After looking at the raw data from the DailyKos/PPP poll released last week, we believe that the recall election for Alberta Darling’s senate seat is tighter than the reported 8 point spread (52-44).
As we noted earlier, prior to re-weighting (in the raw sample), Darling had 51.2% of the responses while a generic Democrat had 44.6%. PPP then re-weighted (presumably based on the demographics of the poll sample) and arrived at the reported spread. It appears to us, however, that there are actually proportionately more Darling voters in the poll, and proper re-weighting would have narrowed her gap from the 6.6% point edge she had in the raw data.
Using results from recent elections in Darling’s senate district and a close look at the poll, we make the following observations about the poll demographics.
- Under representation of Milwaukee County Voters. Voting patterns in district 8 are strongly influenced by geography. Voters in Milwaukee County tend to be more Democratic while voters from outside the county tend to be more Republican. In 2008 Darling received 36% of the vote in Milwaukee county, and the county made up nearly half of the electorate (49.4%). In 2010, a lower turnout election for Democrats, the county made up slightly less of the district 8 electorate (48.3%), and Walker received 38.3% of the district 8 vote in Milwaukee county. In the poll, however, only 44.9% of the responses were from in Milwaukee county. Thus, the representation of Milwaukee county in the poll was less than it was even in the low Democratic turnout election of 2010.
- Under representation of young voters. Only 3% of the polls responses were in the 18-29 age demographic. While we could not find a reliable estimate for the percentage of young voters in this district, it is certainly larger than 3% (nationally, 18-29 year olds made up 19% of the electorate in 2008. If anyone knows where we can find information on the demographics of the district, please contact us). This is probably due to the cell-phone effect (the poll does not include cell phones). Almost no UW-Milwaukee students living off campus have landlines. Young voters, especially those attending college, tend to vote Democratic.
- Biased representation of African Americans. In the polled sample 25% of the African Americans were outside Milwaukee County. We’re not sure what the actual percentage is, but 25% is waaay too high. This is important because out-of-county African Americans supported Darling at a much higher rate (42%) in the poll than those in the county (13%). Again, this is probably due to the cell-phone effect. African Americans in Milwaukee county are much less wealthy and more likely to not have a landline than those from outside the county.
- Over representation of Walker voters Ignoring the “other / don’t recall” responses, Walker voters make up 55.5% of the polled sample. This is only about a point higher than Walker’s performance in 2010 (in district 8). However, given the enthusiasm of Democrats around the recall, it’s hard to imagine many Barrett voters sitting out the recall election.
Taken together then, two of the most under-represented groups in the poll are two of the strongest supporters of Democrats, young voters and African Americans living in Milwaukee county. So, what support should we conclude Darling actually has? If we re-weight only for county residence using the breakdown from 2008 election, it is a four point race (50-46). While 2008 was a high turnout election for Democrats, this re-weighting does not fully correct for the under-representation of young voters or African Americans.
To summarize, it is safe to conclude 1) the race in the recall election is closer than the 8 point gap reported, and 2) turnout will be a huge factor in the election. Since voters that sign the recall petition are more likely to vote in the election, it is essential for us to collect as many recall signatures as possible. Further, we should take advantage of the signature gathering stage to engage voters in demographics that tend to vote Democratic, but frequently don’t vote. Even if the signature yield in these groups are not as high as elsewhere, the effects on election day will be worth the effort.