2006 Election: What Will The Undecided Voters Do

de.licio.us ma.gnolia.com newsvine furl google yahoo netscape

Just 13 days 'til the election. Let me start with a forecast updating projections from here and here. Averaging together the tracking from electoral-vote.com and cqpolitics.com congress should look like this after the election:

House: D-224, R-211
Senate: R-50, D-50

One curiosity about the polling bothers me, however; and that is the significant undecided percentages. Some of the closest races, as tracked by electoral-vote.com, have undecided voters in excess of the margin associated with them:

Race (ST-Dem-Rep) Polling Margin Undecided
AZ-Pederson-Kyl 42% 49% 7% 9%
MD-Cardin-Steele 49% 44% 5% 7%
MO-McCaskill-Talent 45% 46% 1% 9%
MT-Tester-Burns 47% 41% 6% 12%
NJ-Menendez-Kean 46% 41% 5% 13%
OH-Brown-DeWine 48% 40% 8% 12%
TN-Ford-Corker 44% 47% 3% 9%
VA-Webb-Allen 45% 46% 1% 9%

If we apply the same distribution of undecided voters to these races that occurred in the 2004 presidential race in the same state (comparing average polls with final results) this is what we get for the above races:

Race (ST-Dem-Rep) Polling Margin
AZ-Pederson-Kyl 51% 49% 2%
MD-Cardin-Steele 52% 48% 4%
MO-McCaskill-Talent 45% 55% 10%
MT-Tester-Burns 53% 47% 6%
NJ-Menendez-Kean 51% 49% 2%
OH-Brown-DeWine 53% 47% 6%
TN-Ford-Corker 53% 47% 6%
VA-Webb-Allen 45% 54% 9%

In some states in the 2004 race, all the undecideds, plus some Kerry votes, went for Bush and vice-versa. Where that happened, I applied all the undecideds to the appropriate candidate, thus the wild swings compared to the polls. I do fully expect to see this on election day for any number of the following reasons:

  • I suspect that a large percentage of the undecideds in states that went for Bush on 2004 are republicans who are dissatisfied with the administration but not enough so that they will vote for a democrat.
  • I believe some of these surveys could be wildly inaccurate due to the exclusion of some demographics (e.g., cell-phone only households). Pollsters use mathematical weighting of the sampling they get to accommodate these issues, but I believe technology is changing faster than the pollsters can adjust.
  • The possibility of further election-day mishaps like the electronic machines in VA that will leave off party affiliations and some last names during all or part of the voting process due to a font-size issue

If these adjustments prove correct, and the more certain forecasts for the other races hold true, the senate should actually be:

Senate: D-51, R-49

If there is interest, I will look at the close house races the same way. Add a comment below if you're interested.


Related Stories:

Read More About:   Elections | Politics | United States