October 27, 2022

Me vs RCP on Generic Congressional Ballot

The RealClear Politics summary of polls on the generic congressional ballot doesn't work for me.  It weighs polls of all voters or registered voters or likely voters equally.  These are not apples to apples polls.  At this point Likely Voters are the best guage of poll results.  That assumes that pollsters' assignment of likely voters is sound (regardless of whether they are self-identified or determined through other methods).

Polls are not all the same size either.  Does it make sense to aggregate a poll of 500 people with a poll of 7,000 people by giving both polls equal weight? I don't believe it does and in the RCP Average, they do not show margin of error (which is higher with smaller sample sizes), which could be used as a filter for less robust polls.

More recent polls are more telling than older polls.  RCP does account for this by excluding polls over about 2 weeks old. RCP also removes redundant polls where for example a pollster polls twice in the two week period, the older poll gets excluded.

With that in mind I have scraped the RCP polling data and stripped out the duplicates, stripped out polls with no sample size provided (as those two are dubious since they cannot be questioned as to margin of error), stripped out polls that are not of likely voters, and then calculated a weighted average scoring of the total polled across all similar style polls and provided a margin based on that.

Here is what the latest week looks like:

That's only one poll though.  Here's the prior week. 

Better data, but also a bit older.  Here's the last two weeks combined:

I'm seeing a stronger Republican advantage than RCP. I usually do. But it jibes with the sense I'm getting about momentum.  As more recent polls get added I think the change from the prior week above to the current week will solidify more. I expect by early next week we may see enough data to see R+4% looking at just the latest week.

Addendum - just as I finished this about an hour before publication, I see RCP has added a new poll with R+4%.  It doesn't change my overall percentages in the final graph above but supports my theory that +4% seems closer than the +2.3% they are currently showing as their average.

No comments:

Post a Comment

Disagreement is always welcome. Please remain civil. Vulgar or disrespectful comments towards anyone will be removed.

Related Posts Plugin for WordPress, Blogger...

Share This