1

Poster

Comparing Opinions and Preferences across States and Regions: The Fallacy of using Ideological Responses Cormack, Lindsey Nagler, Jonathan Uploaded 07-20-2012 Keywords ideology polarization public opinion Abstract We are interested in differences in ideology and preferences on policies across red and blue states, and across people who say they are liberals versus conservatives. We make several points about measurement of ideology and issue preferences, all in the context of `polarization'. First, the use of ideology for measuring polarization is quite dangerous as the typical ideology question has no fixed scale -- allowing respondents to interpret it quite differently across regions or groups. Second, ideology also has a potential dimensionality problem: it is fundamentally a projection of many dimensions (or issues) onto one dimension, thus allowing respondents to weight lower level dimensions differently across regions or groups. Taken together, this suggests that an electorate may be polarized on some issues, but not on other issues. This could be because the issues exist on distinct dimensions. Or, we could find issues that lie on the same dimensions, but some are simply more discriminating than others. In such cases, `polarization' would exist on the more discriminating issue, but not on the less-discriminating issue. Thus polarization, in the absence of a clear definition, will likely to continue to exist in the eye of the beholder.

2

Poster

The Past is Ever-Present: The Dynamic Nature of Intrastate Conflict Jones, Benjamin Uploaded 07-28-2012 Keywords Event history Survival models Multi-state models Civil war Path dependence Abstract Civil wars pose a grave challenge to international stability as they tend to recur frequently over time. Nevertheless, existing theory treats civil wars as independent events. I reconceptualize civil war as a dynamic process, which creates a new statistical challenge – modeling multi-stage processes through a series of transitions within a longitudinal process. To overcome this problem, I introduce a multi-state event history model, which models the entire civil war process as a series of successive stages in which previous outcomes shape subsequent events, and apply it to a dataset of all civil wars from 1950-2004. The results provide strong evidence that previous outcomes exert both a direct, and indirect effect on subsequent transitions, revealing the conditional nature of factors frequently associated with war and peace.