I've heard various claims over the past few years that poll results are consistently biased toward the Democrats.
This market resolves as follows. At noon US Pacific time the Sunday (Nov. 6) before election day 2022, I will look at "Average Seats Controlled by Party" on the FiveThirtyEight House forecast.
https://projects.fivethirtyeight.com/2022-election-forecast/house/
After the election (and after mainstream US media present the results as final), if more Republicans have seats in the House than FiveThirtyEight predicted, this market will resolve to YES. If the number of Republicans in the House is less than or equal to the FiveThirtyEight prediction, the market will resolve to NO.
The market closes the morning of Sunday Nov. 6.
(If FiveThirtyEight does not have an "Average Seats" prediction as of that date, I will use my best judgement to extract a similar prediction from FiveThirtyEight. In the unlikely event that FiveThirtyEight predictions are unavailable, this market resolves to N/A.)
Related questions
๐ Top traders
| # | Name | Total profit |
|---|---|---|
| 1 | แน292 | |
| 2 | แน131 | |
| 3 | แน16 | |
| 4 | แน11 | |
| 5 | แน11 |
I'm going to resolve NO in the next day or two, unless someone convinces me I should wait.
I don't follow things closely enough to know how final results are. However:
Fox shows 220R - 213D (2 undecided)
CNN shows 221R - 213D (1 undecided)
Wikipedia shows 220R - 213D (2 undecided)
The question was will the Rs win at least 230. At this point it seems the answer is pretty clearly NO.
The prediction now shows 229. (It's half an hour early; if 538's prediction changes in the next half hour, feel free to post and I'll take that into account.)
I may delay resolution until I'm confident in the outcome. Elections have been called wrong in the past
https://en.wikipedia.org/wiki/Dewey_Defeats_Truman
and it sometimes happens that decisive results aren't available for some time. (Slow counting, mail-in votes, recounts, lawsuits, etc.) I'd rather delay resolution that resolve wrong.
@Boklam And (1 pm California time) the prediction is still 229.
So, this resolves YES if 230 or more Rs end up in the House, and NO if 229 or fewer.
While polls often do undercount the support on the right, FiveThirtyEight does make some standard adjustments to the raw data in an effort to remove that (among many other) structural bias. I figure the best outside view here is how often they've done this in the past, and I'm currently trying to figure that out. Sadly, they don't keep a quick index of their previous predictions versus reality around, which is telling about their real commitment to accuracy, I think.
Anyway, the probability distribution (their bottom-most graphic) will tell you their predicted most likely outcome (currently 222 seats for the red team). Theoretically it should be a 50/50 split on above or below that result if they've calibrated correctly, so that seems like the best choice for the market zero to me.
@AndrewHartman See this article with an analysis of polling error over the decades
@AndrewHartman As for "a quick index of their previous predictions", a simple google search for "538 Calibration" would have found this article.
