Steven den Beste has looked at a graph of polls of voting intentions and decided:
In September, I think there was a deliberate attempt to depress Kerry's numbers, so as to set up an "October comeback". Of course, the goal was to engineer a bandwagon.
This seems rather implausible to me. There are very many organizations do the polling. If all the polls are rigged, a huge number of people would have to know about it, and surely one of them would have leaked the information by now. It also seems unlikely that every single poller (including, for example, Fox News) would happen to support Kerry.
The reason that den Beste believes that the polls are rigged is:
The data for September, however, is clearly an anomaly. The data is much too consistent. Compare the amount of jitter present before September to the data during that month. There's no period before that of comparable length where the data was so stable.
Well, if the results in September were much too consistent that would certainly be evidence that the results were fixed, but are they? The pink line in the graph to the left is a detail from den Beste's graph showing Bush's numbers for September. The pink line is a moving average of the results of six polls each of roughly 1000 people, so I constructed my own graph by simulating the results of polls of that size and assuming that the underlying number was not changing. The result is the green line in the graph to the left. As you can see, it shows about the same amount of variation as the pink line. It is not true that the results are much too consistent---there is no evidence for fraud here.
It is true that there is less variation in September than there was in earlier months, but that could well be because people had not made up their minds and so there really were larger swings in support for each candidate.
Den Beste also draws two trend lines for each candidate. However he seem to have constructed his trends by just eye-balling the graph. This is not a good technique for constructing trends since it is far too easy to find the trend you were expecting or hoping to see. To get trends that are not affected by subjective bias you need to use statistics.
If the pre-debate polls were fixed to create the illusion of a post-debate surge for Kerry, then why did Kerry's supporters spend so much time dissecting the polls to argue that Bush's lead was not very big? Ruy Teixeira at Emerging Democratic Majority was very quick to point out that various polls based their results on non-representative party sampling and questionable "likely voter" models, both of which downplayed Kerry's strength.
Den Beste also draws two trend lines for each candidate. However he seem to have constructed his trends by just eye-balling the graph.
---
Yeah, I figured if I could do it, we weren't talking rocket science level statistics.
Thanks for reminding us. There were wild swings in numbers of DEMS v GOP voters.
Chris Goringe plots his Poll of Polls which does not show any noticable consistency in September polling.
Talk about trying to spin results !! Any idiot can see from the graph that Steve used from RCP that President Bush's polling results have been trending up since June and Kerry's polling has remained relatively flat since March. And considering that the spin and BS out of the DNC's camp is becoming even more hateful since April and the startup of their tactics to disrupt the election in many states by voter fraud the only conclusion that can be drawn is that the results noted by Steve are relatively accurate.
To me it looks like another liberal media bias! Its a good thing we have people like den Beste around to show us the truth </sarcasm>
I can't believe that someone would say that. Of course the main speculation for the first debate was that Bush would win. Counter to the image he presents he can be an effective debater but he was so very bad in the first debate it took everyone by surprise.
So lets see, we have a convention bounce followed by a very poor showing in the debate as one theory. The other is that the vast liberal underground conspiracy is pulling all the strings and even arranged George to have a bad first debate (by fitting him with a miniature receiver then blocking the frequency maybe).
Sigh, where is Occam when you need him!
If there's a gigantic conspiracy to have Kerry win among *all* pollsters (even explicitly Republican groups like Strategic Visions), why not just have Kerry 'winning' by 10 points, rather than portraying a deadlocked election going down to the wire?
If the polls are rigged to put Kerry in the lead, it's a conspiracy for Kerry. If the polls are rigged to put Bush in the lead- guess what, it's also a conspiracy for Kerry. I have no doubt that den Beste could construct a conspiracy where unrigged polls also demonstrate an underlying Kerry bias.
den Beste is a sad, sad creature. Whatever data her finds are merely fuel for his paranoia.
Wu
Wu
As I blogged recently, I think the data probably are less noisy in September than earlier because *more polls were taken as we get closer to the election*.
My guess is that if data for Sept is based on averages of 6 polls, the data for July was based on averages of maybe 3 polls.
RCP doesn't report their methodology for choosing how many polls to average (at least I couldn't find it). But if you look at RCP's averages of state polls, you find the average is based on anywhere from 2 to 6 polls, depending on how many have been taken recently.
Is it just me, or do his trend lines start in different places? Just eyeballing the trend lines, it looks to me like if he'd started the Bush trend line at the same place he started the Kerry one, the Bush line would be much less steep.
In the last sentence, you mean "affected" not "effected".
Oops. Fixed.
Very true. He also cherry-picked his data points to make the slope of Bush's line more steep. The dates picked for the Bush line are different than the dates picked from Kerry's line.