WikiSummary, the Social Science Summary Database

 

Johnston, Hagen, and Jamieson. 2001. Dynamics of the 2000 presidential campaign: Evidence from the Annenberg Survey.. APSA conference.

Overview

The Annenberg studies of the 2000 presidential campaign show that our current theories of campaigns are somewhat lacking. Largely an exploratory paper with some preliminary conclusions.

Data

The Annenberg survey took daily random samples from July through November. Smoothing the data (Fig 1) make it possible to investigate the moments in the campaign where things changed. Obviously, there were daily spikes and drops in Gore's support, but, when smoothed, 7 shifts appear to have been permanent (expressed in terms of Gore's support):

1. The drop after Bush's acceptance speech,
2. The surge before the Dem convention,
3. The surge after Gore's acceptance speech,
4. Gore's mid-September surge,
5. Gore's late-September (small) drop,
6. Gore's surge after the last debate,
7. and Gore's (small) drop about a week later.

Theoretical Explanations

The authors use advanced stats to differentiate variables with a time-series (longitudinal) effect from those with only a cross-sectional effect.

• Clinton: Clinton's popularity appears to have had a cross-sectional effect on Gore.
• The macroeconomy: It appears to have had a mild longitudinal effect.
• Character: Drops in Gore's perceived virtue (and rises in Bush's) appear to have had strong longitudinal effects.
• Other: Party ID and ideology matter too.

Speculation

The authors speculate that, of the 7 shifts in Gore's support, shifts 4, 6, and 7 suggest that the media matter. Something in the media or in political ads must have triggered that shift, perhaps by first triggering the changes in Gore's (and Bush's) perceived character qualities.

Comments and Criticisms

• This is aggregate data, so it can't tell us who is changing their mind and why.
• Y is always support for Gore; this assumes that support for Gore is 100 - support for Bush. But what about the undecided vote? See Hillygus and Jackman (2003) fig 2: the proportion of undecideds moves up and down during the campaign. Would that change this study's results?