Not since 1952 has the US encountered a general election campaign involving two candidates that have not yet sat in the executive office. Coupled with a general sentiment for change, low presidential approval ratings and racing primary dates, an unusually large amount of attention has fallen directly on the primary campaigns of 2008. The media attention and disconnect between the White House and the candidates provide an extraordinary opportunity to investigate the relationship between campaigns and the media in the primary setting.
Beginning in the summer of 2007 we employ daily automated content coding of candidate and newspaper websites, coupled with a collection of statewide polls, to measure and test the dynamic mechanisms of presidential primary campaigns.
Who won the invisible primary and does winning the invisible primary predict primary success?
How much do candidates' local, as opposed to national, media profiles relate to primary success?
Does frontloading increase or decrease the importance of Iowa and New Hampshire momentum?
How do campaigns react to the agendas and rhetoric of the opposition and the news media?
How does press attention and coverage differ by a candidate's poll standings?
How persistant is primary momentum and what explains its persistance?
Few studies endeavor to grasp the role of the traditional and informationabundant medium of newspapers; many of those that have tried to do so have relied on a large and often expensive group of coders (e.g Beck et al. 2002). Modern computing capabilities and the duplication of newspaper articles on websites present an easier alternative. Beginning in July of 2007 we employed daily automated content coding of newspaper websites. We have collected national and local newspaper articles from the web using an automated retrieval program.
Downloaded information includes the title, the journal, the time, and the content of the article, as well as the time of its posting. The program was written to retrieve information from a nonrandom sample of newspaper websites; collecting articles from all news sites across the country would indeed be worthwhile, but beyond realizable computational space. Given the necessity of a sample, we set out to balance the data collection across a few potential confounders, including timing of the primary, population size of the state, newspaper syndication size, and even the ideological tendencies of the newspapers. Here we focus on the early state nomination contests of Iowa and New Hampshire in an attempt to see what relationship the two contests have with the national Invisible Primary. This subset of data includes over 4,800 articles spanning 184 days.
We analyze both local and national news media attention since past research has suggested that the two kinds of media cover campaigns in dierent manners (Buell 1987, Just et al. 1996, Shaw & Sparrow 1999, Flowers, Haynes & Crespin 2003). In their analysis of state and national news coverage of the 1992 primary, Haynes & Murray (1998) found candidate spending and visits were much more in uential within state coverage than within national news coverage. They also argue that the national news media's attention to a candidate has a positive in uence among local news coverage. We focus on whether there were noticeable dierences between local and national news media attention levels toward candidates and whether the two sources have a dierent level of in uence on local candidate support. For example, if voters in Iowa are more responsive to local than national news coverage, then candidate activities and spending that in uence local media coverage might be a more successful way for candidates to gain familiarity and viability leverage. However if national media exposure bestows an in- uence both on local news coverage and voter opinions, then perhaps there are extensive gains to be had in winning the Invisible Primary within the national media.
For more information, see our papers in the Data Analysis section