Update on the Colorado senate race.

Update: 10/11/2004

Our new update for the Coors vs. Salazar Senate races comes from the three most recent polls. The Gallup Poll from 10/03-10/06 Gives Salazar a 11 point edge, Salazar 54, Coors 43. Mason/Dixon had a poll from 10/04-10/06. It gave Salazar a 2 point lead, Salazar 46, Coors 44. The most recent poll shows a dead heat. SUSA's poll from 10/05-10/07 had the race at a dead heat. Coors 48, Salazar 48.

As such our new prediction is Coors (((43+44+48)/3)+4)= 49; Salazar ((54+46+48)/3)= 49.33.

Ken Salazar is now our predicted winner. Our old figures were Coors 53.2%, Salazar 46.6% on 9/28/2004 and Coors 50.33%, Salazar 48.66% on 10/08/2004.

Original Post: 9/28/2004
Our most recent projection for the Coors vs. Salazar senate race is Coors 52.3% and Salazar 46.6%.

We looked at several factors in constructing this projection.

Firstly, and surprisingly, there seems to be no strong correlation between the funding advantage that one senate candidate has and the percentage of the vote that that candidate received. For example, in 2002 the Republican candidate got 50.5% of the total campaign donations and 51% of the vote, the Democratic senate candidate got 49.5% of the total campaign donations and received 46% of the vote. In 1998, a similar correlation existed, with the GOP candidate getting 62.5% of the financing pie and 62% of the vote and the Democratic candidate getting 37.5% of the financing pie and 35% of the vote. However in 1996 and 1992 the correlation was reversed. The GOP candidate in 1996 got 43.5% of the total campaign donations and received 51% of the vote, while the Democrat got 56.5% of the campaign donations and got 46% of the vote. In 1992 the Republican got 58% of the financing pie and 43% of the vote, while the Democrat got 42% of the funds and 52% of the vote.

The most recent fundraising numbers that I have seen show that Salazar has out fundraised Coors by a margin of 6-4. If that margin reaches 7-3, then I will take the fundraising advantage into account in the equation, until then, I have to stick to the step wise regression that says that it is not statistically significant.

Throwing out the fundraising numbers, I’m left only with past polls and final election results (registration numbers also are not statistically significant, but we are looking into a “strength of primary win” variable). No Democrat has ever won with over 52% of the vote in a Colorado senate election. Ben Campbell won with that amount in 1992. A Democrat has won a senate election in Colorado two other times since 1978. In 1980 with 50.3 of the vote and in 1986 with 50%.

I also looked at the periodicity of the vote. Maybe Dems scored better during Presidential election years (the big turnout argument). Since 1978 there have been nine senate elections in Colorado. The average Democratic vote percentage during midterm elections was 42.6%, during Presidential election years it was 45.825%. So we can estimate that during an average Presidential election year the Democrat should see a 3% bounce from the higher voter turnout.

As for the polling. There seems to be a consistent polling error of about 7 % against the Republican. In 1998 a Cirulli poll showed Campbell (R) beating Lamm (D) 54%-33%. 2002 showed a similar pattern, the average of two Zogby polls and a USA Today/Gallup poll showed Allard (R) getting 45.6% of the vote and Strickland winning with 47.6%. On election night Allard won with 51% to Strickland’s 46%.

As such, we simply adjust the poll numbers to give the GOP candidate a 4% bump, this should correct for the polling error seen in the last two elections (which were both midterm elections) and the increased voter turnout.

This formula is the most efficient one we have now, however, it is subject to change before election day.

For our calculations we used the three most recent polls, which you can access here.