CSSS Student Seminar
Wednesday, May 21, 4:00–6:30pm
Denny 401

Followed by CSSS Pub Night
6:45 pm @ College Inn
4000 University Way NE

The CSSS student seminar is an informal meeting open to all UW graduate students with interests in interdisciplinary quantitative social science research. The seminar aims to create a friendly supportive environment for students of all disciplines and all levels of technical sophistication to learn from each other. It serves this function in two ways:

(1) By providing a venue for students to practice giving a 'conference style' presentation, present their research at any stage, and receive feedback from fellow graduate students.

(2) By fostering a cross-disciplinary community of graduate students with interests in quantitative research. Fellow graduate students are a great resource for learning about new methods and statistical techniques, and for discussing the 'every day' problems one confronts in research. C6S is a forum for dialogue with and feedback from students in other disciplines. It can be very useful to talk with researchers studying similar questions through other disciplinary lenses, or to see familiar methods used in different disciplinary contexts.

The student presenter is encouraged to talk about their research for ~20 minutes, leaving 10 minutes for questions, feedback, and discussion.

To get a wide range of speakers and participants from different departments, students of different departments act as coordinators. Currently the coordinators are:

We encourage involvement of all departments with interests in research that deals with statistical issues in the social sciences. If you are interested in being a coordinator for your department, send an email to kornrich@u.washington.edu.

If you are interested in presenting at the seminar, read the guidelines and send an email with a short abstract of your talk to the coordinator of your department or to kornrich@u.washington.edu.

Archive of seminars for past quarters.

Abstracts:

1. Loren Collingwood, Political Science
Levels of Political Knowledge and Support for Direct Democracy: A Survey Experiment

Researchers are divided over the normative and political benefits and implications of the surge in American direct democracy over the past century. On the one hand, scholars have argued that voters are not up to making relatively complex voting decisions that direct legislation often requires. On the other hand, a different set of scholars maintain that through the use of voting cues and heuristics, voters are able to vote as though they were fully informed citizens. While elite opinion is essentially divided, voter opinion is certainly not. Public opinion polls suggest that voters overwhelming support direct democracy. Recent research has shown that perhaps the public’s support for direct legislation may not be as deep as surveys suggest. With this in mind, I conducted a split sample survey experiment involving 1) a treatment condition whereby voters were asked how they would vote on a series of upcoming ballot initiatives and then asked their opinion of the ballot initiative process, and 2) a control condition in which voters were asked about the ballot initiative process followed by the series of the same vote questions asked of respondents in the treatment group. A chi-square analysis shows that while there is no difference in support for the ballot initiative process at the treatment group level, respondents with low levels of political knowledge in the treatment condition are statistically significantly different from their counterparts in the control condition. Thus, the paper tentatively concludes that when voters are primed to think about how they will vote on a series of ballot initiatives, their support for direct democracy is moderated by levels of political knowledge.

2. Jennifer Chunn, Statistics
Projecting life expectancy using a Bayesian Hierarchical Model

Projecting life expectancy using a Bayesian Hierarchical Model Every two years, the United Nations Population Division publishes life expectancy and 50 year projections for over 200 countries. However, these projections are entirely deterministic, that is, no level of uncertainty is given with the projections. Current UN methodology uses parametric models of the rate of change in life expectancy. We propose a Bayesian hierarchical model as a natural extension of the current UN methodology while giving probabilistic projections of life expectancy.

3. Xiaoyue Niu, Statistics
A Hidden Markov Model with Latent Eigenfactors for Dynamic Social Networks

The phrase “social network data” generally refers to information on the presence or absence of links between a collection of nodes. The information is often expressed in terms of adjacency matrices. In the talk, we will use the reduced rank approximation to estimate the mean matrices. And we will introduce a hidden Markov model to pool the information across different slices of the network. We will apply the model to an international conflict and cooperation example. Finally, we will mention some more general models and address some computational issues.

4. Sabino Kornrich, Sociology
The Gendered Provision of Income & Spending on Housekeeping Services

The continuing increase in women's labor force participation and the concomitant rise in dual-earner families have meant that more families are subject to work-family conflict and the challenge of fulfilling family obligations with decreasing time. One way that families may deal with these challenges is by outsourcing household labor. Yet the ways in which families outsource household labor are likely linked to gendered notions of the appropriateness of certain types of work. In this talk, I discuss changes in spending on housekeeping over time, the link between women's and men's income provision and housekeeping, and changes in this gendered effect on outsourcing over time.