Arguably, the first important political event in Oregon in 2010 will be the January 26 election result.
Earlier, I made several posts about these survey results:
Top line results are at: http://lindholmcompanyblog.com/?p=1646
Measure 66 demographics are at: http://lindholmcompanyblog.com/?p=1652
Measure 67 demographicsare at: http://lindholmcompanyblog.com/?p=1655
So, what do the survey numbers I posted earlier imply about how the election will turn out?
This post is going to look at what two different social science models of voter decision making imply about where the election is.
It’s well recognized that surveys are only snapshots in time. However, polling from previous elections can give us some idea where the campaigns for measures 66 and 67 will be heading. The idea is to look at what information the voters have when they answer the questions. To examine two possible patterns, I’ll look at public polling on two recent Oregon statewide money measure elections.
The first model describes the polling for November 2007’s Measure 50, a proposed tobacco tax. Tracking surveys are at: http://lindholmcompanyblog.com/?p=1590. The upper graph has the actual survey numbers and the lower graph shows the voters after assignment. The first survey showed the “yes” side well ahead. Subsequent surveys were very good predictors once the undecided voters were allocated to “yes” or “no” in proportion to the already decided “yes” and “no” voters, respectively. Using this model would result in an exactly correct forecast: Yes: 40% and No 60%.
Note that this model also implicitly assumes that the undecided voters are, on average, have attitudes just like already decided voters.
A second model describes the polling for January 2003’s Measure 28, a proposed temporary income tax increase. The public survey, conducted by the University of Oregon’s Survey Research Laboratory, as published in the Register Guard newspaper (Jan. 20, 2003) reported Measure 28 ahead by 46% to 35% with a margin of error of plus or minus 5%. In this case, the undecided voters all should be allocated to the “no” side. Using this model would result in an exactly correct forecast: Yes: 46% and No 54%.
Note that this model implicitly assumes that undecided voters are fundamentally different from decided voters. Undecided voters are “no” voters.
Most ballot measures fit one of these two models. The question is: Which model is more like measures 66 and 67?
If it’s the model of November 2007’s Measure 50, then Measure 66 effectively leads 53% yes to 47% no and Measure 67 effectively leads 58% yes to 42% no. Both measures would be winning!
If it’s the model of January 2003’s Measure 28, then Measure 66 effectively trails 40% yes to 60% no and Measure 67 effectively trails 46% yes to 54% no. Both measures would be losing!
Subsequent research should provide a clearer picture of which model was correct.