Issues, Media and Momentum in the 2008 Primary: Automated Retrieval of Dynamic Content Dino P. Christenson Department of Political Science R ESEARCH Q UESTIONS IA NH 5 4 3 1 0 125 100 50 75 50 40 30 Jul Sep Oct Nov Dec Jan Feb 25 Traditional Republican Issue Dynamics IA NH NV SC Super 0 Education Poverty Labor Environment Military Economy Spending Taxes Military Economy Morals Spending Taxes Morals Dynamic process I Daily censoring at zero I Sampling error from large daily sample sizes I Large volatility in issue mentions I Days of week and holiday media reporting I FEC reporting deadlines 3 M OTIVATING K ALMAN F ILTER 4 5 Healthcare 0 L ARGER P ROJECT Aug 1 I Invisible Early Daily Issue Mentions (filtered) I Traditional Republican Issues 20 I Traditional Democrat Issues 10 I R AW I SSUES D ATA Average Daily Issue Sentences I Explore issue salience in the 2008 presidential primary campaign Expose dynamic nature of the media’s issue coverage as it pertains to both the invisible primary period and the early primary Generate smooth summary measures of the dynamics with Bayesian state space models Find that particular issues dominate the campaigns at the start and finish, but that there are also changes in the relative salience of issues throughout the campaign Showcasing data collection and potential use Just the first few of several steps toward understanding the dynamic relationship of media issue coverage to indicators of candidate status in the 2008 primary 0 I Super 2 S UMMARY NV SC Education Environment Healthcare Labor Poverty 2 I Traditional Democrat Issue Dynamics How does media coverage of issues change over the primary campaign? To what extent is the coverage related to key campaign events? Daily Issue Mentions (filtered) I F ILTERED I SSUES D ATA W ORD C LOUD B AYESIAN S TATE S PACE M ODEL Sample national and local newspapers Test the dynamic mechanisms of primaries Evaluate roles of media, polls, contributions and candidate visits Project with Corwin D. Smidt: http://VisiblePrimary.com Democratic Issues Dynamics Super IA NH NV SC Super ● ● ● ● ● 4 ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 2 ● ● ● ● ● ● ● ●● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● I ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 3 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● I ● ● ● ● ● ● ● ● ● ● ● ● 3.5 ● 1 ● ● ● ● ● 3.0 ● 0 Daily Issue Mentions (logged) ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● I ● ● ● ● I Dec Jan Feb ● Jul Aug I ● Sep Oct Nov Both horse-race & issue coverage increase over the 2008 Primary Defense dominates traditional Rep issues Healthcare strongest among Dem issues Economy important but not the leading concern Moral values & spending with relatively little attention N EXT S TEPS ● ● I I 5 NV SC ● ● I Nov Republican Issues Dynamics IA NH ● Counting number of sentences per day with a salient issue in national newspapers Timeline: July 9, 2007 to February 5, 2008 4,184 articles mentioning an issue 151,702 issue sentences I A GGREGATE F ILTERED S ERIES D YNAMIC I SSUE C ONTENT I I 6.0 I I Bayesian State Space Model: I Forward filtering backward sampling algorithm I Inverse-Wishart priors for residuals I Normal diffuse priors for regressors 5.5 I I θt = θt−1 + υt where υ˜N(0, σ2υ ) 5.0 I Oct Transition equation: 4.5 I Sep C ONCLUSION Issuest = θt + Dayt + FECt + ϵt where ϵ˜N(0, σ2t ) 4.0 I Aug State equation: Original longitudinal dataset culled from newspaper articles Use Perl to automatically retrieve and code articles Daily Issue Mentions (logged) I Jul Dec Jan Feb Jul Aug Sep Oct Nov Dec Jan Feb I Web: http://www.polisci.osu.edu/grads/dchristenson/ Further filter the series by taking into account candidate visits & other cycles Extend timeline to include period leading up to conventions Explicit hypothesis tests of role of issue coverage (dynamic ownership & status) Address theories of bottom-up vs top-down issue salience in campaigns Multivariate perspectives: BVAR Email: christenson.24@polisci.osu.edu