Polling of all sorts always contains a margin of error, usually about 3 percent, though it pays to check each poll. Of course, some pollsters can at times be just plain wrong or, worse, even biased. This should make readers wary of their predictions. Since some online comments in response to more than one of my blogs had touted Rasmussen Reports as a reliable source, I’m passing along what Nate Silver of FiveThirtyEight reported on his election night blog for the New York Times. It's something to keep in mind during the next election campaign:
Rasmussen Reports Polls Were Biased.
While waiting for the remaining results to trickle in from states like Colorado and Alaska, I did a quick check on the accuracy of polls from the firm Rasmussen Reports, which came under heavy criticism this year — including from FiveThirtyEight — because its polls showed a strong lean toward Republican candidates.
Indeed, Rasmussen polls quite consistently turned out to overstate the standing of Republicans tonight. Of the roughly 100 polls released by Rasmussen or its subsidiary Pulse Opinion Research in the final 21 days of the campaign, roughly 70 to 75 percent overestimated the performance of Republican candidates, and on average they were biased against Democrats by 3 to 4 points.
Every pollster is entitled to a bad cycle now and again — and Rasmussen has had some good cycles in the past. But their polling took a major downturn this year.
It is also hard to predict the effect of street money that is used to get out the vote. I am sure the model will be adjusted for next time. Their profits depend upon accuracy.
I find it curious how consistent the American authors are in their assessment of the current political situation. Could we possibly use the word ''bias'' as a descriptor for this behavior. No, no one would ever suspect such a thing on a Catholic site. It stimulates my interest in this blog to see what people say here.
Look at the context of Ms. Smith's introduction, and you'll see that the bias she refers to is not of the mathematical nature that you describe. I stand corrected on the mathematical use of the term in the linked comment.
If the models that Rasmussen used are not able to tweak out the different pressures that leads someone to first vote and then to vote in a certain way, then that is something to look at. But what I also find interesting is the readiness of people to pounce. And that is a different form of bias from statistical bias and statistical bias was never the focus of the original post. Rasmussen is only a factor as long as his organization gets things right and he cannot afford to get things wrong. Too many people are interested in accurate results to pay money for bogus information and to use a favorite expression, all the politiciansl have their own ''internal polls.''
Interestinly the the Gallup poll issued a warning earlier this week to the effect that their polling was showing perameters not encountered before in their long history of polling. They warned that hte usual polling models would likely be thrown off. This elction was deep in uncharted territory.
And sure enough when the acual results were in historic wins of record magnitude for Republicans were recorded all over the place. Did you see some of these results? If anything the Republican wins in the mid mid-west were indeed jugh increase margins of 20% or more from just two years ago. If there was a bias it was one of grossly understating the magnitude of Republican winning. Biut overall the Republicans in fact did unusually well.
The bias here is the denial of the fact the 2010 election was indeed a wave election that did radically revise the political composition of the House of Representative and demonstrated a widespread and intense dissatifaticion with the policies aof congress and the President in historic levels of intensity that noone could predict.
As usuual a more calm and realistic assessment of the election is needed, without making moral issues out of technical issues. Why invent moral problems that do not exist to make a political point?