Final was R +3. Within +/-3 the whole time. Gradually becoming more D ward from 6 to 2.5. Average August through October pretty much exactly where it ended up, which is freaking amazing.
DSeid talks to DSeid: “Ya know DSeid I’m seeing zero evidence of GOP skew? How about you?” “Indubitably DSeid! None.”
Seriously a hypothesis of R skewing that somehow disappeared in the last weeks would not be pretty much on the money most of the cycle getting most D leaning towards the end before bumping a bit back R ward last two or three weeks hitting numbers historically very much on the money.
I’m not saying there was a skew, but if there were pollsters purposely reporting things more R-leaning than reality in the summer and they gradually stopped doing it until October when no one was doing it, we would expect a time series exactly like that.
There are a lot of things that could have caused that, of course, but there is no way you can look at that graph and say it definitely didn’t happen.
“Polls moved from R+6.5 in early June to R+2.4 in early October, so, as we can all see, there’s no possible way dishonest pollsters were moving the poll average rightward in summer, but had stopped doing it closer to the election,” is a ridiculous position.
So the hypothesis is that they skewed as hard as they could in June and gradually eased up, to the point of starting to skew ever so slightly to the D side by October, until they stopped skewing, with polls then moving more in favor for GOP wins, to give historically excellent results in the last three weeks. Okay!
This hypothesis is consistent with the data. I’m not saying that this happened, I’m saying we can not reject this hypothesis based on this data.
If you disagree, answer this question: What would the graph of averaged polling from June to October look like if a handful of pollsters were skewing things in the same direction in June and gradually stopped doing it until they stopped completely by October, all other things being equal?
Oy. My task is not to prove anything but to demonstrate the lack of evidence.
The results in the last few weeks were by historical standards very good. But maybe they were skewing the months before? The two months before ran more D ward. Well maybe they skewed in June and then stopped?
Answer me why there would be a GOP interest in creating that narrative of doing best in June and gradually losing ground?
The farthest the aggregated polls were from the end aggregated poll result was three points. Most of the most active period, August and September, averaged right where the actual vote landed.
I’ve got good news and bad news. The good news is you’ve accomplished the first part. You haven’t proved anything, nor has anyone asked you to.
The bad news is that you’ve failed to do the second part. You haven’t demonstrated a lack of evidence. You brought evidence that is consistent with the hypothesis and declared that you demonstrated a lack of evidence. That’s not the same thing.
No one, as far as I’ve seen, has said the GOP is involved.
I usually find it annoying when people sign their posts, but this time I’m cool with it.
Extraordinary claims require more definitive evidence, not just facts that fail to contradict them.
The kind of pervasive (yet oddly time-period-confined) “skewing” that’s being claimed isn’t super extraordinary, but it’s prima facie less likely than the explanations DSeid provided.
Premise: aggregations that are within historical norms, let alone very close to end results are inconsistent with significant skewing.
Finding: throughout the cycle aggregate polling was better than historical norms, pretty much right on target of the end results for the August through September period.
This may have been addressed already, but I don’t get the motivation for this. If Harris is truly 1 or 2 percent ahead, how does it help the other side to make it sound like she’s 1 or 2 percent behind?
The question you should be asking is, “Why would someone who tells lies for money tell lies?”
And the answer is: For money.
A dishonest pollster can freely put out whatever poll results they want in the summer and early fall because those polls are not checked by anyone for accuracy. They could publish results that are right in line with the honest pollsters, or they could publish something that will get a ton of airplay on Fox News, Newsmax, and OAN. If they are cool with telling lies to make money, which option is more appealing?
Whether or not there is “flooding” going on is not the question. The question is if the aggregations are somehow “skewed” by such, assuming such occurred to significance. The Hopium gang states “folks should continue to be skeptical, given what happened in 2022 … polling averages. They were gamed by a flood of right-wing polls in 2022” (Please note that they are not claiming something that happened only in June then disappeared as the race heated up in August and September.)
I am looking at “what happened in 2022”: the aggregations were on the money. There may have been an attempt to game the system but to the degree that the aggregations included those polls in the mix they may have only made it more accurate.
Yes be skeptical of RCP, which does averages alone.
As “skew” is being used by those of us who are politically interested and not statisticians, it exactly is. We are not meaning the mathematical skewness coefficient; we are interested in whether or not the number is shifted off.
That’s simply an incorrect way to look at time series. It is not meaningful to just look at the end result and compare everything to that. Time series change over time and this particular time series should change near the end as the proportion of undecided voters decreases. Even if the undecided voters break perfectly proportionally to each party the margin between the two parties will tend to increase.
You are assuming perfect polling would be constant over time and trying to determine skew over time by measuring deviation from the baseline, but even if this was a good idea (it isn’t) you are using the wrong baseline.
Most pollsters lose money on political polling. It is a loss leader to burnish their brand. Some are hired by campaigns to collect confidential data. These few are apparently hired by campaigns to attempt to create a narrative. Of course they do it for the money, but why are the campaigns paying them to do it?
The “skew” story goes that these GOP hired shill pollsters do not have a brand otherwise, that sells things like consumer product polling that brings in the money, nor are they selling news services otherwise like CNN or NYT.
These specific ones seem to disappear between cycles and allegedly exist only to try to create these narratives that candidates are doing better than they actually are.