted just 15 of the 308 seats - an impressive record, especially given that they extrapolate from national party stocks to individual riding outcomes.

But what surprised me was that the Election Prediction project actually did better than either of these methods, incorrectly predicting only 12 of the district outcomes. For a supposedly less scientific way of predicting the future, the guesses of pundits and the public seemed to outdo the informed opinion of stock market traders.

Posted by Phil Jones at 10:22 PM | Comments (1)

5 February 2006

CAPS Seminar--Larry Bartels, "What's the Matter with What's the Matter with Kansas?"

On Friday, February 3, Larry Bartels presented his paper, "What's the Matter with What's the Matter with Kansas" for the Center for American Political Studies. Franks in the original work "What's the Matter with Kansas" argues that lower class voters in states like Kansas ignore their economic interests and support the Republican Party because they believe that Republicans are closer to them on important social and moral issues. Moreover, the Democratic Party fails to assert their advantage on economic issues to these working- and lower-class voters. Bartels presented data from the past 50 years of the American National Election Studies that shows that whites without a college degree are more likely to support the Republican Party. However, among whites without college degrees, the voters in the lower-third of the income distribution are more likely to be Democratic supporters, a finding that directly contradicts Franks's arguments. Bartels also finds that social or cultural issues such as abortion or women's rights play less of a role in determining vote choice than do beliefs about government spending.

The audience response was overwhelmingly positive. Most of the discussion centered around what Bartels's analysis meant for the prospects of the Democratic Party in the next election. However, the audience also pressed Bartels to discuss the implications of the analysis and some alternative explanations for recent Republican success.

Posted by Traci Burch at 10:17 AM

3 February 2006

McDermott on "Cognitive Neuroscience and Politics"

Today, the Political Psychology and Behavior Workshop sponsored a talk by Prof. Rose McDermott of UC-Santa Barbara. Prof. McDermott presented a chapter from a forthcoming edited volume in which she reviews some of the latest findings in the field of cognitive neuroscience, their application to the study of political behavior, and the methodological/theoretical challenges faced by this emerging strand of research. Such challenges include the recognition that "emotion" is a term that describes a variegated class of phenomena, the difficulty in administering suitable stimuli and in measuring emotional responses, and the complexity of the mechanisms by which emotions shape cognition. McDermott discusses some of the next steps that this literature should take to formulate a "comprehensive theory of emotion." In particular, she points to recent advances in evolutionary biology and psychology suggesting that "behavior can be understood as a function of genetics interacting with the environment." The chapter concludes with an exhortation to extend empirical testing beyond American voting behavior and to employ a more eclectic set of methodologies and modelling strategies drawn from a variety of research traditions.

Posted by Federico Ferrara at 2:05 PM