This post takes another angle at the tracking surveys from the November 2007 election. It looks at the the interaction of the demographic variables and the dates of the tracking surveys.
The responses from the three tracking surveys were pooled and then segmenting using a standard CHAID analysis. This leads to a tree diagram.
Usually CHAID is used to analyze the demographics of a single survey, a single snapshot in time. Earlier posts on this blog have done that. This analysis pools all three tracking surveys and uses survey date as another demographic variable. This allows one to see how key demographic groups responded to the campaign over time.
This analysis shows that Republicans, Other Registrants (NAVs and 3rd parties), and Democrats with less than a college degree all shifted significantly against Measure 50 between the September 10-12 survey (“9″) and the October 15-16 (“101″) survey. These groups remained opposing the measure throughout the rest of the campaign.
Only Democrats with a college degree or graduate school did not move strongly against the tax.
The three tracking surveys are designated 9 (Sept, 10-12 survey), 101 (Oct. 15-16 survey), and 102 (Oct. 29-30 survey).
QUESTION WORDING: “Now, I would like to ask you about the ballot measures on the November election. If the election were held today, would you vote yes or no on Measure 50: Amends constitution: dedicates funds to provide health care for children, fund tobacco prevention, through increased tobacco tax?”
IF YES/NO: Is that strongly or somewhat? IF DON’T KNOW: Which way do you lean?Response categories:
1. Strongly Yes
2. Somewhat Yes
3. Lean Yes
4. Don’t Know
5. Lean No
6. Somewhat No
7. Strongly No
METHODOLOGY: All three surveys consisted of 400 live telephone interviews of likely Special Election voters. The margin of error is five percent at the overall sample median. Likelihood was determined using a combination of modeling and screening potential respondents. Dates are as indicated in the graphs.