All political eyes are focused on the U.S. Senate Race here in Washington state. Will 3-term Patty Murray–a Democrat–be beaten by the two-time Gubernatorial loser Republican Dino Rossi?
Friday, October 08, 2010– www.rasmussenreports.com leads with this header:
Washington Senate: Rossi (R) Inches Ahead of Murray (D)
First line: Republican challenger Dino Rossi has edged slightly ahead of incumbent Democrat Patty Murray in Washington’s U.S. Senate race.
They provide their data:
A new Rasmussen Reports telephone survey of Likely Voters in Washington shows Rossi with 49% of the vote, while Murray’s support stands at 46%. Three percent (3%) prefer some other candidate in the race, and two percent (2%) are undecided.
Unfortunately, their poll has a 4 percent margin of error. This means that there needs to be more than a 4 percent difference in order to draw a statistically sound conclusion that one is actually ahead (and there is still a 5% chance of being wrong). These polling results are too close to call–and that is what should have been reported if Rasmussen were being honest about this.
Of course, Murray should be nervous that she is not doing better in the polls but Rossi cannot assume that he is, in fact, ahead, based on 750 “likely” voters.
From a research perspective, there are a couple of things we would need to know. One is how they define “likely.” The second thing is what percent of those participating were Republicans and Democrats. The third thing is whether they tried to weight the data to balance the percent Republicans and Democrats who answered their survey with the distribution of Republicans and Democrats in the population. And if they did, exactly how did they do that? It is very likely that some error was built into this weighting.
Lastly, given Washington politics, it would be important to know whether the respondents were representative of the east-west divide; the east side of Washington tends to be more Republican than the West side, especially the King County population center.
While we all want to know how the elections are likely to go, there is the risk that polls can be distorted to create a spin that might influence actual voter behavior.
It is important to ask questions of research results.
Before I leave this–I came across Nate Silver’s blog–which confirms my hunch that there might be a problem with the weighting of the sample results. He reports:
“Rasmussen Reports has had some problems in Washington in the past, overestimating the performance of the Republican candidate by an average of 4 or 5 points since 2000.”
See his data at: http://fivethirtyeight.blogs.nytimes.com/2010/10/13/mail-voting-in-washington-state-may-confound-pollsters/#more-2127
One more: The Elway Poll:
Murray Firming Up Lead Over Rossi, Tops 50%
“week before ballots start arriving in mailboxes and days before their first debate, Democrat incumbent Patty Murray appears to be solidifying her lead over Republican challenger Dino Rossi. In the Elway Poll of 450 likely voters completed last night, Murray led Rossi by a margin of 51% to 38%, with 11% still undecided. When those 11% were asked toward which candidate they were “leaning,” an additional 4% said Murray and 2% said Rossi, giving Murray a 55‐40% lead. One month ago, Murray led by a 50‐41% margin. Murray has gained in each of the last three Elway Polls, from 47% to 50% to 55%, while Rossi has stalled at 40‐41%.”
http://www.scribd.com/doc/39223757/Elway-Poll-OCT12-Senate
Still–the same questions apply. 450 voters means a higher sampling error of 5 percent. If the difference is more than 10 percent, the results are significant.
Another article appeared today warning about election polling results that leave out cell phone users:
Published on Tuesday, October 19, 2010 by McClatchy Newspapers
Beware of Polls That Exclude Cell Phone-Only Voters
by Steven Thomma
WASHINGTON — Watching the polls to figure out who’s up and who’s down this election season? Be careful. The poll may have a pro-Republican bias.
The ranks of Americans who use cell phones only have skyrocketed. Some public polls don’t survey them, however, and they miss a group of people who are more likely to vote Democratic, including the young, the poor, Hispanics and African-Americans.
The nonpartisan Pew Research Center recently found that polls that contact only those with land-line phones gave Republicans a 4 to 6 percentage point edge over Democrats compared with polls that included cell phones, in four out of five national polls this year.
See source document: http://pewresearch.org/pubs/1761/cell-phones-and-election-polls-2010-midterm-elections